ip_output.c revision 97074
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 97074 2002-05-21 18:52:24Z arr $ 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#include <netinet/ip_dummynet.h> 82 83#ifdef IPFIREWALL_FORWARD_DEBUG 84#define print_ip(a) printf("%ld.%ld.%ld.%ld",(ntohl(a.s_addr)>>24)&0xFF,\ 85 (ntohl(a.s_addr)>>16)&0xFF,\ 86 (ntohl(a.s_addr)>>8)&0xFF,\ 87 (ntohl(a.s_addr))&0xFF); 88#endif 89 90u_short ip_id; 91 92static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 93static struct ifnet *ip_multicast_if(struct in_addr *, int *); 94static void ip_mloopback 95 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 96static int ip_getmoptions 97 (struct sockopt *, struct ip_moptions *); 98static int ip_pcbopts(int, struct mbuf **, struct mbuf *); 99static int ip_setmoptions 100 (struct sockopt *, struct ip_moptions **); 101 102int ip_optcopy(struct ip *, struct ip *); 103 104 105extern struct protosw inetsw[]; 106 107/* 108 * IP output. The packet in mbuf chain m contains a skeletal IP 109 * header (with len, off, ttl, proto, tos, src, dst). 110 * The mbuf chain containing the packet will be freed. 111 * The mbuf opt, if present, will not be freed. 112 */ 113int 114ip_output(m0, opt, ro, flags, imo) 115 struct mbuf *m0; 116 struct mbuf *opt; 117 struct route *ro; 118 int flags; 119 struct ip_moptions *imo; 120{ 121 struct ip *ip, *mhip; 122 struct ifnet *ifp; 123 struct mbuf *m = m0; 124 int hlen = sizeof (struct ip); 125 int len, off, error = 0; 126 struct sockaddr_in *dst; 127 struct in_ifaddr *ia; 128 int isbroadcast, sw_csum; 129 struct in_addr pkt_dst; 130#ifdef IPSEC 131 struct route iproute; 132 struct socket *so = NULL; 133 struct secpolicy *sp = NULL; 134#endif 135 u_int16_t divert_cookie; /* firewall cookie */ 136#ifdef PFIL_HOOKS 137 struct packet_filter_hook *pfh; 138 struct mbuf *m1; 139 int rv; 140#endif /* PFIL_HOOKS */ 141#ifdef IPFIREWALL_FORWARD 142 int fwd_rewrite_src = 0; 143#endif 144 struct ip_fw *rule = NULL; 145 146#ifdef IPDIVERT 147 /* Get and reset firewall cookie */ 148 divert_cookie = ip_divert_cookie; 149 ip_divert_cookie = 0; 150#else 151 divert_cookie = 0; 152#endif 153 154 /* 155 * dummynet packet are prepended a vestigial mbuf with 156 * m_type = MT_DUMMYNET and m_data pointing to the matching 157 * rule. 158 */ 159 if (m->m_type == MT_DUMMYNET) { 160 /* 161 * the packet was already tagged, so part of the 162 * processing was already done, and we need to go down. 163 * Get parameters from the header. 164 */ 165 rule = (struct ip_fw *)(m->m_data) ; 166 opt = NULL ; 167 ro = & ( ((struct dn_pkt *)m)->ro ) ; 168 imo = NULL ; 169 dst = ((struct dn_pkt *)m)->dn_dst ; 170 ifp = ((struct dn_pkt *)m)->ifp ; 171 flags = ((struct dn_pkt *)m)->flags ; 172 173 m0 = m = m->m_next ; 174#ifdef IPSEC 175 so = ipsec_getsocket(m); 176 (void)ipsec_setsocket(m, NULL); 177#endif 178 ip = mtod(m, struct ip *); 179 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ; 180 ia = ifatoia(ro->ro_rt->rt_ifa); 181 goto sendit; 182 } else 183 rule = NULL ; 184#ifdef IPSEC 185 so = ipsec_getsocket(m); 186 (void)ipsec_setsocket(m, NULL); 187#endif 188 189 KASSERT((m->m_flags & M_PKTHDR) != 0, ("ip_output: no HDR")); 190 191 KASSERT(ro != NULL, ("ip_output: no route, proto %d", 192 mtod(m, struct ip *)->ip_p)); 193 194 if (opt) { 195 m = ip_insertoptions(m, opt, &len); 196 hlen = len; 197 } 198 ip = mtod(m, struct ip *); 199 pkt_dst = ip_fw_fwd_addr == NULL 200 ? ip->ip_dst : ip_fw_fwd_addr->sin_addr; 201 202 /* 203 * Fill in IP header. 204 */ 205 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 206 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2); 207 ip->ip_off &= IP_DF; 208#ifdef RANDOM_IP_ID 209 ip->ip_id = ip_randomid(); 210#else 211 ip->ip_id = htons(ip_id++); 212#endif 213 ipstat.ips_localout++; 214 } else { 215 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 216 } 217 218 dst = (struct sockaddr_in *)&ro->ro_dst; 219 /* 220 * If there is a cached route, 221 * check that it is to the same destination 222 * and is still up. If not, free it and try again. 223 * The address family should also be checked in case of sharing the 224 * cache with IPv6. 225 */ 226 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 227 dst->sin_family != AF_INET || 228 dst->sin_addr.s_addr != pkt_dst.s_addr)) { 229 RTFREE(ro->ro_rt); 230 ro->ro_rt = (struct rtentry *)0; 231 } 232 if (ro->ro_rt == 0) { 233 bzero(dst, sizeof(*dst)); 234 dst->sin_family = AF_INET; 235 dst->sin_len = sizeof(*dst); 236 dst->sin_addr = pkt_dst; 237 } 238 /* 239 * If routing to interface only, 240 * short circuit routing lookup. 241 */ 242 if (flags & IP_ROUTETOIF) { 243 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 244 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 245 ipstat.ips_noroute++; 246 error = ENETUNREACH; 247 goto bad; 248 } 249 ifp = ia->ia_ifp; 250 ip->ip_ttl = 1; 251 isbroadcast = in_broadcast(dst->sin_addr, ifp); 252 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 253 imo != NULL && imo->imo_multicast_ifp != NULL) { 254 /* 255 * Bypass the normal routing lookup for multicast 256 * packets if the interface is specified. 257 */ 258 ifp = imo->imo_multicast_ifp; 259 IFP_TO_IA(ifp, ia); 260 isbroadcast = 0; /* fool gcc */ 261 } else { 262 /* 263 * If this is the case, we probably don't want to allocate 264 * a protocol-cloned route since we didn't get one from the 265 * ULP. This lets TCP do its thing, while not burdening 266 * forwarding or ICMP with the overhead of cloning a route. 267 * Of course, we still want to do any cloning requested by 268 * the link layer, as this is probably required in all cases 269 * for correct operation (as it is for ARP). 270 */ 271 if (ro->ro_rt == 0) 272 rtalloc_ign(ro, RTF_PRCLONING); 273 if (ro->ro_rt == 0) { 274 ipstat.ips_noroute++; 275 error = EHOSTUNREACH; 276 goto bad; 277 } 278 ia = ifatoia(ro->ro_rt->rt_ifa); 279 ifp = ro->ro_rt->rt_ifp; 280 ro->ro_rt->rt_use++; 281 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 282 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 283 if (ro->ro_rt->rt_flags & RTF_HOST) 284 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 285 else 286 isbroadcast = in_broadcast(dst->sin_addr, ifp); 287 } 288 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) { 289 struct in_multi *inm; 290 291 m->m_flags |= M_MCAST; 292 /* 293 * IP destination address is multicast. Make sure "dst" 294 * still points to the address in "ro". (It may have been 295 * changed to point to a gateway address, above.) 