ip_output.c revision 1.149.2.4
1 /* $NetBSD: ip_output.c,v 1.149.2.4 2007/01/28 19:37:20 tron Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69 /* 70 * Copyright (c) 1982, 1986, 1988, 1990, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. Neither the name of the University nor the names of its contributors 82 * may be used to endorse or promote products derived from this software 83 * without specific prior written permission. 84 * 85 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 86 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 87 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 88 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 89 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 90 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 91 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 92 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 93 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 94 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 95 * SUCH DAMAGE. 96 * 97 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 98 */ 99 100 #include <sys/cdefs.h> 101 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.149.2.4 2007/01/28 19:37:20 tron Exp $"); 102 103 #include "opt_pfil_hooks.h" 104 #include "opt_inet.h" 105 #include "opt_ipsec.h" 106 #include "opt_mrouting.h" 107 108 #include <sys/param.h> 109 #include <sys/malloc.h> 110 #include <sys/mbuf.h> 111 #include <sys/errno.h> 112 #include <sys/protosw.h> 113 #include <sys/socket.h> 114 #include <sys/socketvar.h> 115 #ifdef FAST_IPSEC 116 #include <sys/domain.h> 117 #endif 118 #include <sys/systm.h> 119 #include <sys/proc.h> 120 121 #include <net/if.h> 122 #include <net/route.h> 123 #include <net/pfil.h> 124 125 #include <netinet/in.h> 126 #include <netinet/in_systm.h> 127 #include <netinet/ip.h> 128 #include <netinet/in_pcb.h> 129 #include <netinet/in_var.h> 130 #include <netinet/ip_var.h> 131 132 #ifdef MROUTING 133 #include <netinet/ip_mroute.h> 134 #endif 135 136 #include <machine/stdarg.h> 137 138 #ifdef IPSEC 139 #include <netinet6/ipsec.h> 140 #include <netkey/key.h> 141 #include <netkey/key_debug.h> 142 #ifdef IPSEC_NAT_T 143 #include <netinet/udp.h> 144 #endif 145 #endif /*IPSEC*/ 146 147 #ifdef FAST_IPSEC 148 #include <netipsec/ipsec.h> 149 #include <netipsec/key.h> 150 #include <netipsec/xform.h> 151 #endif /* FAST_IPSEC*/ 152 153 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 154 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 155 static void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *); 156 static int ip_getoptval(struct mbuf *, u_int8_t *, u_int); 157 158 #ifdef PFIL_HOOKS 159 extern struct pfil_head inet_pfil_hook; /* XXX */ 160 #endif 161 162 int udp_do_loopback_cksum = 0; 163 int tcp_do_loopback_cksum = 0; 164 int ip_do_loopback_cksum = 0; 165 166 #define IN_NEED_CHECKSUM(ifp, csum_flags) \ 167 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 168 (((csum_flags) & M_CSUM_UDPv4) != 0 && udp_do_loopback_cksum) || \ 169 (((csum_flags) & M_CSUM_TCPv4) != 0 && tcp_do_loopback_cksum) || \ 170 (((csum_flags) & M_CSUM_IPv4) != 0 && ip_do_loopback_cksum))) 171 172 /* 173 * IP output. The packet in mbuf chain m contains a skeletal IP 174 * header (with len, off, ttl, proto, tos, src, dst). 175 * The mbuf chain containing the packet will be freed. 176 * The mbuf opt, if present, will not be freed. 177 */ 178 int 179 ip_output(struct mbuf *m0, ...) 