ip_output.c revision 88931
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 88931 2002-01-05 18:23:53Z fenner $ 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 __P((struct mbuf *, struct mbuf *, int *)); 93static struct ifnet *ip_multicast_if __P((struct in_addr *, int *)); 94static void ip_mloopback 95 __P((struct ifnet *, struct mbuf *, struct sockaddr_in *, int)); 96static int ip_getmoptions 97 __P((struct sockopt *, struct ip_moptions *)); 98static int ip_pcbopts __P((int, struct mbuf **, struct mbuf *)); 99static int ip_setmoptions 100 __P((struct sockopt *, struct ip_moptions **)); 101 102int ip_optcopy __P((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 route iproute; 127 struct sockaddr_in *dst; 128 struct in_ifaddr *ia; 129 int isbroadcast, sw_csum; 130 struct in_addr pkt_dst; 131#ifdef IPSEC 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#ifdef DIAGNOSTIC 190 if ((m->m_flags & M_PKTHDR) == 0) 191 panic("ip_output no HDR"); 192#endif 193 if (opt) { 194 m = ip_insertoptions(m, opt, &len); 195 hlen = len; 196 } 197 ip = mtod(m, struct ip *); 198 pkt_dst = ip_fw_fwd_addr == NULL 199 ? ip->ip_dst : ip_fw_fwd_addr->sin_addr; 200 201 /* 202 * Fill in IP header. 203 */ 204 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 205 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2); 206 ip->ip_off &= IP_DF; 207#ifdef RANDOM_IP_ID 208 ip->ip_id = ip_randomid(); 209#else 210 ip->ip_id = htons(ip_id++); 211#endif 212 ipstat.ips_localout++; 213 } else { 214 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 215 } 216 217 /* Route packet. */ 218 if (ro == NULL) { 219 ro = &iproute; 220 bzero(ro, sizeof(*ro)); 221 } 222 dst = (struct sockaddr_in *)&ro->ro_dst; 223 /* 224 * If there is a cached route, 225 * check that it is to the same destination 226 * and is still up. If not, free it and try again. 227 */ 228 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 229 dst->sin_addr.s_addr != pkt_dst.s_addr)) { 230 RTFREE(ro->ro_rt); 231 ro->ro_rt = (struct rtentry *)0; 232 } 233 if (ro->ro_rt == 0) { 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 * and 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 HTONS(ip->ip_len); 513 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 NTOHS(ip->ip_len); 573 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(&ip, 612 hlen, ifp, &divert_cookie, &m, &rule, &dst); 613 /* 614 * On return we must do the following: 615 * m == NULL -> drop the pkt (old interface, deprecated) 616 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 617 * 1<=off<= 0xffff -> DIVERT 618 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 619 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 620 * dst != old -> IPFIREWALL_FORWARD 621 * off==0, dst==old -> accept 622 * If some of the above modules are not compiled in, then 623 * we should't have to check the corresponding condition 624 * (because the ipfw control socket should not accept 625 * unsupported rules), but better play safe and drop 626 * packets in case of doubt. 627 */ 628 if (off & IP_FW_PORT_DENY_FLAG) { /* XXX new interface-denied */ 629 if (m) 630 m_freem(m); 631 error = EACCES; 632 goto done; 633 } 634 if (!m) { /* firewall said to reject */ 635 static int __debug=10; 636 637 if (__debug > 0) { 638 printf( 639 "firewall returns NULL, please update!\n"); 640 __debug--; 641 } 642 error = EACCES; 643 goto done; 644 } 645 if (off == 0 && dst == old) /* common case */ 646 goto pass; 647 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 648 /* 649 * pass the pkt to dummynet. Need to include 650 * pipe number, m, ifp, ro, dst because these are 651 * not recomputed in the next pass. 652 * All other parameters have been already used and 653 * so they are not needed anymore. 654 * XXX note: if the ifp or ro entry are deleted 655 * while a pkt is in dummynet, we are in trouble! 656 */ 657 error = ip_dn_io_ptr(off & 0xffff, DN_TO_IP_OUT, m, 658 ifp, ro, dst, rule, flags); 659 goto done; 660 } 661#ifdef IPDIVERT 662 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 663 struct mbuf *clone = NULL; 664 665 /* Clone packet if we're doing a 'tee' */ 666 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 667 clone = m_dup(m, M_DONTWAIT); 668 669 /* 670 * XXX 671 * delayed checksums are not currently compatible 672 * with divert sockets. 673 */ 674 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 675 in_delayed_cksum(m); 676 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 677 } 678 679 /* Restore packet header fields to original values */ 680 HTONS(ip->ip_len); 681 HTONS(ip->ip_off); 682 683 /* Deliver packet to divert input routine */ 684 ip_divert_cookie = divert_cookie; 685 divert_packet(m, 0, off & 0xffff); 686 687 /* If 'tee', continue with original packet */ 688 if (clone != NULL) { 689 m = clone; 690 ip = mtod(m, struct ip *); 691 goto pass; 692 } 693 goto done; 694 } 695#endif 696 697#ifdef IPFIREWALL_FORWARD 698 /* Here we check dst to make sure it's directly reachable on the 699 * interface we previously thought it was. 700 * If it isn't (which may be likely in some situations) we have 701 * to re-route it (ie, find a route for the next-hop and the 702 * associated interface) and set them here. This is nested 703 * forwarding which in most cases is undesirable, except where 704 * such control is nigh impossible. So we do it here. 705 * And I'm babbling. 