296 */ 297 dst = (struct sockaddr_in *)&ro->ro_dst; 298 /* 299 * See if the caller provided any multicast options 300 */ 301 if (imo != NULL) { 302 ip->ip_ttl = imo->imo_multicast_ttl; 303 if (imo->imo_multicast_vif != -1) 304 ip->ip_src.s_addr = 305 ip_mcast_src(imo->imo_multicast_vif); 306 } else 307 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 308 /* 309 * Confirm that the outgoing interface supports multicast. 310 */ 311 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 312 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 313 ipstat.ips_noroute++; 314 error = ENETUNREACH; 315 goto bad; 316 } 317 } 318 /* 319 * If source address not specified yet, use address 320 * of outgoing interface. 321 */ 322 if (ip->ip_src.s_addr == INADDR_ANY) { 323 /* Interface may have no addresses. */ 324 if (ia != NULL) 325 ip->ip_src = IA_SIN(ia)->sin_addr; 326 } 327 328 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 329 /* 330 * XXX 331 * delayed checksums are not currently 332 * compatible with IP multicast routing 333 */ 334 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 335 in_delayed_cksum(m); 336 m->m_pkthdr.csum_flags &= 337 ~CSUM_DELAY_DATA; 338 } 339 } 340 IN_LOOKUP_MULTI(pkt_dst, ifp, inm); 341 if (inm != NULL && 342 (imo == NULL || imo->imo_multicast_loop)) { 343 /* 344 * If we belong to the destination multicast group 345 * on the outgoing interface, and the caller did not 346 * forbid loopback, loop back a copy. 347 */ 348 ip_mloopback(ifp, m, dst, hlen); 349 } 350 else { 351 /* 352 * If we are acting as a multicast router, perform 353 * multicast forwarding as if the packet had just 354 * arrived on the interface to which we are about 355 * to send. The multicast forwarding function 356 * recursively calls this function, using the 357 * IP_FORWARDING flag to prevent infinite recursion. 358 * 359 * Multicasts that are looped back by ip_mloopback(), 360 * above, will be forwarded by the ip_input() routine, 361 * if necessary. 362 */ 363 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 364 /* 365 * Check if rsvp daemon is running. If not, don't 366 * set ip_moptions. This ensures that the packet 367 * is multicast and not just sent down one link 368 * as prescribed by rsvpd. 369 */ 370 if (!rsvp_on) 371 imo = NULL; 372 if (ip_mforward(ip, ifp, m, imo) != 0) { 373 m_freem(m); 374 goto done; 375 } 376 } 377 } 378 379 /* 380 * Multicasts with a time-to-live of zero may be looped- 381 * back, above, but must not be transmitted on a network. 382 * Also, multicasts addressed to the loopback interface 383 * are not sent -- the above call to ip_mloopback() will 384 * loop back a copy if this host actually belongs to the 385 * destination group on the loopback interface. 386 */ 387 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 388 m_freem(m); 389 goto done; 390 } 391 392 goto sendit; 393 } 394#ifndef notdef 395 /* 396 * If source address not specified yet, use address 397 * of outgoing interface. 398 */ 399 if (ip->ip_src.s_addr == INADDR_ANY) { 400 /* Interface may have no addresses. */ 401 if (ia != NULL) { 402 ip->ip_src = IA_SIN(ia)->sin_addr; 403#ifdef IPFIREWALL_FORWARD 404 /* Keep note that we did this - if the firewall changes 405 * the next-hop, our interface may change, changing the 406 * default source IP. It's a shame so much effort happens 407 * twice. Oh well. 408 */ 409 fwd_rewrite_src++; 410#endif /* IPFIREWALL_FORWARD */ 411 } 412 } 413#endif /* notdef */ 414 /* 415 * Verify that we have any chance at all of being able to queue 416 * the packet or packet fragments 417 */ 418 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 419 ifp->if_snd.ifq_maxlen) { 420 error = ENOBUFS; 421 ipstat.ips_odropped++; 422 goto bad; 423 } 424 425 /* 426 * Look for broadcast address and 427 * verify user is allowed to send 428 * such a packet. 429 */ 430 if (isbroadcast) { 431 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 432 error = EADDRNOTAVAIL; 433 goto bad; 434 } 435 if ((flags & IP_ALLOWBROADCAST) == 0) { 436 error = EACCES; 437 goto bad; 438 } 439 /* don't allow broadcast messages to be fragmented */ 440 if ((u_short)ip->ip_len > ifp->if_mtu) { 441 error = EMSGSIZE; 442 goto bad; 443 } 444 m->m_flags |= M_BCAST; 445 } else { 446 m->m_flags &= ~M_BCAST; 447 } 448 449sendit: 450#ifdef IPSEC 451 /* get SP for this packet */ 452 if (so == NULL) 453 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 454 else 455 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 456 457 if (sp == NULL) { 458 ipsecstat.out_inval++; 459 goto bad; 460 } 461 462 error = 0; 463 464 /* check policy */ 465 switch (sp->policy) { 466 case IPSEC_POLICY_DISCARD: 467 /* 468 * This packet is just discarded. 469 */ 470 ipsecstat.out_polvio++; 471 goto bad; 472 473 case IPSEC_POLICY_BYPASS: 474 case IPSEC_POLICY_NONE: 475 /* no need to do IPsec. */ 476 goto skip_ipsec; 477 478 case IPSEC_POLICY_IPSEC: 479 if (sp->req == NULL) { 480 /* acquire a policy */ 481 error = key_spdacquire(sp); 482 goto bad; 483 } 484 break; 485 486 case IPSEC_POLICY_ENTRUST: 487 default: 488 printf("ip_output: Invalid policy found. %d\n", sp->policy); 489 } 490 { 491 struct ipsec_output_state state; 492 bzero(&state, sizeof(state)); 493 state.m = m; 494 if (flags & IP_ROUTETOIF) { 495 state.ro = &iproute; 496 bzero(&iproute, sizeof(iproute)); 497 } else 498 state.ro = ro; 499 state.dst = (struct sockaddr *)dst; 500 501 ip->ip_sum = 0; 502 503 /* 504 * XXX 505 * delayed checksums are not currently compatible with IPsec 506 */ 507 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 508 in_delayed_cksum(m); 509 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 510 } 511 512 ip->ip_len = htons(ip->ip_len); 513 ip->ip_off = htons(ip->ip_off); 514 515 error = ipsec4_output(&state, sp, flags); 516 517 m = state.m; 518 if (flags & IP_ROUTETOIF) { 519 /* 520 * if we have tunnel mode SA, we may need to ignore 521 * IP_ROUTETOIF. 