180 { 181 struct ip *ip; 182 struct ifnet *ifp; 183 struct mbuf *m = m0; 184 int hlen = sizeof (struct ip); 185 int len, error = 0; 186 struct route iproute; 187 struct sockaddr_in *dst; 188 struct in_ifaddr *ia; 189 struct mbuf *opt; 190 struct route *ro; 191 int flags, sw_csum; 192 int *mtu_p; 193 u_long mtu; 194 struct ip_moptions *imo; 195 struct socket *so; 196 va_list ap; 197 #ifdef IPSEC 198 struct secpolicy *sp = NULL; 199 #ifdef IPSEC_NAT_T 200 int natt_frag = 0; 201 #endif 202 #endif /*IPSEC*/ 203 #ifdef FAST_IPSEC 204 struct inpcb *inp; 205 struct m_tag *mtag; 206 struct secpolicy *sp = NULL; 207 struct tdb_ident *tdbi; 208 int s; 209 #endif 210 u_int16_t ip_len; 211 212 len = 0; 213 va_start(ap, m0); 214 opt = va_arg(ap, struct mbuf *); 215 ro = va_arg(ap, struct route *); 216 flags = va_arg(ap, int); 217 imo = va_arg(ap, struct ip_moptions *); 218 so = va_arg(ap, struct socket *); 219 if (flags & IP_RETURNMTU) 220 mtu_p = va_arg(ap, int *); 221 else 222 mtu_p = NULL; 223 va_end(ap); 224 225 MCLAIM(m, &ip_tx_mowner); 226 #ifdef FAST_IPSEC 227 if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET) 228 inp = (struct inpcb *)so->so_pcb; 229 else 230 inp = NULL; 231 #endif /* FAST_IPSEC */ 232 233 #ifdef DIAGNOSTIC 234 if ((m->m_flags & M_PKTHDR) == 0) 235 panic("ip_output no HDR"); 236 #endif 237 if (opt) { 238 m = ip_insertoptions(m, opt, &len); 239 if (len >= sizeof(struct ip)) 240 hlen = len; 241 } 242 ip = mtod(m, struct ip *); 243 /* 244 * Fill in IP header. 245 */ 246 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 247 ip->ip_v = IPVERSION; 248 ip->ip_off = htons(0); 249 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 250 ip->ip_id = ip_newid(); 251 } else { 252 253 /* 254 * TSO capable interfaces (typically?) increment 255 * ip_id for each segment. 256 * "allocate" enough ids here to increase the chance 257 * for them to be unique. 258 * 259 * note that the following calculation is not 260 * needed to be precise. wasting some ip_id is fine. 261 */ 262 263 unsigned int segsz = m->m_pkthdr.segsz; 264 unsigned int datasz = ntohs(ip->ip_len) - hlen; 265 unsigned int num = howmany(datasz, segsz); 266 267 ip->ip_id = ip_newid_range(num); 268 } 269 ip->ip_hl = hlen >> 2; 270 ipstat.ips_localout++; 271 } else { 272 hlen = ip->ip_hl << 2; 273 } 274 /* 275 * Route packet. 276 */ 277 if (ro == 0) { 278 ro = &iproute; 279 bzero((caddr_t)ro, sizeof (*ro)); 280 } 281 dst = satosin(&ro->ro_dst); 282 /* 283 * If there is a cached route, 284 * check that it is to the same destination 285 * and is still up. If not, free it and try again. 286 * The address family should also be checked in case of sharing the 287 * cache with IPv6. 288 */ 289 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 290 dst->sin_family != AF_INET || 291 !in_hosteq(dst->sin_addr, ip->ip_dst))) { 292 RTFREE(ro->ro_rt); 293 ro->ro_rt = (struct rtentry *)0; 294 } 295 if (ro->ro_rt == 0) { 296 bzero(dst, sizeof(*dst)); 297 dst->sin_family = AF_INET; 298 dst->sin_len = sizeof(*dst); 299 dst->sin_addr = ip->ip_dst; 300 } 301 /* 302 * If routing to interface only, 303 * short circuit routing lookup. 304 */ 305 if (flags & IP_ROUTETOIF) { 306 if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) { 307 ipstat.ips_noroute++; 308 error = ENETUNREACH; 309 goto bad; 310 } 311 ifp = ia->ia_ifp; 312 mtu = ifp->if_mtu; 313 ip->ip_ttl = 1; 314 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 315 ip->ip_dst.s_addr == INADDR_BROADCAST) && 316 imo != NULL && imo->imo_multicast_ifp != NULL) { 317 ifp = imo->imo_multicast_ifp; 318 mtu = ifp->if_mtu; 319 IFP_TO_IA(ifp, ia); 320 } else { 321 if (ro->ro_rt == 0) 322 rtalloc(ro); 323 if (ro->ro_rt == 0) { 324 ipstat.ips_noroute++; 325 error = EHOSTUNREACH; 326 goto bad; 327 } 328 ia = ifatoia(ro->ro_rt->rt_ifa); 329 ifp = ro->ro_rt->rt_ifp; 330 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 331 mtu = ifp->if_mtu; 332 ro->ro_rt->rt_use++; 333 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 334 dst = satosin(ro->ro_rt->rt_gateway); 335 } 336 if (IN_MULTICAST(ip->ip_dst.s_addr) || 337 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 338 struct in_multi *inm; 339 340 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 341 M_BCAST : M_MCAST; 342 /* 343 * IP destination address is multicast. Make sure "dst" 344 * still points to the address in "ro". (It may have been 345 * changed to point to a gateway address, above.) 346 */ 347 dst = satosin(&ro->ro_dst); 348 /* 349 * See if the caller provided any multicast options 350 */ 351 if (imo != NULL) 352 ip->ip_ttl = imo->imo_multicast_ttl; 353 else 354 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 355 356 /* 357 * if we don't know the outgoing ifp yet, we can't generate 358 * output 359 */ 360 if (!ifp) { 361 ipstat.ips_noroute++; 362 error = ENETUNREACH; 363 goto bad; 364 } 365 366 /* 367 * If the packet is multicast or broadcast, confirm that 368 * the outgoing interface can transmit it. 369 */ 370 if (((m->m_flags & M_MCAST) && 371 (ifp->if_flags & IFF_MULTICAST) == 0) || 372 ((m->m_flags & M_BCAST) && 373 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 374 ipstat.ips_noroute++; 375 error = ENETUNREACH; 376 goto bad; 377 } 378 /* 379 * If source address not specified yet, use an address 380 * of outgoing interface. 381 */ 382 if (in_nullhost(ip->ip_src)) { 383 struct in_ifaddr *ia; 384 385 IFP_TO_IA(ifp, ia); 386 if (!ia) { 387 error = EADDRNOTAVAIL; 388 goto bad; 389 } 390 ip->ip_src = ia->ia_addr.sin_addr; 391 } 392 393 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 394 if (inm != NULL && 395 (imo == NULL || imo->imo_multicast_loop)) { 396 /* 397 * If we belong to the destination multicast group 398 * on the outgoing interface, and the caller did not 399 * forbid loopback, loop back a copy. 400 */ 401 ip_mloopback(ifp, m, dst); 402 } 403 #ifdef MROUTING 404 else { 405 /* 406 * If we are acting as a multicast router, perform 407 * multicast forwarding as if the packet had just 408 * arrived on the interface to which we are about 409 * to send. The multicast forwarding function 410 * recursively calls this function, using the 411 * IP_FORWARDING flag to prevent infinite recursion. 412 * 413 * Multicasts that are looped back by ip_mloopback(), 414 * above, will be forwarded by the ip_input() routine, 415 * if necessary. 416 */ 417 extern struct socket *ip_mrouter; 418 419 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 420 if (ip_mforward(m, ifp) != 0) { 421 m_freem(m); 422 goto done; 423 } 424 } 425 } 426 #endif 427 /* 428 * Multicasts with a time-to-live of zero may be looped- 429 * back, above, but must not be transmitted on a network. 430 * Also, multicasts addressed to the loopback interface 431 * are not sent -- the above call to ip_mloopback() will 432 * loop back a copy if this host actually belongs to the 433 * destination group on the loopback interface. 434 */ 435 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 436 m_freem(m); 437 goto done; 438 } 439 440 goto sendit; 441 } 442 #ifndef notdef 443 /* 444 * If source address not specified yet, use address 445 * of outgoing interface. 