706 */ 707 if (off == 0 && old != dst) { 708 struct in_ifaddr *ia; 709 710 /* It's changed... */ 711 /* There must be a better way to do this next line... */ 712 static struct route sro_fwd, *ro_fwd = &sro_fwd; 713#ifdef IPFIREWALL_FORWARD_DEBUG 714 printf("IPFIREWALL_FORWARD: New dst ip: "); 715 print_ip(dst->sin_addr); 716 printf("\n"); 717#endif 718 /* 719 * We need to figure out if we have been forwarded 720 * to a local socket. If so then we should somehow 721 * "loop back" to ip_input, and get directed to the 722 * PCB as if we had received this packet. This is 723 * because it may be dificult to identify the packets 724 * you want to forward until they are being output 725 * and have selected an interface. (e.g. locally 726 * initiated packets) If we used the loopback inteface, 727 * we would not be able to control what happens 728 * as the packet runs through ip_input() as 729 * it is done through a ISR. 730 */ 731 LIST_FOREACH(ia, 732 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 733 /* 734 * If the addr to forward to is one 735 * of ours, we pretend to 736 * be the destination for this packet. 737 */ 738 if (IA_SIN(ia)->sin_addr.s_addr == 739 dst->sin_addr.s_addr) 740 break; 741 } 742 if (ia) { 743 /* tell ip_input "dont filter" */ 744 ip_fw_fwd_addr = dst; 745 if (m->m_pkthdr.rcvif == NULL) 746 m->m_pkthdr.rcvif = ifunit("lo0"); 747 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 748 m->m_pkthdr.csum_flags |= 749 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 750 m0->m_pkthdr.csum_data = 0xffff; 751 } 752 m->m_pkthdr.csum_flags |= 753 CSUM_IP_CHECKED | CSUM_IP_VALID; 754 HTONS(ip->ip_len); 755 HTONS(ip->ip_off); 756 ip_input(m); 757 goto done; 758 } 759 /* Some of the logic for this was 760 * nicked from above. 761 * 762 * This rewrites the cached route in a local PCB. 763 * Is this what we want to do? 764 */ 765 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 766 767 ro_fwd->ro_rt = 0; 768 rtalloc_ign(ro_fwd, RTF_PRCLONING); 769 770 if (ro_fwd->ro_rt == 0) { 771 ipstat.ips_noroute++; 772 error = EHOSTUNREACH; 773 goto bad; 774 } 775 776 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 777 ifp = ro_fwd->ro_rt->rt_ifp; 778 ro_fwd->ro_rt->rt_use++; 779 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 780 dst = (struct sockaddr_in *)ro_fwd->ro_rt->rt_gateway; 781 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 782 isbroadcast = 783 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 784 else 785 isbroadcast = in_broadcast(dst->sin_addr, ifp); 786 if (ro->ro_rt) 787 RTFREE(ro->ro_rt); 788 ro->ro_rt = ro_fwd->ro_rt; 789 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 790 791 /* 792 * If we added a default src ip earlier, 793 * which would have been gotten from the-then 794 * interface, do it again, from the new one. 795 */ 796 if (fwd_rewrite_src) 797 ip->ip_src = IA_SIN(ia)->sin_addr; 798 goto pass ; 799 } 800#endif /* IPFIREWALL_FORWARD */ 801 /* 802 * if we get here, none of the above matches, and 803 * we have to drop the pkt 804 */ 805 m_freem(m); 806 error = EACCES; /* not sure this is the right error msg */ 807 goto done; 808 } 809 810 ip_fw_fwd_addr = NULL; 811pass: 812 m->m_pkthdr.csum_flags |= CSUM_IP; 813 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 814 if (sw_csum & CSUM_DELAY_DATA) { 815 in_delayed_cksum(m); 816 sw_csum &= ~CSUM_DELAY_DATA; 817 } 818 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 819 820 /* 821 * If small enough for interface, or the interface will take 822 * care of the fragmentation for us, can just send directly. 823 */ 824 if ((u_short)ip->ip_len <= ifp->if_mtu || 825 ifp->if_hwassist & CSUM_FRAGMENT) { 826 HTONS(ip->ip_len); 827 HTONS(ip->ip_off); 828 ip->ip_sum = 0; 829 if (sw_csum & CSUM_DELAY_IP) { 830 if (ip->ip_vhl == IP_VHL_BORING) { 831 ip->ip_sum = in_cksum_hdr(ip); 832 } else { 833 ip->ip_sum = in_cksum(m, hlen); 834 } 835 } 836 837 /* Record statistics for this interface address. */ 838 if (!(flags & IP_FORWARDING) && ia) { 839 ia->ia_ifa.if_opackets++; 840 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 841 } 842 843#ifdef IPSEC 844 /* clean ipsec history once it goes out of the node */ 845 ipsec_delaux(m); 846#endif 847 848 error = (*ifp->if_output)(ifp, m, 849 (struct sockaddr *)dst, ro->ro_rt); 850 goto done; 851 } 852 /* 853 * Too large for interface; fragment if possible. 