522 */ 523 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 524 flags &= ~IP_ROUTETOIF; 525 ro = state.ro; 526 } 527 } else 528 ro = state.ro; 529 dst = (struct sockaddr_in *)state.dst; 530 if (error) { 531 /* mbuf is already reclaimed in ipsec4_output. */ 532 m0 = NULL; 533 switch (error) { 534 case EHOSTUNREACH: 535 case ENETUNREACH: 536 case EMSGSIZE: 537 case ENOBUFS: 538 case ENOMEM: 539 break; 540 default: 541 printf("ip4_output (ipsec): error code %d\n", error); 542 /*fall through*/ 543 case ENOENT: 544 /* don't show these error codes to the user */ 545 error = 0; 546 break; 547 } 548 goto bad; 549 } 550 } 551 552 /* be sure to update variables that are affected by ipsec4_output() */ 553 ip = mtod(m, struct ip *); 554#ifdef _IP_VHL 555 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 556#else 557 hlen = ip->ip_hl << 2; 558#endif 559 if (ro->ro_rt == NULL) { 560 if ((flags & IP_ROUTETOIF) == 0) { 561 printf("ip_output: " 562 "can't update route after IPsec processing\n"); 563 error = EHOSTUNREACH; /*XXX*/ 564 goto bad; 565 } 566 } else { 567 ia = ifatoia(ro->ro_rt->rt_ifa); 568 ifp = ro->ro_rt->rt_ifp; 569 } 570 571 /* make it flipped, again. */ 572 ip->ip_len = ntohs(ip->ip_len); 573 ip->ip_off = ntohs(ip->ip_off); 574skip_ipsec: 575#endif /*IPSEC*/ 576 577 /* 578 * IpHack's section. 579 * - Xlate: translate packet's addr/port (NAT). 580 * - Firewall: deny/allow/etc. 581 * - Wrap: fake packet's addr/port <unimpl.> 582 * - Encapsulate: put it in another IP and send out. <unimp.> 583 */ 584#ifdef PFIL_HOOKS 585 /* 586 * Run through list of hooks for output packets. 587 */ 588 m1 = m; 589 pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 590 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 591 if (pfh->pfil_func) { 592 rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1); 593 if (rv) { 594 error = EHOSTUNREACH; 595 goto done; 596 } 597 m = m1; 598 if (m == NULL) 599 goto done; 600 ip = mtod(m, struct ip *); 601 } 602#endif /* PFIL_HOOKS */ 603 604 /* 605 * Check with the firewall... 606 * but not if we are already being fwd'd from a firewall. 607 */ 608 if (fw_enable && IPFW_LOADED && !ip_fw_fwd_addr) { 609 struct sockaddr_in *old = dst; 610 611 off = ip_fw_chk_ptr(&m, ifp, &divert_cookie, &rule, &dst); 612 /* 613 * On return we must do the following: 614 * m == NULL -> drop the pkt (old interface, deprecated) 615 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 616 * 1<=off<= 0xffff -> DIVERT 617 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 618 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 619 * dst != old -> IPFIREWALL_FORWARD 620 * off==0, dst==old -> accept 621 * If some of the above modules are not compiled in, then 622 * we should't have to check the corresponding condition 623 * (because the ipfw control socket should not accept 624 * unsupported rules), but better play safe and drop 625 * packets in case of doubt. 626 */ 627 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) { 628 if (m) 629 m_freem(m); 630 error = EACCES; 631 goto done; 632 } 633 ip = mtod(m, struct ip *); 634 if (off == 0 && dst == old) /* common case */ 635 goto pass; 636 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 637 /* 638 * pass the pkt to dummynet. Need to include 639 * pipe number, m, ifp, ro, dst because these are 640 * not recomputed in the next pass. 641 * All other parameters have been already used and 642 * so they are not needed anymore. 643 * XXX note: if the ifp or ro entry are deleted 644 * while a pkt is in dummynet, we are in trouble! 645 */ 646 error = ip_dn_io_ptr(off & 0xffff, DN_TO_IP_OUT, m, 647 ifp, ro, dst, rule, flags); 648 goto done; 649 } 650#ifdef IPDIVERT 651 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 652 struct mbuf *clone = NULL; 653 654 /* Clone packet if we're doing a 'tee' */ 655 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 656 clone = m_dup(m, M_DONTWAIT); 657 658 /* 659 * XXX 660 * delayed checksums are not currently compatible 661 * with divert sockets. 662 */ 663 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 664 in_delayed_cksum(m); 665 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 666 } 667 668 /* Restore packet header fields to original values */ 669 ip->ip_len = htons(ip->ip_len); 670 ip->ip_off = htons(ip->ip_off); 671 672 /* Deliver packet to divert input routine */ 673 ip_divert_cookie = divert_cookie; 674 divert_packet(m, 0, off & 0xffff); 675 676 /* If 'tee', continue with original packet */ 677 if (clone != NULL) { 678 m = clone; 679 ip = mtod(m, struct ip *); 680 goto pass; 681 } 682 goto done; 683 } 684#endif 685 686#ifdef IPFIREWALL_FORWARD 687 /* Here we check dst to make sure it's directly reachable on the 688 * interface we previously thought it was. 689 * If it isn't (which may be likely in some situations) we have 690 * to re-route it (ie, find a route for the next-hop and the 691 * associated interface) and set them here. This is nested 692 * forwarding which in most cases is undesirable, except where 693 * such control is nigh impossible. So we do it here. 694 * And I'm babbling. 695 */ 696 if (off == 0 && old != dst) { 697 struct in_ifaddr *ia; 698 699 /* It's changed... */ 700 /* There must be a better way to do this next line... */ 701 static struct route sro_fwd, *ro_fwd = &sro_fwd; 702#ifdef IPFIREWALL_FORWARD_DEBUG 703 printf("IPFIREWALL_FORWARD: New dst ip: "); 704 print_ip(dst->sin_addr); 705 printf("\n"); 706#endif 707 /* 708 * We need to figure out if we have been forwarded 709 * to a local socket. If so then we should somehow 710 * "loop back" to ip_input, and get directed to the 711 * PCB as if we had received this packet. This is 712 * because it may be dificult to identify the packets 713 * you want to forward until they are being output 714 * and have selected an interface. (e.g. locally 715 * initiated packets) If we used the loopback inteface, 716 * we would not be able to control what happens 717 * as the packet runs through ip_input() as 718 * it is done through a ISR. 719 */ 720 LIST_FOREACH(ia, 721 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 722 /* 723 * If the addr to forward to is one 724 * of ours, we pretend to 725 * be the destination for this packet. 726 */ 727 if (IA_SIN(ia)->sin_addr.s_addr == 728 dst->sin_addr.s_addr) 729 break; 730 } 731 if (ia) { 732 /* tell ip_input "dont filter" */ 733 ip_fw_fwd_addr = dst; 734 if (m->m_pkthdr.rcvif == NULL) 735 m->m_pkthdr.rcvif = ifunit("lo0"); 736 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 737 m->m_pkthdr.csum_flags |= 738 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 739 m0->m_pkthdr.csum_data = 0xffff; 740 } 741 m->m_pkthdr.csum_flags |= 742 CSUM_IP_CHECKED | CSUM_IP_VALID; 743 ip->ip_len = htons(ip->ip_len); 744 ip->ip_off = htons(ip->ip_off); 745 ip_input(m); 746 goto done; 747 } 748 /* Some of the logic for this was 749 * nicked from above. 