446 */ 447 if (in_nullhost(ip->ip_src)) 448 ip->ip_src = ia->ia_addr.sin_addr; 449 #endif 450 451 /* 452 * packets with Class-D address as source are not valid per 453 * RFC 1112 454 */ 455 if (IN_MULTICAST(ip->ip_src.s_addr)) { 456 ipstat.ips_odropped++; 457 error = EADDRNOTAVAIL; 458 goto bad; 459 } 460 461 /* 462 * Look for broadcast address and 463 * and verify user is allowed to send 464 * such a packet. 465 */ 466 if (in_broadcast(dst->sin_addr, ifp)) { 467 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 468 error = EADDRNOTAVAIL; 469 goto bad; 470 } 471 if ((flags & IP_ALLOWBROADCAST) == 0) { 472 error = EACCES; 473 goto bad; 474 } 475 /* don't allow broadcast messages to be fragmented */ 476 if (ntohs(ip->ip_len) > ifp->if_mtu) { 477 error = EMSGSIZE; 478 goto bad; 479 } 480 m->m_flags |= M_BCAST; 481 } else 482 m->m_flags &= ~M_BCAST; 483 484 sendit: 485 /* 486 * If we're doing Path MTU Discovery, we need to set DF unless 487 * the route's MTU is locked. 488 */ 489 if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL && 490 (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 491 ip->ip_off |= htons(IP_DF); 492 493 /* Remember the current ip_len */ 494 ip_len = ntohs(ip->ip_len); 495 496 #ifdef IPSEC 497 /* get SP for this packet */ 498 if (so == NULL) 499 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 500 flags, &error); 501 else { 502 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 503 IPSEC_DIR_OUTBOUND)) 504 goto skip_ipsec; 505 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 506 } 507 508 if (sp == NULL) { 509 ipsecstat.out_inval++; 510 goto bad; 511 } 512 513 error = 0; 514 515 /* check policy */ 516 switch (sp->policy) { 517 case IPSEC_POLICY_DISCARD: 518 /* 519 * This packet is just discarded. 520 */ 521 ipsecstat.out_polvio++; 522 goto bad; 523 524 case IPSEC_POLICY_BYPASS: 525 case IPSEC_POLICY_NONE: 526 /* no need to do IPsec. */ 527 goto skip_ipsec; 528 529 case IPSEC_POLICY_IPSEC: 530 if (sp->req == NULL) { 531 /* XXX should be panic ? */ 532 printf("ip_output: No IPsec request specified.\n"); 533 error = EINVAL; 534 goto bad; 535 } 536 break; 537 538 case IPSEC_POLICY_ENTRUST: 539 default: 540 printf("ip_output: Invalid policy found. %d\n", sp->policy); 541 } 542 543 #ifdef IPSEC_NAT_T 544 /* 545 * NAT-T ESP fragmentation: don't do IPSec processing now, 546 * we'll do it on each fragmented packet. 547 */ 548 if (sp->req->sav && 549 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 550 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 551 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 552 natt_frag = 1; 553 mtu = sp->req->sav->esp_frag; 554 goto skip_ipsec; 555 } 556 } 557 #endif /* IPSEC_NAT_T */ 558 559 /* 560 * ipsec4_output() expects ip_len and ip_off in network 561 * order. They have been set to network order above. 562 */ 563 564 { 565 struct ipsec_output_state state; 566 bzero(&state, sizeof(state)); 567 state.m = m; 568 if (flags & IP_ROUTETOIF) { 569 state.ro = &iproute; 570 bzero(&iproute, sizeof(iproute)); 571 } else 572 state.ro = ro; 573 state.dst = (struct sockaddr *)dst; 574 575 /* 576 * We can't defer the checksum of payload data if 577 * we're about to encrypt/authenticate it. 578 * 579 * XXX When we support crypto offloading functions of 580 * XXX network interfaces, we need to reconsider this, 581 * XXX since it's likely that they'll support checksumming, 582 * XXX as well. 583 */ 584 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 585 in_delayed_cksum(m); 586 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 587 } 588 589 error = ipsec4_output(&state, sp, flags); 590 591 m = state.m; 592 if (flags & IP_ROUTETOIF) { 593 /* 594 * if we have tunnel mode SA, we may need to ignore 595 * IP_ROUTETOIF. 