854 * Must be able to put at least 8 bytes per fragment. 855 */ 856 if (ip->ip_off & IP_DF) { 857 error = EMSGSIZE; 858 /* 859 * This case can happen if the user changed the MTU 860 * of an interface after enabling IP on it. Because 861 * most netifs don't keep track of routes pointing to 862 * them, there is no way for one to update all its 863 * routes when the MTU is changed. 864 */ 865 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) 866 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) 867 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 868 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 869 } 870 ipstat.ips_cantfrag++; 871 goto bad; 872 } 873 len = (ifp->if_mtu - hlen) &~ 7; 874 if (len < 8) { 875 error = EMSGSIZE; 876 goto bad; 877 } 878 879 /* 880 * if the interface will not calculate checksums on 881 * fragmented packets, then do it here. 882 */ 883 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 884 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) { 885 in_delayed_cksum(m); 886 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 887 } 888 889 { 890 int mhlen, firstlen = len; 891 struct mbuf **mnext = &m->m_nextpkt; 892 int nfrags = 1; 893 894 /* 895 * Loop through length of segment after first fragment, 896 * make new header and copy data of each part and link onto chain. 897 */ 898 m0 = m; 899 mhlen = sizeof (struct ip); 900 for (off = hlen + len; off < (u_short)ip->ip_len; off += len) { 901 MGETHDR(m, M_DONTWAIT, MT_HEADER); 902 if (m == 0) { 903 error = ENOBUFS; 904 ipstat.ips_odropped++; 905 goto sendorfree; 906 } 907 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 908 m->m_data += max_linkhdr; 909 mhip = mtod(m, struct ip *); 910 *mhip = *ip; 911 if (hlen > sizeof (struct ip)) { 912 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 913 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2); 914 } 915 m->m_len = mhlen; 916 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 917 if (off + len >= (u_short)ip->ip_len) 918 len = (u_short)ip->ip_len - off; 919 else 920 mhip->ip_off |= IP_MF; 921 mhip->ip_len = htons((u_short)(len + mhlen)); 922 m->m_next = m_copy(m0, off, len); 923 if (m->m_next == 0) { 924 (void) m_free(m); 925 error = ENOBUFS; /* ??? */ 926 ipstat.ips_odropped++; 927 goto sendorfree; 928 } 929 m->m_pkthdr.len = mhlen + len; 930 m->m_pkthdr.rcvif = (struct ifnet *)0; 931 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 932 HTONS(mhip->ip_off); 933 mhip->ip_sum = 0; 934 if (sw_csum & CSUM_DELAY_IP) { 935 if (mhip->ip_vhl == IP_VHL_BORING) { 936 mhip->ip_sum = in_cksum_hdr(mhip); 937 } else { 938 mhip->ip_sum = in_cksum(m, mhlen); 939 } 940 } 941 *mnext = m; 942 mnext = &m->m_nextpkt; 943 nfrags++; 944 } 945 ipstat.ips_ofragments += nfrags; 946 947 /* set first/last markers for fragment chain */ 948 m->m_flags |= M_LASTFRAG; 949 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 950 m0->m_pkthdr.csum_data = nfrags; 951 952 /* 953 * Update first fragment by trimming what's been copied out 954 * and updating header, then send each fragment (in order). 955 */ 956 m = m0; 957 m_adj(m, hlen + firstlen - (u_short)ip->ip_len); 958 m->m_pkthdr.len = hlen + firstlen; 959 ip->ip_len = htons((u_short)m->m_pkthdr.len); 960 ip->ip_off |= IP_MF; 961 HTONS(ip->ip_off); 962 ip->ip_sum = 0; 963 if (sw_csum & CSUM_DELAY_IP) { 964 if (ip->ip_vhl == IP_VHL_BORING) { 965 ip->ip_sum = in_cksum_hdr(ip); 966 } else { 967 ip->ip_sum = in_cksum(m, hlen); 968 } 969 } 970sendorfree: 971 for (m = m0; m; m = m0) { 972 m0 = m->m_nextpkt; 973 m->m_nextpkt = 0; 974#ifdef IPSEC 975 /* clean ipsec history once it goes out of the node */ 976 ipsec_delaux(m); 977#endif 978 if (error == 0) { 979 /* Record statistics for this interface address. */ 980 if (ia != NULL) { 981 ia->ia_ifa.if_opackets++; 982 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 983 } 984 985 error = (*ifp->if_output)(ifp, m, 986 (struct sockaddr *)dst, ro->ro_rt); 987 } else 988 m_freem(m); 989 } 990 991 if (error == 0) 992 ipstat.ips_fragmented++; 993 } 994done: 995#ifdef IPSEC 996 if (ro == &iproute && ro->ro_rt) { 997 RTFREE(ro->ro_rt); 998 ro->ro_rt = NULL; 999 } 1000 if (sp != NULL) { 1001 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1002 printf("DP ip_output call free SP:%p\n", sp)); 1003 key_freesp(sp); 1004 } 1005#endif /* IPSEC */ 1006 return (error); 1007bad: 1008 m_freem(m); 1009 goto done; 1010} 1011 1012void 1013in_delayed_cksum(struct mbuf *m) 1014{ 1015 struct ip *ip; 1016 u_short csum, offset; 1017 1018 ip = mtod(m, struct ip *); 1019 offset = IP_VHL_HL(ip->ip_vhl) << 2 ; 1020 csum = in_cksum_skip(m, ip->ip_len, offset); 1021 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1022 csum = 0xffff; 1023 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1024 1025 if (offset + sizeof(u_short) > m->m_len) { 1026 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1027 m->m_len, offset, ip->ip_p); 1028 /* 1029 * XXX 1030 * this shouldn't happen, but if it does, the 1031 * correct behavior may be to insert the checksum 1032 * in the existing chain instead of rearranging it. 1033 */ 1034 m = m_pullup(m, offset + sizeof(u_short)); 1035 } 1036 *(u_short *)(m->m_data + offset) = csum; 1037} 1038 1039/* 1040 * Insert IP options into preformed packet. 