750 * 751 * This rewrites the cached route in a local PCB. 752 * Is this what we want to do? 753 */ 754 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 755 756 ro_fwd->ro_rt = 0; 757 rtalloc_ign(ro_fwd, RTF_PRCLONING); 758 759 if (ro_fwd->ro_rt == 0) { 760 ipstat.ips_noroute++; 761 error = EHOSTUNREACH; 762 goto bad; 763 } 764 765 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 766 ifp = ro_fwd->ro_rt->rt_ifp; 767 ro_fwd->ro_rt->rt_use++; 768 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 769 dst = (struct sockaddr_in *)ro_fwd->ro_rt->rt_gateway; 770 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 771 isbroadcast = 772 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 773 else 774 isbroadcast = in_broadcast(dst->sin_addr, ifp); 775 if (ro->ro_rt) 776 RTFREE(ro->ro_rt); 777 ro->ro_rt = ro_fwd->ro_rt; 778 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 779 780 /* 781 * If we added a default src ip earlier, 782 * which would have been gotten from the-then 783 * interface, do it again, from the new one. 784 */ 785 if (fwd_rewrite_src) 786 ip->ip_src = IA_SIN(ia)->sin_addr; 787 goto pass ; 788 } 789#endif /* IPFIREWALL_FORWARD */ 790 /* 791 * if we get here, none of the above matches, and 792 * we have to drop the pkt 793 */ 794 m_freem(m); 795 error = EACCES; /* not sure this is the right error msg */ 796 goto done; 797 } 798 799 ip_fw_fwd_addr = NULL; 800pass: 801 /* 127/8 must not appear on wire - RFC1122. */ 802 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 803 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 804 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 805 ipstat.ips_badaddr++; 806 error = EADDRNOTAVAIL; 807 goto bad; 808 } 809 } 810 811 m->m_pkthdr.csum_flags |= CSUM_IP; 812 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 813 if (sw_csum & CSUM_DELAY_DATA) { 814 in_delayed_cksum(m); 815 sw_csum &= ~CSUM_DELAY_DATA; 816 } 817 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 818 819 /* 820 * If small enough for interface, or the interface will take 821 * care of the fragmentation for us, can just send directly. 822 */ 823 if ((u_short)ip->ip_len <= ifp->if_mtu || 824 ifp->if_hwassist & CSUM_FRAGMENT) { 825 ip->ip_len = htons(ip->ip_len); 826 ip->ip_off = htons(ip->ip_off); 827 ip->ip_sum = 0; 828 if (sw_csum & CSUM_DELAY_IP) { 829 if (ip->ip_vhl == IP_VHL_BORING) { 830 ip->ip_sum = in_cksum_hdr(ip); 831 } else { 832 ip->ip_sum = in_cksum(m, hlen); 833 } 834 } 835 836 /* Record statistics for this interface address. */ 837 if (!(flags & IP_FORWARDING) && ia) { 838 ia->ia_ifa.if_opackets++; 839 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 840 } 841 842#ifdef IPSEC 843 /* clean ipsec history once it goes out of the node */ 844 ipsec_delaux(m); 845#endif 846 847 error = (*ifp->if_output)(ifp, m, 848 (struct sockaddr *)dst, ro->ro_rt); 849 goto done; 850 } 851 /* 852 * Too large for interface; fragment if possible. 853 * Must be able to put at least 8 bytes per fragment. 854 */ 855 if (ip->ip_off & IP_DF) { 856 error = EMSGSIZE; 857 /* 858 * This case can happen if the user changed the MTU 859 * of an interface after enabling IP on it. Because 860 * most netifs don't keep track of routes pointing to 861 * them, there is no way for one to update all its 862 * routes when the MTU is changed. 863 */ 864 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) 865 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) 866 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 867 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 868 } 869 ipstat.ips_cantfrag++; 870 goto bad; 871 } 872 len = (ifp->if_mtu - hlen) &~ 7; 873 if (len < 8) { 874 error = EMSGSIZE; 875 goto bad; 876 } 877 878 /* 879 * if the interface will not calculate checksums on 880 * fragmented packets, then do it here. 881 */ 882 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 883 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) { 884 in_delayed_cksum(m); 885 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 886 } 887 888 { 889 int mhlen, firstlen = len; 890 struct mbuf **mnext = &m->m_nextpkt; 891 int nfrags = 1; 892 893 /* 894 * Loop through length of segment after first fragment, 895 * make new header and copy data of each part and link onto chain. 896 */ 897 m0 = m; 898 mhlen = sizeof (struct ip); 899 for (off = hlen + len; off < (u_short)ip->ip_len; off += len) { 900 MGETHDR(m, M_DONTWAIT, MT_HEADER); 901 if (m == 0) { 902 error = ENOBUFS; 903 ipstat.ips_odropped++; 904 goto sendorfree; 905 } 906 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 907 m->m_data += max_linkhdr; 908 mhip = mtod(m, struct ip *); 909 *mhip = *ip; 910 if (hlen > sizeof (struct ip)) { 911 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 912 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2); 913 } 914 m->m_len = mhlen; 915 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 916 if (off + len >= (u_short)ip->ip_len) 917 len = (u_short)ip->ip_len - off; 918 else 919 mhip->ip_off |= IP_MF; 920 mhip->ip_len = htons((u_short)(len + mhlen)); 921 m->m_next = m_copy(m0, off, len); 922 if (m->m_next == 0) { 923 (void) m_free(m); 924 error = ENOBUFS; /* ??? */ 925 ipstat.ips_odropped++; 926 goto sendorfree; 927 } 928 m->m_pkthdr.len = mhlen + len; 929 m->m_pkthdr.rcvif = (struct ifnet *)0; 930 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 931 mhip->ip_off = htons(mhip->ip_off); 932 mhip->ip_sum = 0; 933 if (sw_csum & CSUM_DELAY_IP) { 934 if (mhip->ip_vhl == IP_VHL_BORING) { 935 mhip->ip_sum = in_cksum_hdr(mhip); 936 } else { 937 mhip->ip_sum = in_cksum(m, mhlen); 938 } 939 } 940 *mnext = m; 941 mnext = &m->m_nextpkt; 942 nfrags++; 943 } 944 ipstat.ips_ofragments += nfrags; 945 946 /* set first/last markers for fragment chain */ 947 m->m_flags |= M_LASTFRAG; 948 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 949 m0->m_pkthdr.csum_data = nfrags; 950 951 /* 952 * Update first fragment by trimming what's been copied out 953 * and updating header, then send each fragment (in order). 