596 */ 597 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 598 flags &= ~IP_ROUTETOIF; 599 ro = state.ro; 600 } 601 } else 602 ro = state.ro; 603 dst = (struct sockaddr_in *)state.dst; 604 if (error) { 605 /* mbuf is already reclaimed in ipsec4_output. */ 606 m0 = NULL; 607 switch (error) { 608 case EHOSTUNREACH: 609 case ENETUNREACH: 610 case EMSGSIZE: 611 case ENOBUFS: 612 case ENOMEM: 613 break; 614 default: 615 printf("ip4_output (ipsec): error code %d\n", error); 616 /*fall through*/ 617 case ENOENT: 618 /* don't show these error codes to the user */ 619 error = 0; 620 break; 621 } 622 goto bad; 623 } 624 625 /* be sure to update variables that are affected by ipsec4_output() */ 626 ip = mtod(m, struct ip *); 627 hlen = ip->ip_hl << 2; 628 ip_len = ntohs(ip->ip_len); 629 630 if (ro->ro_rt == NULL) { 631 if ((flags & IP_ROUTETOIF) == 0) { 632 printf("ip_output: " 633 "can't update route after IPsec processing\n"); 634 error = EHOSTUNREACH; /*XXX*/ 635 goto bad; 636 } 637 } else { 638 /* nobody uses ia beyond here */ 639 if (state.encap) { 640 ifp = ro->ro_rt->rt_ifp; 641 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 642 mtu = ifp->if_mtu; 643 } 644 } 645 } 646 skip_ipsec: 647 #endif /*IPSEC*/ 648 #ifdef FAST_IPSEC 649 /* 650 * Check the security policy (SP) for the packet and, if 651 * required, do IPsec-related processing. There are two 652 * cases here; the first time a packet is sent through 653 * it will be untagged and handled by ipsec4_checkpolicy. 654 * If the packet is resubmitted to ip_output (e.g. after 655 * AH, ESP, etc. processing), there will be a tag to bypass 656 * the lookup and related policy checking. 657 */ 658 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 659 s = splsoftnet(); 660 if (mtag != NULL) { 661 tdbi = (struct tdb_ident *)(mtag + 1); 662 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 663 if (sp == NULL) 664 error = -EINVAL; /* force silent drop */ 665 m_tag_delete(m, mtag); 666 } else { 667 if (inp != NULL && 668 IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND)) 669 goto spd_done; 670 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 671 &error, inp); 672 } 673 /* 674 * There are four return cases: 675 * sp != NULL apply IPsec policy 676 * sp == NULL, error == 0 no IPsec handling needed 677 * sp == NULL, error == -EINVAL discard packet w/o error 678 * sp == NULL, error != 0 discard packet, report error 679 */ 680 if (sp != NULL) { 681 /* Loop detection, check if ipsec processing already done */ 682 IPSEC_ASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 683 for (mtag = m_tag_first(m); mtag != NULL; 684 mtag = m_tag_next(m, mtag)) { 685 #ifdef MTAG_ABI_COMPAT 686 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 687 continue; 688 #endif 689 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 690 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 691 continue; 692 /* 693 * Check if policy has an SA associated with it. 694 * This can happen when an SP has yet to acquire 695 * an SA; e.g. on first reference. If it occurs, 696 * then we let ipsec4_process_packet do its thing. 697 */ 698 if (sp->req->sav == NULL) 699 break; 700 tdbi = (struct tdb_ident *)(mtag + 1); 701 if (tdbi->spi == sp->req->sav->spi && 702 tdbi->proto == sp->req->sav->sah->saidx.proto && 703 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 704 sizeof (union sockaddr_union)) == 0) { 705 /* 706 * No IPsec processing is needed, free 707 * reference to SP. 708 * 709 * NB: null pointer to avoid free at 710 * done: below. 