1041 * Adjust IP destination as required for IP source routing, 1042 * as indicated by a non-zero in_addr at the start of the options. 1043 * 1044 * XXX This routine assumes that the packet has no options in place. 1045 */ 1046static struct mbuf * 1047ip_insertoptions(m, opt, phlen) 1048 register struct mbuf *m; 1049 struct mbuf *opt; 1050 int *phlen; 1051{ 1052 register struct ipoption *p = mtod(opt, struct ipoption *); 1053 struct mbuf *n; 1054 register struct ip *ip = mtod(m, struct ip *); 1055 unsigned optlen; 1056 1057 optlen = opt->m_len - sizeof(p->ipopt_dst); 1058 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) 1059 return (m); /* XXX should fail */ 1060 if (p->ipopt_dst.s_addr) 1061 ip->ip_dst = p->ipopt_dst; 1062 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1063 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1064 if (n == 0) 1065 return (m); 1066 n->m_pkthdr.rcvif = (struct ifnet *)0; 1067 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1068 m->m_len -= sizeof(struct ip); 1069 m->m_data += sizeof(struct ip); 1070 n->m_next = m; 1071 m = n; 1072 m->m_len = optlen + sizeof(struct ip); 1073 m->m_data += max_linkhdr; 1074 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1075 } else { 1076 m->m_data -= optlen; 1077 m->m_len += optlen; 1078 m->m_pkthdr.len += optlen; 1079 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1080 } 1081 ip = mtod(m, struct ip *); 1082 bcopy(p->ipopt_list, ip + 1, optlen); 1083 *phlen = sizeof(struct ip) + optlen; 1084 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2); 1085 ip->ip_len += optlen; 1086 return (m); 1087} 1088 1089/* 1090 * Copy options from ip to jp, 1091 * omitting those not copied during fragmentation. 1092 */ 1093int 1094ip_optcopy(ip, jp) 1095 struct ip *ip, *jp; 1096{ 1097 register u_char *cp, *dp; 1098 int opt, optlen, cnt; 1099 1100 cp = (u_char *)(ip + 1); 1101 dp = (u_char *)(jp + 1); 1102 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1103 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1104 opt = cp[0]; 1105 if (opt == IPOPT_EOL) 1106 break; 1107 if (opt == IPOPT_NOP) { 1108 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1109 *dp++ = IPOPT_NOP; 1110 optlen = 1; 1111 continue; 1112 } 1113#ifdef DIAGNOSTIC 1114 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1115 panic("malformed IPv4 option passed to ip_optcopy"); 1116#endif 1117 optlen = cp[IPOPT_OLEN]; 1118#ifdef DIAGNOSTIC 1119 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1120 panic("malformed IPv4 option passed to ip_optcopy"); 1121#endif 1122 /* bogus lengths should have been caught by ip_dooptions */ 1123 if (optlen > cnt) 1124 optlen = cnt; 1125 if (IPOPT_COPIED(opt)) { 1126 bcopy(cp, dp, optlen); 1127 dp += optlen; 1128 } 1129 } 1130 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1131 *dp++ = IPOPT_EOL; 1132 return (optlen); 1133} 1134 1135/* 1136 * IP socket option processing. 1137 */ 1138int 1139ip_ctloutput(so, sopt) 1140 struct socket *so; 1141 struct sockopt *sopt; 1142{ 1143 struct inpcb *inp = sotoinpcb(so); 1144 int error, optval; 1145 1146 error = optval = 0; 1147 if (sopt->sopt_level != IPPROTO_IP) { 1148 return (EINVAL); 1149 } 1150 1151 switch (sopt->sopt_dir) { 1152 case SOPT_SET: 1153 switch (sopt->sopt_name) { 1154 case IP_OPTIONS: 1155#ifdef notyet 1156 case IP_RETOPTS: 1157#endif 1158 { 1159 struct mbuf *m; 1160 if (sopt->sopt_valsize > MLEN) { 1161 error = EMSGSIZE; 1162 break; 1163 } 1164 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1165 if (m == 0) { 1166 error = ENOBUFS; 1167 break; 1168 } 1169 m->m_len = sopt->sopt_valsize; 1170 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1171 m->m_len); 1172 1173 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1174 m)); 1175 } 1176 1177 case IP_TOS: 1178 case IP_TTL: 1179 case IP_RECVOPTS: 1180 case IP_RECVRETOPTS: 1181 case IP_RECVDSTADDR: 1182 case IP_RECVIF: 1183 case IP_FAITH: 1184 error = sooptcopyin(sopt, &optval, sizeof optval, 1185 sizeof optval); 1186 if (error) 1187 break; 1188 1189 switch (sopt->sopt_name) { 1190 case IP_TOS: 1191 inp->inp_ip_tos = optval; 1192 break; 1193 1194 case IP_TTL: 1195 inp->inp_ip_ttl = optval; 1196 break; 1197#define OPTSET(bit) \ 1198 if (optval) \ 1199 inp->inp_flags |= bit; \ 1200 else \ 1201 inp->inp_flags &= ~bit; 1202 1203 case IP_RECVOPTS: 1204 OPTSET(INP_RECVOPTS); 1205 break; 1206 1207 case IP_RECVRETOPTS: 1208 OPTSET(INP_RECVRETOPTS); 1209 break; 