954 */ 955 m = m0; 956 m_adj(m, hlen + firstlen - (u_short)ip->ip_len); 957 m->m_pkthdr.len = hlen + firstlen; 958 ip->ip_len = htons((u_short)m->m_pkthdr.len); 959 ip->ip_off |= IP_MF; 960 ip->ip_off = htons(ip->ip_off); 961 ip->ip_sum = 0; 962 if (sw_csum & CSUM_DELAY_IP) { 963 if (ip->ip_vhl == IP_VHL_BORING) { 964 ip->ip_sum = in_cksum_hdr(ip); 965 } else { 966 ip->ip_sum = in_cksum(m, hlen); 967 } 968 } 969sendorfree: 970 for (m = m0; m; m = m0) { 971 m0 = m->m_nextpkt; 972 m->m_nextpkt = 0; 973#ifdef IPSEC 974 /* clean ipsec history once it goes out of the node */ 975 ipsec_delaux(m); 976#endif 977 if (error == 0) { 978 /* Record statistics for this interface address. */ 979 if (ia != NULL) { 980 ia->ia_ifa.if_opackets++; 981 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 982 } 983 984 error = (*ifp->if_output)(ifp, m, 985 (struct sockaddr *)dst, ro->ro_rt); 986 } else 987 m_freem(m); 988 } 989 990 if (error == 0) 991 ipstat.ips_fragmented++; 992 } 993done: 994#ifdef IPSEC 995 if (ro == &iproute && ro->ro_rt) { 996 RTFREE(ro->ro_rt); 997 ro->ro_rt = NULL; 998 } 999 if (sp != NULL) { 1000 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1001 printf("DP ip_output call free SP:%p\n", sp)); 1002 key_freesp(sp); 1003 } 1004#endif /* IPSEC */ 1005 return (error); 1006bad: 1007 m_freem(m); 1008 goto done; 1009} 1010 1011void 1012in_delayed_cksum(struct mbuf *m) 1013{ 1014 struct ip *ip; 1015 u_short csum, offset; 1016 1017 ip = mtod(m, struct ip *); 1018 offset = IP_VHL_HL(ip->ip_vhl) << 2 ; 1019 csum = in_cksum_skip(m, ip->ip_len, offset); 1020 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1021 csum = 0xffff; 1022 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1023 1024 if (offset + sizeof(u_short) > m->m_len) { 1025 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1026 m->m_len, offset, ip->ip_p); 1027 /* 1028 * XXX 1029 * this shouldn't happen, but if it does, the 1030 * correct behavior may be to insert the checksum 1031 * in the existing chain instead of rearranging it. 1032 */ 1033 m = m_pullup(m, offset + sizeof(u_short)); 1034 } 1035 *(u_short *)(m->m_data + offset) = csum; 1036} 1037 1038/* 1039 * Insert IP options into preformed packet. 1040 * Adjust IP destination as required for IP source routing, 1041 * as indicated by a non-zero in_addr at the start of the options. 1042 * 1043 * XXX This routine assumes that the packet has no options in place. 1044 */ 1045static struct mbuf * 1046ip_insertoptions(m, opt, phlen) 1047 register struct mbuf *m; 1048 struct mbuf *opt; 1049 int *phlen; 1050{ 1051 register struct ipoption *p = mtod(opt, struct ipoption *); 1052 struct mbuf *n; 1053 register struct ip *ip = mtod(m, struct ip *); 1054 unsigned optlen; 1055 1056 optlen = opt->m_len - sizeof(p->ipopt_dst); 1057 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) 1058 return (m); /* XXX should fail */ 1059 if (p->ipopt_dst.s_addr) 1060 ip->ip_dst = p->ipopt_dst; 1061 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1062 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1063 if (n == 0) 1064 return (m); 1065 n->m_pkthdr.rcvif = (struct ifnet *)0; 1066 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1067 m->m_len -= sizeof(struct ip); 1068 m->m_data += sizeof(struct ip); 1069 n->m_next = m; 1070 m = n; 1071 m->m_len = optlen + sizeof(struct ip); 1072 m->m_data += max_linkhdr; 1073 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1074 } else { 1075 m->m_data -= optlen; 1076 m->m_len += optlen; 1077 m->m_pkthdr.len += optlen; 1078 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1079 } 1080 ip = mtod(m, struct ip *); 1081 bcopy(p->ipopt_list, ip + 1, optlen); 1082 *phlen = sizeof(struct ip) + optlen; 1083 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2); 1084 ip->ip_len += optlen; 1085 return (m); 1086} 1087 1088/* 1089 * Copy options from ip to jp, 1090 * omitting those not copied during fragmentation. 1091 */ 1092int 1093ip_optcopy(ip, jp) 1094 struct ip *ip, *jp; 1095{ 1096 register u_char *cp, *dp; 1097 int opt, optlen, cnt; 1098 1099 cp = (u_char *)(ip + 1); 1100 dp = (u_char *)(jp + 1); 1101 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1102 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1103 opt = cp[0]; 1104 if (opt == IPOPT_EOL) 1105 break; 1106 if (opt == IPOPT_NOP) { 1107 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1108 *dp++ = IPOPT_NOP; 1109 optlen = 1; 1110 continue; 1111 } 1112 1113 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1114 ("ip_optcopy: malformed ipv4 option")); 1115 optlen = cp[IPOPT_OLEN]; 1116 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1117 ("ip_optcopy: malformed ipv4 option")); 1118 1119 /* bogus lengths should have been caught by ip_dooptions */ 1120 if (optlen > cnt) 1121 optlen = cnt; 1122 if (IPOPT_COPIED(opt)) { 1123 bcopy(cp, dp, optlen); 1124 dp += optlen; 1125 } 1126 } 1127 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1128 *dp++ = IPOPT_EOL; 1129 return (optlen); 1130} 1131 1132/* 1133 * IP socket option processing. 1134 */ 1135int 1136ip_ctloutput(so, sopt) 1137 struct socket *so; 1138 struct sockopt *sopt; 1139{ 1140 struct inpcb *inp = sotoinpcb(so); 1141 int error, optval; 1142 1143 error = optval = 0; 1144 if (sopt->sopt_level != IPPROTO_IP) { 1145 return (EINVAL); 1146 } 1147 1148 switch (sopt->sopt_dir) { 1149 case SOPT_SET: 1150 switch (sopt->sopt_name) { 1151 case IP_OPTIONS: 1152#ifdef notyet 1153 case IP_RETOPTS: 1154#endif 1155 { 1156 struct mbuf *m; 1157 if (sopt->sopt_valsize > MLEN) { 1158 error = EMSGSIZE; 1159 break; 1160 } 1161 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1162 if (m == 0) { 1163 error = ENOBUFS; 1164 break; 1165 } 1166 m->m_len = sopt->sopt_valsize; 1167 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1168 m->m_len); 1169 1170 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1171 m)); 1172 } 1173 1174 case IP_TOS: 1175 case IP_TTL: 1176 case IP_RECVOPTS: 1177 case IP_RECVRETOPTS: 1178 case IP_RECVDSTADDR: 1179 case IP_RECVIF: 1180 case IP_FAITH: 1181 error = sooptcopyin(sopt, &optval, sizeof optval, 1182 sizeof optval); 1183 if (error) 1184 break; 1185 1186 switch (sopt->sopt_name) { 1187 case IP_TOS: 1188 inp->inp_ip_tos = optval; 1189 break; 1190 1191 case IP_TTL: 1192 inp->inp_ip_ttl = optval; 1193 break; 1194#define OPTSET(bit) \ 1195 if (optval) \ 1196 inp->inp_flags |= bit; \ 1197 else \ 1198 inp->inp_flags &= ~bit; 1199 1200 case IP_RECVOPTS: 1201 OPTSET(INP_RECVOPTS); 1202 break; 1203 1204 case IP_RECVRETOPTS: 1205 OPTSET(INP_RECVRETOPTS); 1206 break; 1207 1208 case IP_RECVDSTADDR: 1209 OPTSET(INP_RECVDSTADDR); 1210 break; 1211 1212 case IP_RECVIF: 1213 OPTSET(INP_RECVIF); 1214 break; 1215 1216 case IP_FAITH: 1217 OPTSET(INP_FAITH); 1218 break; 1219 } 1220 