711 */ 712 KEY_FREESP(&sp), sp = NULL; 713 splx(s); 714 goto spd_done; 715 } 716 } 717 718 /* 719 * Do delayed checksums now because we send before 720 * this is done in the normal processing path. 721 */ 722 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 723 in_delayed_cksum(m); 724 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 725 } 726 727 #ifdef __FreeBSD__ 728 ip->ip_len = htons(ip->ip_len); 729 ip->ip_off = htons(ip->ip_off); 730 #endif 731 732 /* NB: callee frees mbuf */ 733 error = ipsec4_process_packet(m, sp->req, flags, 0); 734 /* 735 * Preserve KAME behaviour: ENOENT can be returned 736 * when an SA acquire is in progress. Don't propagate 737 * this to user-level; it confuses applications. 738 * 739 * XXX this will go away when the SADB is redone. 740 */ 741 if (error == ENOENT) 742 error = 0; 743 splx(s); 744 goto done; 745 } else { 746 splx(s); 747 748 if (error != 0) { 749 /* 750 * Hack: -EINVAL is used to signal that a packet 751 * should be silently discarded. This is typically 752 * because we asked key management for an SA and 753 * it was delayed (e.g. kicked up to IKE). 754 */ 755 if (error == -EINVAL) 756 error = 0; 757 goto bad; 758 } else { 759 /* No IPsec processing for this packet. */ 760 } 761 #ifdef notyet 762 /* 763 * If deferred crypto processing is needed, check that 764 * the interface supports it. 765 */ 766 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 767 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 768 /* notify IPsec to do its own crypto */ 769 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 770 error = EHOSTUNREACH; 771 goto bad; 772 } 773 #endif 774 } 775 spd_done: 776 #endif /* FAST_IPSEC */ 777 778 #ifdef PFIL_HOOKS 779 /* 780 * Run through list of hooks for output packets. 781 */ 782 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 783 goto done; 784 if (m == NULL) 785 goto done; 786 787 ip = mtod(m, struct ip *); 788 hlen = ip->ip_hl << 2; 789 ip_len = ntohs(ip->ip_len); 790 #endif /* PFIL_HOOKS */ 791 792 m->m_pkthdr.csum_data |= hlen << 16; 793 794 #if IFA_STATS 795 /* 796 * search for the source address structure to 797 * maintain output statistics. 798 */ 799 INADDR_TO_IA(ip->ip_src, ia); 800 #endif 801 802 /* Maybe skip checksums on loopback interfaces. */ 803 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 804 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 805 } 806 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 807 /* 808 * If small enough for mtu of path, or if using TCP segmentation 809 * offload, can just send directly. 810 */ 811 if (ip_len <= mtu || 812 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 813 #if IFA_STATS 814 if (ia) 815 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 816 #endif 817 /* 818 * Always initialize the sum to 0! Some HW assisted 819 * checksumming requires this. 820 */ 821 ip->ip_sum = 0; 822 823 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 824 /* 825 * Perform any checksums that the hardware can't do 826 * for us. 827 * 828 * XXX Does any hardware require the {th,uh}_sum 829 * XXX fields to be 0? 