1210 1211 case IP_RECVDSTADDR: 1212 OPTSET(INP_RECVDSTADDR); 1213 break; 1214 1215 case IP_RECVIF: 1216 OPTSET(INP_RECVIF); 1217 break; 1218 1219 case IP_FAITH: 1220 OPTSET(INP_FAITH); 1221 break; 1222 } 1223 break; 1224#undef OPTSET 1225 1226 case IP_MULTICAST_IF: 1227 case IP_MULTICAST_VIF: 1228 case IP_MULTICAST_TTL: 1229 case IP_MULTICAST_LOOP: 1230 case IP_ADD_MEMBERSHIP: 1231 case IP_DROP_MEMBERSHIP: 1232 error = ip_setmoptions(sopt, &inp->inp_moptions); 1233 break; 1234 1235 case IP_PORTRANGE: 1236 error = sooptcopyin(sopt, &optval, sizeof optval, 1237 sizeof optval); 1238 if (error) 1239 break; 1240 1241 switch (optval) { 1242 case IP_PORTRANGE_DEFAULT: 1243 inp->inp_flags &= ~(INP_LOWPORT); 1244 inp->inp_flags &= ~(INP_HIGHPORT); 1245 break; 1246 1247 case IP_PORTRANGE_HIGH: 1248 inp->inp_flags &= ~(INP_LOWPORT); 1249 inp->inp_flags |= INP_HIGHPORT; 1250 break; 1251 1252 case IP_PORTRANGE_LOW: 1253 inp->inp_flags &= ~(INP_HIGHPORT); 1254 inp->inp_flags |= INP_LOWPORT; 1255 break; 1256 1257 default: 1258 error = EINVAL; 1259 break; 1260 } 1261 break; 1262 1263#ifdef IPSEC 1264 case IP_IPSEC_POLICY: 1265 { 1266 caddr_t req; 1267 size_t len = 0; 1268 int priv; 1269 struct mbuf *m; 1270 int optname; 1271 1272 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1273 break; 1274 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1275 break; 1276 priv = (sopt->sopt_td != NULL && 1277 suser_td(sopt->sopt_td) != 0) ? 0 : 1; 1278 req = mtod(m, caddr_t); 1279 len = m->m_len; 1280 optname = sopt->sopt_name; 1281 error = ipsec4_set_policy(inp, optname, req, len, priv); 1282 m_freem(m); 1283 break; 1284 } 1285#endif /*IPSEC*/ 1286 1287 default: 1288 error = ENOPROTOOPT; 1289 break; 1290 } 1291 break; 1292 1293 case SOPT_GET: 1294 switch (sopt->sopt_name) { 1295 case IP_OPTIONS: 1296 case IP_RETOPTS: 1297 if (inp->inp_options) 1298 error = sooptcopyout(sopt, 1299 mtod(inp->inp_options, 1300 char *), 1301 inp->inp_options->m_len); 1302 else 1303 sopt->sopt_valsize = 0; 1304 break; 1305 1306 case IP_TOS: 1307 case IP_TTL: 1308 case IP_RECVOPTS: 1309 case IP_RECVRETOPTS: 1310 case IP_RECVDSTADDR: 1311 case IP_RECVIF: 1312 case IP_PORTRANGE: 1313 case IP_FAITH: 1314 switch (sopt->sopt_name) { 1315 1316 case IP_TOS: 1317 optval = inp->inp_ip_tos; 1318 break; 1319 1320 case IP_TTL: 1321 optval = inp->inp_ip_ttl; 1322 break; 1323 1324#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1325 1326 case IP_RECVOPTS: 1327 optval = OPTBIT(INP_RECVOPTS); 1328 break; 1329 1330 case IP_RECVRETOPTS: 1331 optval = OPTBIT(INP_RECVRETOPTS); 1332 break; 1333 1334 case IP_RECVDSTADDR: 1335 optval = OPTBIT(INP_RECVDSTADDR); 1336 break; 1337 1338 case IP_RECVIF: 1339 optval = OPTBIT(INP_RECVIF); 1340 break; 1341 1342 case IP_PORTRANGE: 1343 if (inp->inp_flags & INP_HIGHPORT) 1344 optval = IP_PORTRANGE_HIGH; 1345 else if (inp->inp_flags & INP_LOWPORT) 1346 optval = IP_PORTRANGE_LOW; 1347 else 1348 optval = 0; 1349 break; 1350 1351 case IP_FAITH: 1352 optval = OPTBIT(INP_FAITH); 1353 break; 1354 } 1355 error = sooptcopyout(sopt, &optval, sizeof optval); 1356 break; 1357 1358 case IP_MULTICAST_IF: 1359 case IP_MULTICAST_VIF: 1360 case IP_MULTICAST_TTL: 1361 case IP_MULTICAST_LOOP: 1362 case IP_ADD_MEMBERSHIP: 1363 case IP_DROP_MEMBERSHIP: 1364 error = ip_getmoptions(sopt, inp->inp_moptions); 1365 break; 1366 1367#ifdef IPSEC 1368 case IP_IPSEC_POLICY: 1369 { 1370 struct mbuf *m = NULL; 1371 caddr_t req = NULL; 1372 size_t len = 0; 1373 1374 if (m != 0) { 1375 req = mtod(m, caddr_t); 1376 len = m->m_len; 1377 } 1378 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1379 if (error == 0) 1380 error = soopt_mcopyout(sopt, m); /* XXX */ 1381 if (error == 0) 1382 m_freem(m); 1383 break; 1384 } 1385#endif /*IPSEC*/ 1386 1387 default: 1388 error = ENOPROTOOPT; 1389 break; 1390 } 1391 break; 1392 } 1393 return (error); 1394} 1395 1396/* 1397 * Set up IP options in pcb for insertion in output packets. 1398 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1399 * with destination address if source routed. 1400 */ 1401static int 1402ip_pcbopts(optname, pcbopt, m) 1403 int optname; 1404 struct mbuf **pcbopt; 1405 register struct mbuf *m; 1406{ 1407 register int cnt, optlen; 1408 register u_char *cp; 1409 u_char opt; 1410 1411 /* turn off any old options */ 1412 if (*pcbopt) 1413 (void)m_free(*pcbopt); 1414 *pcbopt = 0; 1415 if (m == (struct mbuf *)0 || m->m_len == 0) { 1416 /* 1417 * Only turning off any previous options. 1418 */ 1419 if (m) 1420 (void)m_free(m); 1421 return (0); 1422 } 1423 1424 if (m->m_len % sizeof(int32_t)) 1425 goto bad; 1426 /* 1427 * IP first-hop destination address will be stored before 1428 * actual options; move other options back 1429 * and clear it when none present. 