break; 1221#undef OPTSET 1222 1223 case IP_MULTICAST_IF: 1224 case IP_MULTICAST_VIF: 1225 case IP_MULTICAST_TTL: 1226 case IP_MULTICAST_LOOP: 1227 case IP_ADD_MEMBERSHIP: 1228 case IP_DROP_MEMBERSHIP: 1229 error = ip_setmoptions(sopt, &inp->inp_moptions); 1230 break; 1231 1232 case IP_PORTRANGE: 1233 error = sooptcopyin(sopt, &optval, sizeof optval, 1234 sizeof optval); 1235 if (error) 1236 break; 1237 1238 switch (optval) { 1239 case IP_PORTRANGE_DEFAULT: 1240 inp->inp_flags &= ~(INP_LOWPORT); 1241 inp->inp_flags &= ~(INP_HIGHPORT); 1242 break; 1243 1244 case IP_PORTRANGE_HIGH: 1245 inp->inp_flags &= ~(INP_LOWPORT); 1246 inp->inp_flags |= INP_HIGHPORT; 1247 break; 1248 1249 case IP_PORTRANGE_LOW: 1250 inp->inp_flags &= ~(INP_HIGHPORT); 1251 inp->inp_flags |= INP_LOWPORT; 1252 break; 1253 1254 default: 1255 error = EINVAL; 1256 break; 1257 } 1258 break; 1259 1260#ifdef IPSEC 1261 case IP_IPSEC_POLICY: 1262 { 1263 caddr_t req; 1264 size_t len = 0; 1265 int priv; 1266 struct mbuf *m; 1267 int optname; 1268 1269 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1270 break; 1271 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1272 break; 1273 priv = (sopt->sopt_td != NULL && 1274 suser(sopt->sopt_td) != 0) ? 0 : 1; 1275 req = mtod(m, caddr_t); 1276 len = m->m_len; 1277 optname = sopt->sopt_name; 1278 error = ipsec4_set_policy(inp, optname, req, len, priv); 1279 m_freem(m); 1280 break; 1281 } 1282#endif /*IPSEC*/ 1283 1284 default: 1285 error = ENOPROTOOPT; 1286 break; 1287 } 1288 break; 1289 1290 case SOPT_GET: 1291 switch (sopt->sopt_name) { 1292 case IP_OPTIONS: 1293 case IP_RETOPTS: 1294 if (inp->inp_options) 1295 error = sooptcopyout(sopt, 1296 mtod(inp->inp_options, 1297 char *), 1298 inp->inp_options->m_len); 1299 else 1300 sopt->sopt_valsize = 0; 1301 break; 1302 1303 case IP_TOS: 1304 case IP_TTL: 1305 case IP_RECVOPTS: 1306 case IP_RECVRETOPTS: 1307 case IP_RECVDSTADDR: 1308 case IP_RECVIF: 1309 case IP_PORTRANGE: 1310 case IP_FAITH: 1311 switch (sopt->sopt_name) { 1312 1313 case IP_TOS: 1314 optval = inp->inp_ip_tos; 1315 break; 1316 1317 case IP_TTL: 1318 optval = inp->inp_ip_ttl; 1319 break; 1320 1321#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1322 1323 case IP_RECVOPTS: 1324 optval = OPTBIT(INP_RECVOPTS); 1325 break; 1326 1327 case IP_RECVRETOPTS: 1328 optval = OPTBIT(INP_RECVRETOPTS); 1329 break; 1330 1331 case IP_RECVDSTADDR: 1332 optval = OPTBIT(INP_RECVDSTADDR); 1333 break; 1334 1335 case IP_RECVIF: 1336 optval = OPTBIT(INP_RECVIF); 1337 break; 1338 1339 case IP_PORTRANGE: 1340 if (inp->inp_flags & INP_HIGHPORT) 1341 optval = IP_PORTRANGE_HIGH; 1342 else if (inp->inp_flags & INP_LOWPORT) 1343 optval = IP_PORTRANGE_LOW; 1344 else 1345 optval = 0; 1346 break; 1347 1348 case IP_FAITH: 1349 optval = OPTBIT(INP_FAITH); 1350 break; 1351 } 1352 error = sooptcopyout(sopt, &optval, sizeof optval); 1353 break; 1354 1355 case IP_MULTICAST_IF: 1356 case IP_MULTICAST_VIF: 1357 case IP_MULTICAST_TTL: 1358 case IP_MULTICAST_LOOP: 1359 case IP_ADD_MEMBERSHIP: 1360 case IP_DROP_MEMBERSHIP: 1361 error = ip_getmoptions(sopt, inp->inp_moptions); 1362 break; 1363 1364#ifdef IPSEC 1365 case IP_IPSEC_POLICY: 1366 { 1367 struct mbuf *m = NULL; 1368 caddr_t req = NULL; 1369 size_t len = 0; 1370 1371 if (m != 0) { 1372 req = mtod(m, caddr_t); 1373 len = m->m_len; 1374 } 1375 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1376 if (error == 0) 1377 error = soopt_mcopyout(sopt, m); /* XXX */ 1378 if (error == 0) 1379 m_freem(m); 1380 break; 1381 } 1382#endif /*IPSEC*/ 1383 1384 default: 1385 error = ENOPROTOOPT; 1386 break; 1387 } 1388 break; 1389 } 1390 return (error); 1391} 1392 1393/* 1394 * Set up IP options in pcb for insertion in output packets. 1395 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1396 * with destination address if source routed. 1397 */ 1398static int 1399ip_pcbopts(optname, pcbopt, m) 1400 int optname; 1401 struct mbuf **pcbopt; 1402 register struct mbuf *m; 1403{ 1404 register int cnt, optlen; 1405 register u_char *cp; 1406 u_char opt; 1407 1408 /* turn off any old options */ 1409 if (*pcbopt) 1410 (void)m_free(*pcbopt); 1411 *pcbopt = 0; 1412 if (m == (struct mbuf *)0 || m->m_len == 0) { 1413 /* 1414 * Only turning off any previous options. 1415 */ 1416 if (m) 1417 (void)m_free(m); 1418 return (0); 1419 } 1420 1421 if (m->m_len % sizeof(int32_t)) 1422 goto bad; 1423 /* 1424 * IP first-hop destination address will be stored before 1425 * actual options; move other options back 1426 * and clear it when none present. 1427 */ 1428 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1429 goto bad; 1430 cnt = m->m_len; 1431 m->m_len += sizeof(struct in_addr); 1432 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1433 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt); 1434 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1435 1436 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1437 opt = cp[IPOPT_OPTVAL]; 1438 if (opt == IPOPT_EOL) 1439 break; 1440 if (opt == IPOPT_NOP) 1441 optlen = 1; 1442 else { 1443 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1444 goto bad; 1445 optlen = cp[IPOPT_OLEN]; 1446 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1447 goto bad; 1448 } 1449 switch (opt) { 1450 1451 default: 1452 break; 1453 1454 case IPOPT_LSRR: 1455 case IPOPT_SSRR: 1456 /* 1457 * user process specifies route as: 1458 * ->A->B->C->D 1459 * D must be our final destination (but we can't 1460 * check that since we may not have connected yet). 1461 * A is first hop destination, which doesn't appear in 1462 * actual IP option, but is stored before the options. 1463 */ 1464 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1465 goto bad; 1466 m->m_len -= sizeof(struct in_addr); 1467 cnt -= sizeof(struct in_addr); 1468 optlen -= sizeof(struct in_addr); 1469 cp[IPOPT_OLEN] = optlen; 1470 /* 1471 * Move first hop before start of options. 1472 */ 1473 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1474 sizeof(struct in_addr)); 1475 /* 1476 * Then copy rest of options back 1477 * to close up the deleted entry. 1478 */ 1479 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] + 1480 sizeof(struct in_addr)), 1481 (caddr_t)&cp[IPOPT_OFFSET+1], 1482 (unsigned)cnt + sizeof(struct in_addr)); 1483 break; 1484 } 1485 } 1486 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1487 goto bad; 1488 *pcbopt = m; 1489 return (0); 1490 1491bad: 1492 (void)m_free(m); 1493 return (EINVAL); 1494} 1495 1496/* 1497 * XXX 1498 * The whole multicast option thing needs to be re-thought. 