830 */ 831 if (sw_csum & M_CSUM_IPv4) { 832 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 833 ip->ip_sum = in_cksum(m, hlen); 834 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 835 } 836 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 837 if (IN_NEED_CHECKSUM(ifp, 838 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 839 in_delayed_cksum(m); 840 } 841 m->m_pkthdr.csum_flags &= 842 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 843 } 844 } 845 846 #ifdef IPSEC 847 /* clean ipsec history once it goes out of the node */ 848 ipsec_delaux(m); 849 #endif 850 error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt); 851 goto done; 852 } 853 854 /* 855 * We can't use HW checksumming if we're about to 856 * to fragment the packet. 857 * 858 * XXX Some hardware can do this. 859 */ 860 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 861 if (IN_NEED_CHECKSUM(ifp, 862 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 863 in_delayed_cksum(m); 864 } 865 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 866 } 867 868 /* 869 * Too large for interface; fragment if possible. 870 * Must be able to put at least 8 bytes per fragment. 871 */ 872 if (ntohs(ip->ip_off) & IP_DF) { 873 if (flags & IP_RETURNMTU) 874 *mtu_p = mtu; 875 error = EMSGSIZE; 876 ipstat.ips_cantfrag++; 877 goto bad; 878 } 879 880 error = ip_fragment(m, ifp, mtu); 881 if (error) { 882 m = NULL; 883 goto bad; 884 } 885 886 for (; m; m = m0) { 887 m0 = m->m_nextpkt; 888 m->m_nextpkt = 0; 889 if (error == 0) { 890 #if IFA_STATS 891 if (ia) 892 ia->ia_ifa.ifa_data.ifad_outbytes += 893 ntohs(ip->ip_len); 894 #endif 895 #ifdef IPSEC 896 /* clean ipsec history once it goes out of the node */ 897 ipsec_delaux(m); 898 899 #ifdef IPSEC_NAT_T 900 /* 901 * If we get there, the packet has not been handeld by 902 * IPSec whereas it should have. Now that it has been 903 * fragmented, re-inject it in ip_output so that IPsec 904 * processing can occur. 905 */ 906 if (natt_frag) { 907 error = ip_output(m, opt, 908 ro, flags, imo, so, mtu_p); 909 } else 910 #endif /* IPSEC_NAT_T */ 911 #endif /* IPSEC */ 912 { 913 KASSERT((m->m_pkthdr.csum_flags & 914 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 915 error = (*ifp->if_output)(ifp, m, sintosa(dst), 916 ro->ro_rt); 917 } 918 } else 919 m_freem(m); 920 } 921 922 if (error == 0) 923 ipstat.ips_fragmented++; 924 done: 925 if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) { 926 RTFREE(ro->ro_rt); 927 ro->ro_rt = 0; 928 } 929 930 #ifdef IPSEC 931 if (sp != NULL) { 932 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 933 printf("DP ip_output call free SP:%p\n", sp)); 934 key_freesp(sp); 935 } 936 #endif /* IPSEC */ 937 #ifdef FAST_IPSEC 938 if (sp != NULL) 939 KEY_FREESP(&sp); 940 #endif /* FAST_IPSEC */ 941 942 return (error); 943 bad: 944 m_freem(m); 945 goto done; 946 } 947 948 int 949 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 950 { 951 struct ip *ip, *mhip; 952 struct mbuf *m0; 953 int len, hlen, off; 954 int mhlen, firstlen; 955 struct mbuf **mnext; 956 int sw_csum = m->m_pkthdr.csum_flags; 957 int fragments = 0; 958 int s; 959 int error = 0; 960 961 ip = mtod(m, struct ip *); 962 hlen = ip->ip_hl << 2; 963 if (ifp != NULL) 964 sw_csum &= ~ifp->if_csum_flags_tx; 965 966 len = (mtu - hlen) &~ 7; 967 if (len < 8) { 968 m_freem(m); 969 return (EMSGSIZE); 970 } 971 972 firstlen = len; 973 mnext = &m->m_nextpkt; 974 975 /* 976 * Loop through length of segment after first fragment, 977 * make new header and copy data of each part and link onto chain. 