1430 */ 1431 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1432 goto bad; 1433 cnt = m->m_len; 1434 m->m_len += sizeof(struct in_addr); 1435 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1436 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt); 1437 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1438 1439 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1440 opt = cp[IPOPT_OPTVAL]; 1441 if (opt == IPOPT_EOL) 1442 break; 1443 if (opt == IPOPT_NOP) 1444 optlen = 1; 1445 else { 1446 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1447 goto bad; 1448 optlen = cp[IPOPT_OLEN]; 1449 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1450 goto bad; 1451 } 1452 switch (opt) { 1453 1454 default: 1455 break; 1456 1457 case IPOPT_LSRR: 1458 case IPOPT_SSRR: 1459 /* 1460 * user process specifies route as: 1461 * ->A->B->C->D 1462 * D must be our final destination (but we can't 1463 * check that since we may not have connected yet). 1464 * A is first hop destination, which doesn't appear in 1465 * actual IP option, but is stored before the options. 1466 */ 1467 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1468 goto bad; 1469 m->m_len -= sizeof(struct in_addr); 1470 cnt -= sizeof(struct in_addr); 1471 optlen -= sizeof(struct in_addr); 1472 cp[IPOPT_OLEN] = optlen; 1473 /* 1474 * Move first hop before start of options. 1475 */ 1476 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1477 sizeof(struct in_addr)); 1478 /* 1479 * Then copy rest of options back 1480 * to close up the deleted entry. 1481 */ 1482 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] + 1483 sizeof(struct in_addr)), 1484 (caddr_t)&cp[IPOPT_OFFSET+1], 1485 (unsigned)cnt + sizeof(struct in_addr)); 1486 break; 1487 } 1488 } 1489 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1490 goto bad; 1491 *pcbopt = m; 1492 return (0); 1493 1494bad: 1495 (void)m_free(m); 1496 return (EINVAL); 1497} 1498 1499/* 1500 * XXX 1501 * The whole multicast option thing needs to be re-thought. 1502 * Several of these options are equally applicable to non-multicast 1503 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1504 * standard option (IP_TTL). 1505 */ 1506 1507/* 1508 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1509 */ 1510static struct ifnet * 1511ip_multicast_if(a, ifindexp) 1512 struct in_addr *a; 1513 int *ifindexp; 1514{ 1515 int ifindex; 1516 struct ifnet *ifp; 1517 1518 if (ifindexp) 1519 *ifindexp = 0; 1520 if (ntohl(a->s_addr) >> 24 == 0) { 1521 ifindex = ntohl(a->s_addr) & 0xffffff; 1522 if (ifindex < 0 || if_index < ifindex) 1523 return NULL; 1524 ifp = ifnet_byindex(ifindex); 1525 if (ifindexp) 1526 *ifindexp = ifindex; 1527 } else { 1528 INADDR_TO_IFP(*a, ifp); 1529 } 1530 return ifp; 1531} 1532 1533/* 1534 * Set the IP multicast options in response to user setsockopt(). 1535 */ 1536static int 1537ip_setmoptions(sopt, imop) 1538 struct sockopt *sopt; 1539 struct ip_moptions **imop; 1540{ 1541 int error = 0; 1542 int i; 1543 struct in_addr addr; 1544 struct ip_mreq mreq; 1545 struct ifnet *ifp; 1546 struct ip_moptions *imo = *imop; 1547 struct route ro; 1548 struct sockaddr_in *dst; 1549 int ifindex; 1550 int s; 1551 1552 if (imo == NULL) { 1553 /* 1554 * No multicast option buffer attached to the pcb; 1555 * allocate one and initialize to default values. 1556 */ 1557 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1558 M_WAITOK); 1559 1560 if (imo == NULL) 1561 return (ENOBUFS); 1562 *imop = imo; 1563 imo->imo_multicast_ifp = NULL; 1564 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1565 imo->imo_multicast_vif = -1; 1566 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1567 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1568 imo->imo_num_memberships = 0; 1569 } 1570 1571 switch (sopt->sopt_name) { 1572 /* store an index number for the vif you wanna use in the send */ 1573 case IP_MULTICAST_VIF: 1574 if (legal_vif_num == 0) { 1575 error = EOPNOTSUPP; 1576 break; 1577 } 1578 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1579 if (error) 1580 break; 1581 if (!legal_vif_num(i) && (i != -1)) { 1582 error = EINVAL; 1583 break; 1584 } 1585 imo->imo_multicast_vif = i; 1586 break; 1587 1588 case IP_MULTICAST_IF: 1589 /* 1590 * Select the interface for outgoing multicast packets. 1591 */ 1592 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1593 if (error) 1594 break; 1595 /* 1596 * INADDR_ANY is used to remove a previous selection. 1597 * When no interface is selected, a default one is 1598 * chosen every time a multicast packet is sent. 1599 */ 1600 if (addr.s_addr == INADDR_ANY) { 1601 imo->imo_multicast_ifp = NULL; 1602 break; 1603 } 1604 /* 1605 * The selected interface is identified by its local 1606 * IP address. Find the interface and confirm that 1607 * it supports multicasting. 