1499 * Several of these options are equally applicable to non-multicast 1500 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1501 * standard option (IP_TTL). 1502 */ 1503 1504/* 1505 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1506 */ 1507static struct ifnet * 1508ip_multicast_if(a, ifindexp) 1509 struct in_addr *a; 1510 int *ifindexp; 1511{ 1512 int ifindex; 1513 struct ifnet *ifp; 1514 1515 if (ifindexp) 1516 *ifindexp = 0; 1517 if (ntohl(a->s_addr) >> 24 == 0) { 1518 ifindex = ntohl(a->s_addr) & 0xffffff; 1519 if (ifindex < 0 || if_index < ifindex) 1520 return NULL; 1521 ifp = ifnet_byindex(ifindex); 1522 if (ifindexp) 1523 *ifindexp = ifindex; 1524 } else { 1525 INADDR_TO_IFP(*a, ifp); 1526 } 1527 return ifp; 1528} 1529 1530/* 1531 * Set the IP multicast options in response to user setsockopt(). 1532 */ 1533static int 1534ip_setmoptions(sopt, imop) 1535 struct sockopt *sopt; 1536 struct ip_moptions **imop; 1537{ 1538 int error = 0; 1539 int i; 1540 struct in_addr addr; 1541 struct ip_mreq mreq; 1542 struct ifnet *ifp; 1543 struct ip_moptions *imo = *imop; 1544 struct route ro; 1545 struct sockaddr_in *dst; 1546 int ifindex; 1547 int s; 1548 1549 if (imo == NULL) { 1550 /* 1551 * No multicast option buffer attached to the pcb; 1552 * allocate one and initialize to default values. 1553 */ 1554 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1555 M_WAITOK); 1556 1557 if (imo == NULL) 1558 return (ENOBUFS); 1559 *imop = imo; 1560 imo->imo_multicast_ifp = NULL; 1561 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1562 imo->imo_multicast_vif = -1; 1563 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1564 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1565 imo->imo_num_memberships = 0; 1566 } 1567 1568 switch (sopt->sopt_name) { 1569 /* store an index number for the vif you wanna use in the send */ 1570 case IP_MULTICAST_VIF: 1571 if (legal_vif_num == 0) { 1572 error = EOPNOTSUPP; 1573 break; 1574 } 1575 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1576 if (error) 1577 break; 1578 if (!legal_vif_num(i) && (i != -1)) { 1579 error = EINVAL; 1580 break; 1581 } 1582 imo->imo_multicast_vif = i; 1583 break; 1584 1585 case IP_MULTICAST_IF: 1586 /* 1587 * Select the interface for outgoing multicast packets. 1588 */ 1589 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1590 if (error) 1591 break; 1592 /* 1593 * INADDR_ANY is used to remove a previous selection. 1594 * When no interface is selected, a default one is 1595 * chosen every time a multicast packet is sent. 1596 */ 1597 if (addr.s_addr == INADDR_ANY) { 1598 imo->imo_multicast_ifp = NULL; 1599 break; 1600 } 1601 /* 1602 * The selected interface is identified by its local 1603 * IP address. Find the interface and confirm that 1604 * it supports multicasting. 1605 */ 1606 s = splimp(); 1607 ifp = ip_multicast_if(&addr, &ifindex); 1608 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1609 splx(s); 1610 error = EADDRNOTAVAIL; 1611 break; 1612 } 1613 imo->imo_multicast_ifp = ifp; 1614 if (ifindex) 1615 imo->imo_multicast_addr = addr; 1616 else 1617 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1618 splx(s); 1619 break; 1620 1621 case IP_MULTICAST_TTL: 1622 /* 1623 * Set the IP time-to-live for outgoing multicast packets. 1624 * The original multicast API required a char argument, 1625 * which is inconsistent with the rest of the socket API. 1626 * We allow either a char or an int. 1627 */ 1628 if (sopt->sopt_valsize == 1) { 1629 u_char ttl; 1630 error = sooptcopyin(sopt, &ttl, 1, 1); 1631 if (error) 1632 break; 1633 imo->imo_multicast_ttl = ttl; 1634 } else { 1635 u_int ttl; 1636 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1637 sizeof ttl); 1638 if (error) 1639 break; 1640 if (ttl > 255) 1641 error = EINVAL; 1642 else 1643 imo->imo_multicast_ttl = ttl; 1644 } 1645 break; 1646 1647 case IP_MULTICAST_LOOP: 1648 /* 1649 * Set the loopback flag for outgoing multicast packets. 1650 * Must be zero or one. The original multicast API required a 1651 * char argument, which is inconsistent with the rest 1652 * of the socket API. We allow either a char or an int. 1653 */ 1654 if (sopt->sopt_valsize == 1) { 1655 u_char loop; 1656 error = sooptcopyin(sopt, &loop, 1, 1); 1657 if (error) 1658 break; 1659 imo->imo_multicast_loop = !!loop; 1660 } else { 1661 u_int loop; 1662 error = sooptcopyin(sopt, &loop, sizeof loop, 1663 sizeof loop); 1664 if (error) 1665 break; 1666 imo->imo_multicast_loop = !!loop; 1667 } 1668 break; 1669 1670 case IP_ADD_MEMBERSHIP: 1671 /* 1672 * Add a multicast group membership. 1673 * Group must be a valid IP multicast address. 1674 */ 1675 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1676 if (error) 1677 break; 1678 1679 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1680 error = EINVAL; 1681 break; 1682 } 1683 s = splimp(); 1684 /* 1685 * If no interface address was provided, use the interface of 1686 * the route to the given multicast address. 1687 */ 1688 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1689 bzero((caddr_t)&ro, sizeof(ro)); 1690 dst = (struct sockaddr_in *)&ro.ro_dst; 1691 dst->sin_len = sizeof(*dst); 1692 dst->sin_family = AF_INET; 1693 dst->sin_addr = mreq.imr_multiaddr; 1694 rtalloc(&ro); 1695 if (ro.ro_rt == NULL) { 1696 error = EADDRNOTAVAIL; 1697 splx(s); 1698 break; 1699 } 1700 ifp = ro.ro_rt->rt_ifp; 1701 rtfree(ro.ro_rt); 1702 } 1703 else { 1704 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1705 } 1706 1707 /* 1708 * See if we found an interface, and confirm that it 1709 * supports multicast. 1710 */ 1711 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1712 error = EADDRNOTAVAIL; 1713 splx(s); 1714 break; 1715 } 1716 /* 1717 * See if the membership already exists or if all the 1718 * membership slots are full. 1719 */ 1720 for (i = 0; i < imo->imo_num_memberships; ++i) { 1721 if (imo->imo_membership[i]->inm_ifp == ifp && 1722 imo->imo_membership[i]->inm_addr.s_addr 1723 == mreq.imr_multiaddr.s_addr) 1724 break; 1725 } 1726 if (i < imo->imo_num_memberships) { 1727 error = EADDRINUSE; 1728 splx(s); 1729 break; 1730 } 1731 if (i == IP_MAX_MEMBERSHIPS) { 1732 error = ETOOMANYREFS; 1733 splx(s); 1734 break; 1735 } 1736 /* 1737 * Everything looks good; add a new record to the multicast 1738 * address list for the given interface. 