1608 */ 1609 s = splimp(); 1610 ifp = ip_multicast_if(&addr, &ifindex); 1611 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1612 splx(s); 1613 error = EADDRNOTAVAIL; 1614 break; 1615 } 1616 imo->imo_multicast_ifp = ifp; 1617 if (ifindex) 1618 imo->imo_multicast_addr = addr; 1619 else 1620 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1621 splx(s); 1622 break; 1623 1624 case IP_MULTICAST_TTL: 1625 /* 1626 * Set the IP time-to-live for outgoing multicast packets. 1627 * The original multicast API required a char argument, 1628 * which is inconsistent with the rest of the socket API. 1629 * We allow either a char or an int. 1630 */ 1631 if (sopt->sopt_valsize == 1) { 1632 u_char ttl; 1633 error = sooptcopyin(sopt, &ttl, 1, 1); 1634 if (error) 1635 break; 1636 imo->imo_multicast_ttl = ttl; 1637 } else { 1638 u_int ttl; 1639 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1640 sizeof ttl); 1641 if (error) 1642 break; 1643 if (ttl > 255) 1644 error = EINVAL; 1645 else 1646 imo->imo_multicast_ttl = ttl; 1647 } 1648 break; 1649 1650 case IP_MULTICAST_LOOP: 1651 /* 1652 * Set the loopback flag for outgoing multicast packets. 1653 * Must be zero or one. The original multicast API required a 1654 * char argument, which is inconsistent with the rest 1655 * of the socket API. We allow either a char or an int. 1656 */ 1657 if (sopt->sopt_valsize == 1) { 1658 u_char loop; 1659 error = sooptcopyin(sopt, &loop, 1, 1); 1660 if (error) 1661 break; 1662 imo->imo_multicast_loop = !!loop; 1663 } else { 1664 u_int loop; 1665 error = sooptcopyin(sopt, &loop, sizeof loop, 1666 sizeof loop); 1667 if (error) 1668 break; 1669 imo->imo_multicast_loop = !!loop; 1670 } 1671 break; 1672 1673 case IP_ADD_MEMBERSHIP: 1674 /* 1675 * Add a multicast group membership. 1676 * Group must be a valid IP multicast address. 1677 */ 1678 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1679 if (error) 1680 break; 1681 1682 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1683 error = EINVAL; 1684 break; 1685 } 1686 s = splimp(); 1687 /* 1688 * If no interface address was provided, use the interface of 1689 * the route to the given multicast address. 1690 */ 1691 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1692 bzero((caddr_t)&ro, sizeof(ro)); 1693 dst = (struct sockaddr_in *)&ro.ro_dst; 1694 dst->sin_len = sizeof(*dst); 1695 dst->sin_family = AF_INET; 1696 dst->sin_addr = mreq.imr_multiaddr; 1697 rtalloc(&ro); 1698 if (ro.ro_rt == NULL) { 1699 error = EADDRNOTAVAIL; 1700 splx(s); 1701 break; 1702 } 1703 ifp = ro.ro_rt->rt_ifp; 1704 rtfree(ro.ro_rt); 1705 } 1706 else { 1707 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1708 } 1709 1710 /* 1711 * See if we found an interface, and confirm that it 1712 * supports multicast. 1713 */ 1714 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1715 error = EADDRNOTAVAIL; 1716 splx(s); 1717 break; 1718 } 1719 /* 1720 * See if the membership already exists or if all the 1721 * membership slots are full. 1722 */ 1723 for (i = 0; i < imo->imo_num_memberships; ++i) { 1724 if (imo->imo_membership[i]->inm_ifp == ifp && 1725 imo->imo_membership[i]->inm_addr.s_addr 1726 == mreq.imr_multiaddr.s_addr) 1727 break; 1728 } 1729 if (i < imo->imo_num_memberships) { 1730 error = EADDRINUSE; 1731 splx(s); 1732 break; 1733 } 1734 if (i == IP_MAX_MEMBERSHIPS) { 1735 error = ETOOMANYREFS; 1736 splx(s); 1737 break; 1738 } 1739 /* 1740 * Everything looks good; add a new record to the multicast 1741 * address list for the given interface. 1742 */ 1743 if ((imo->imo_membership[i] = 1744 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1745 error = ENOBUFS; 1746 splx(s); 1747 break; 1748 } 1749 ++imo->imo_num_memberships; 1750 splx(s); 1751 break; 1752 1753 case IP_DROP_MEMBERSHIP: 1754 /* 1755 * Drop a multicast group membership. 1756 * Group must be a valid IP multicast address. 1757 */ 1758 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1759 if (error) 1760 break; 1761 1762 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1763 error = EINVAL; 1764 break; 1765 } 1766 1767 s = splimp(); 1768 /* 1769 * If an interface address was specified, get a pointer 1770 * to its ifnet structure. 1771 */ 1772 if (mreq.imr_interface.s_addr == INADDR_ANY) 1773 ifp = NULL; 1774 else { 1775 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1776 if (ifp == NULL) { 1777 error = EADDRNOTAVAIL; 1778 splx(s); 1779 break; 1780 } 1781 } 1782 /* 1783 * Find the membership in the membership array. 1784 */ 1785 for (i = 0; i < imo->imo_num_memberships; ++i) { 1786 if ((ifp == NULL || 1787 imo->imo_membership[i]->inm_ifp == ifp) && 1788 imo->imo_membership[i]->inm_addr.s_addr == 1789 mreq.imr_multiaddr.s_addr) 1790 break; 1791 } 1792 if (i == imo->imo_num_memberships) { 1793 error = EADDRNOTAVAIL; 1794 splx(s); 1795 break; 1796 } 1797 /* 1798 * Give up the multicast address record to which the 1799 * membership points. 1800 */ 1801 in_delmulti(imo->imo_membership[i]); 1802 /* 1803 * Remove the gap in the membership array. 