1739 */ 1740 if ((imo->imo_membership[i] = 1741 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1742 error = ENOBUFS; 1743 splx(s); 1744 break; 1745 } 1746 ++imo->imo_num_memberships; 1747 splx(s); 1748 break; 1749 1750 case IP_DROP_MEMBERSHIP: 1751 /* 1752 * Drop a multicast group membership. 1753 * Group must be a valid IP multicast address. 1754 */ 1755 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1756 if (error) 1757 break; 1758 1759 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1760 error = EINVAL; 1761 break; 1762 } 1763 1764 s = splimp(); 1765 /* 1766 * If an interface address was specified, get a pointer 1767 * to its ifnet structure. 1768 */ 1769 if (mreq.imr_interface.s_addr == INADDR_ANY) 1770 ifp = NULL; 1771 else { 1772 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1773 if (ifp == NULL) { 1774 error = EADDRNOTAVAIL; 1775 splx(s); 1776 break; 1777 } 1778 } 1779 /* 1780 * Find the membership in the membership array. 1781 */ 1782 for (i = 0; i < imo->imo_num_memberships; ++i) { 1783 if ((ifp == NULL || 1784 imo->imo_membership[i]->inm_ifp == ifp) && 1785 imo->imo_membership[i]->inm_addr.s_addr == 1786 mreq.imr_multiaddr.s_addr) 1787 break; 1788 } 1789 if (i == imo->imo_num_memberships) { 1790 error = EADDRNOTAVAIL; 1791 splx(s); 1792 break; 1793 } 1794 /* 1795 * Give up the multicast address record to which the 1796 * membership points. 1797 */ 1798 in_delmulti(imo->imo_membership[i]); 1799 /* 1800 * Remove the gap in the membership array. 1801 */ 1802 for (++i; i < imo->imo_num_memberships; ++i) 1803 imo->imo_membership[i-1] = imo->imo_membership[i]; 1804 --imo->imo_num_memberships; 1805 splx(s); 1806 break; 1807 1808 default: 1809 error = EOPNOTSUPP; 1810 break; 1811 } 1812 1813 /* 1814 * If all options have default values, no need to keep the mbuf. 1815 */ 1816 if (imo->imo_multicast_ifp == NULL && 1817 imo->imo_multicast_vif == -1 && 1818 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1819 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1820 imo->imo_num_memberships == 0) { 1821 free(*imop, M_IPMOPTS); 1822 *imop = NULL; 1823 } 1824 1825 return (error); 1826} 1827 1828/* 1829 * Return the IP multicast options in response to user getsockopt(). 1830 */ 1831static int 1832ip_getmoptions(sopt, imo) 1833 struct sockopt *sopt; 1834 register struct ip_moptions *imo; 1835{ 1836 struct in_addr addr; 1837 struct in_ifaddr *ia; 1838 int error, optval; 1839 u_char coptval; 1840 1841 error = 0; 1842 switch (sopt->sopt_name) { 1843 case IP_MULTICAST_VIF: 1844 if (imo != NULL) 1845 optval = imo->imo_multicast_vif; 1846 else 1847 optval = -1; 1848 error = sooptcopyout(sopt, &optval, sizeof optval); 1849 break; 1850 1851 case IP_MULTICAST_IF: 1852 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1853 addr.s_addr = INADDR_ANY; 1854 else if (imo->imo_multicast_addr.s_addr) { 1855 /* return the value user has set */ 1856 addr = imo->imo_multicast_addr; 1857 } else { 1858 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1859 addr.s_addr = (ia == NULL) ? INADDR_ANY 1860 : IA_SIN(ia)->sin_addr.s_addr; 1861 } 1862 error = sooptcopyout(sopt, &addr, sizeof addr); 1863 break; 1864 1865 case IP_MULTICAST_TTL: 1866 if (imo == 0) 1867 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1868 else 1869 optval = coptval = imo->imo_multicast_ttl; 1870 if (sopt->sopt_valsize == 1) 1871 error = sooptcopyout(sopt, &coptval, 1); 1872 else 1873 error = sooptcopyout(sopt, &optval, sizeof optval); 1874 break; 1875 1876 case IP_MULTICAST_LOOP: 1877 if (imo == 0) 1878 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1879 else 1880 optval = coptval = imo->imo_multicast_loop; 1881 if (sopt->sopt_valsize == 1) 1882 error = sooptcopyout(sopt, &coptval, 1); 1883 else 1884 error = sooptcopyout(sopt, &optval, sizeof optval); 1885 break; 1886 1887 default: 1888 error = ENOPROTOOPT; 1889 break; 1890 } 1891 return (error); 1892} 1893 1894/* 1895 * Discard the IP multicast options. 1896 */ 1897void 1898ip_freemoptions(imo) 1899 register struct ip_moptions *imo; 1900{ 1901 register int i; 1902 1903 if (imo != NULL) { 1904 for (i = 0; i < imo->imo_num_memberships; ++i) 1905 in_delmulti(imo->imo_membership[i]); 1906 free(imo, M_IPMOPTS); 1907 } 1908} 1909 1910/* 1911 * Routine called from ip_output() to loop back a copy of an IP multicast 1912 * packet to the input queue of a specified interface. Note that this 1913 * calls the output routine of the loopback "driver", but with an interface 1914 * pointer that might NOT be a loopback interface -- evil, but easier than 1915 * replicating that code here. 1916 */ 1917static void 1918ip_mloopback(ifp, m, dst, hlen) 1919 struct ifnet *ifp; 1920 register struct mbuf *m; 1921 register struct sockaddr_in *dst; 1922 int hlen; 1923{ 1924 register struct ip *ip; 1925 struct mbuf *copym; 1926 1927 copym = m_copy(m, 0, M_COPYALL); 1928 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1929 copym = m_pullup(copym, hlen); 1930 if (copym != NULL) { 1931 /* 1932 * We don't bother to fragment if the IP length is greater 1933 * than the interface's MTU. Can this possibly matter? 1934 */ 1935 ip = mtod(copym, struct ip *); 1936 ip->ip_len = htons(ip->ip_len); 1937 ip->ip_off = htons(ip->ip_off); 1938 ip->ip_sum = 0; 1939 if (ip->ip_vhl == IP_VHL_BORING) { 1940 ip->ip_sum = in_cksum_hdr(ip); 1941 } else { 1942 ip->ip_sum = in_cksum(copym, hlen); 1943 } 1944 /* 1945 * NB: 1946 * It's not clear whether there are any lingering 1947 * reentrancy problems in other areas which might 1948 * be exposed by using ip_input directly (in 1949 * particular, everything which modifies the packet 1950 * in-place). Yet another option is using the 1951 * protosw directly to deliver the looped back 1952 * packet. For the moment, we'll err on the side 1953 * of safety by using if_simloop(). 1954 */ 1955#if 1 /* XXX */ 1956 if (dst->sin_family != AF_INET) { 1957 printf("ip_mloopback: bad address family %d\n", 1958 dst->sin_family); 1959 dst->sin_family = AF_INET; 1960 } 1961#endif 1962 1963#ifdef notdef 1964 copym->m_pkthdr.rcvif = ifp; 1965 ip_input(copym); 1966#else 1967 /* if the checksum hasn't been computed, mark it as valid */ 1968 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1969 copym->m_pkthdr.csum_flags |= 1970 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1971 copym->m_pkthdr.csum_data = 0xffff; 1972 } 1973 if_simloop(ifp, copym, dst->sin_family, 0); 1974#endif 1975 } 1976} 1977