1804 */ 1805 for (++i; i < imo->imo_num_memberships; ++i) 1806 imo->imo_membership[i-1] = imo->imo_membership[i]; 1807 --imo->imo_num_memberships; 1808 splx(s); 1809 break; 1810 1811 default: 1812 error = EOPNOTSUPP; 1813 break; 1814 } 1815 1816 /* 1817 * If all options have default values, no need to keep the mbuf. 1818 */ 1819 if (imo->imo_multicast_ifp == NULL && 1820 imo->imo_multicast_vif == -1 && 1821 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1822 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1823 imo->imo_num_memberships == 0) { 1824 free(*imop, M_IPMOPTS); 1825 *imop = NULL; 1826 } 1827 1828 return (error); 1829} 1830 1831/* 1832 * Return the IP multicast options in response to user getsockopt(). 1833 */ 1834static int 1835ip_getmoptions(sopt, imo) 1836 struct sockopt *sopt; 1837 register struct ip_moptions *imo; 1838{ 1839 struct in_addr addr; 1840 struct in_ifaddr *ia; 1841 int error, optval; 1842 u_char coptval; 1843 1844 error = 0; 1845 switch (sopt->sopt_name) { 1846 case IP_MULTICAST_VIF: 1847 if (imo != NULL) 1848 optval = imo->imo_multicast_vif; 1849 else 1850 optval = -1; 1851 error = sooptcopyout(sopt, &optval, sizeof optval); 1852 break; 1853 1854 case IP_MULTICAST_IF: 1855 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1856 addr.s_addr = INADDR_ANY; 1857 else if (imo->imo_multicast_addr.s_addr) { 1858 /* return the value user has set */ 1859 addr = imo->imo_multicast_addr; 1860 } else { 1861 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1862 addr.s_addr = (ia == NULL) ? INADDR_ANY 1863 : IA_SIN(ia)->sin_addr.s_addr; 1864 } 1865 error = sooptcopyout(sopt, &addr, sizeof addr); 1866 break; 1867 1868 case IP_MULTICAST_TTL: 1869 if (imo == 0) 1870 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1871 else 1872 optval = coptval = imo->imo_multicast_ttl; 1873 if (sopt->sopt_valsize == 1) 1874 error = sooptcopyout(sopt, &coptval, 1); 1875 else 1876 error = sooptcopyout(sopt, &optval, sizeof optval); 1877 break; 1878 1879 case IP_MULTICAST_LOOP: 1880 if (imo == 0) 1881 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1882 else 1883 optval = coptval = imo->imo_multicast_loop; 1884 if (sopt->sopt_valsize == 1) 1885 error = sooptcopyout(sopt, &coptval, 1); 1886 else 1887 error = sooptcopyout(sopt, &optval, sizeof optval); 1888 break; 1889 1890 default: 1891 error = ENOPROTOOPT; 1892 break; 1893 } 1894 return (error); 1895} 1896 1897/* 1898 * Discard the IP multicast options. 1899 */ 1900void 1901ip_freemoptions(imo) 1902 register struct ip_moptions *imo; 1903{ 1904 register int i; 1905 1906 if (imo != NULL) { 1907 for (i = 0; i < imo->imo_num_memberships; ++i) 1908 in_delmulti(imo->imo_membership[i]); 1909 free(imo, M_IPMOPTS); 1910 } 1911} 1912 1913/* 1914 * Routine called from ip_output() to loop back a copy of an IP multicast 1915 * packet to the input queue of a specified interface. Note that this 1916 * calls the output routine of the loopback "driver", but with an interface 1917 * pointer that might NOT be a loopback interface -- evil, but easier than 1918 * replicating that code here. 1919 */ 1920static void 1921ip_mloopback(ifp, m, dst, hlen) 1922 struct ifnet *ifp; 1923 register struct mbuf *m; 1924 register struct sockaddr_in *dst; 1925 int hlen; 1926{ 1927 register struct ip *ip; 1928 struct mbuf *copym; 1929 1930 copym = m_copy(m, 0, M_COPYALL); 1931 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1932 copym = m_pullup(copym, hlen); 1933 if (copym != NULL) { 1934 /* 1935 * We don't bother to fragment if the IP length is greater 1936 * than the interface's MTU. Can this possibly matter? 1937 */ 1938 ip = mtod(copym, struct ip *); 1939 HTONS(ip->ip_len); 1940 HTONS(ip->ip_off); 1941 ip->ip_sum = 0; 1942 if (ip->ip_vhl == IP_VHL_BORING) { 1943 ip->ip_sum = in_cksum_hdr(ip); 1944 } else { 1945 ip->ip_sum = in_cksum(copym, hlen); 1946 } 1947 /* 1948 * NB: 1949 * It's not clear whether there are any lingering 1950 * reentrancy problems in other areas which might 1951 * be exposed by using ip_input directly (in 1952 * particular, everything which modifies the packet 1953 * in-place). Yet another option is using the 1954 * protosw directly to deliver the looped back 1955 * packet. For the moment, we'll err on the side 1956 * of safety by using if_simloop(). 1957 */ 1958#if 1 /* XXX */ 1959 if (dst->sin_family != AF_INET) { 1960 printf("ip_mloopback: bad address family %d\n", 1961 dst->sin_family); 1962 dst->sin_family = AF_INET; 1963 } 1964#endif 1965 1966#ifdef notdef 1967 copym->m_pkthdr.rcvif = ifp; 1968 ip_input(copym); 1969#else 1970 /* if the checksum hasn't been computed, mark it as valid */ 1971 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1972 copym->m_pkthdr.csum_flags |= 1973 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1974 copym->m_pkthdr.csum_data = 0xffff; 1975 } 1976 if_simloop(ifp, copym, dst->sin_family, 0); 1977#endif 1978 } 1979} 1980