978 */ 979 m0 = m; 980 mhlen = sizeof (struct ip); 981 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 982 MGETHDR(m, M_DONTWAIT, MT_HEADER); 983 if (m == 0) { 984 error = ENOBUFS; 985 ipstat.ips_odropped++; 986 goto sendorfree; 987 } 988 MCLAIM(m, m0->m_owner); 989 *mnext = m; 990 mnext = &m->m_nextpkt; 991 m->m_data += max_linkhdr; 992 mhip = mtod(m, struct ip *); 993 *mhip = *ip; 994 /* we must inherit MCAST and BCAST flags */ 995 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 996 if (hlen > sizeof (struct ip)) { 997 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 998 mhip->ip_hl = mhlen >> 2; 999 } 1000 m->m_len = mhlen; 1001 mhip->ip_off = ((off - hlen) >> 3) + 1002 (ntohs(ip->ip_off) & ~IP_MF); 1003 if (ip->ip_off & htons(IP_MF)) 1004 mhip->ip_off |= IP_MF; 1005 if (off + len >= ntohs(ip->ip_len)) 1006 len = ntohs(ip->ip_len) - off; 1007 else 1008 mhip->ip_off |= IP_MF; 1009 HTONS(mhip->ip_off); 1010 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 1011 m->m_next = m_copy(m0, off, len); 1012 if (m->m_next == 0) { 1013 error = ENOBUFS; /* ??? */ 1014 ipstat.ips_odropped++; 1015 goto sendorfree; 1016 } 1017 m->m_pkthdr.len = mhlen + len; 1018 m->m_pkthdr.rcvif = (struct ifnet *)0; 1019 mhip->ip_sum = 0; 1020 if (sw_csum & M_CSUM_IPv4) { 1021 mhip->ip_sum = in_cksum(m, mhlen); 1022 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 1023 } else { 1024 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 1025 m->m_pkthdr.csum_data |= mhlen << 16; 1026 } 1027 ipstat.ips_ofragments++; 1028 fragments++; 1029 } 1030 /* 1031 * Update first fragment by trimming what's been copied out 1032 * and updating header, then send each fragment (in order). 1033 */ 1034 m = m0; 1035 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 1036 m->m_pkthdr.len = hlen + firstlen; 1037 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 1038 ip->ip_off |= htons(IP_MF); 1039 ip->ip_sum = 0; 1040 if (sw_csum & M_CSUM_IPv4) { 1041 ip->ip_sum = in_cksum(m, hlen); 1042 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 1043 } else { 1044 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 1045 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 1046 sizeof(struct ip)); 1047 } 1048 sendorfree: 1049 /* 1050 * If there is no room for all the fragments, don't queue 1051 * any of them. 1052 */ 1053 if (ifp != NULL) { 1054 s = splnet(); 1055 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 1056 error == 0) { 1057 error = ENOBUFS; 1058 ipstat.ips_odropped++; 1059 IFQ_INC_DROPS(&ifp->if_snd); 1060 } 1061 splx(s); 1062 } 1063 if (error) { 1064 for (m = m0; m; m = m0) { 1065 m0 = m->m_nextpkt; 1066 m->m_nextpkt = NULL; 1067 m_freem(m); 1068 } 1069 } 1070 return (error); 1071 } 1072 1073 /* 1074 * Process a delayed payload checksum calculation. 1075 */ 1076 void 1077 in_delayed_cksum(struct mbuf *m) 1078 { 1079 struct ip *ip; 1080 u_int16_t csum, offset; 1081 1082 ip = mtod(m, struct ip *); 1083 offset = ip->ip_hl << 2; 1084 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 1085 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 1086 csum = 0xffff; 1087 1088 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 1089 1090 if ((offset + sizeof(u_int16_t)) > m->m_len) { 1091 /* This happen when ip options were inserted 1092 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 1093 m->m_len, offset, ip->ip_p); 1094 */ 1095 m_copyback(m, offset, sizeof(csum), (caddr_t) &csum); 1096 } else 1097 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum; 1098 } 1099 1100 /* 1101 * Determine the maximum length of the options to be inserted; 1102 * we would far rather allocate too much space rather than too little. 1103 */ 1104 1105 u_int 1106 ip_optlen(struct inpcb *inp) 1107 { 1108 struct mbuf *m = inp->inp_options; 1109 1110 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 1111 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 1112 else 1113 return 0; 1114 } 1115 1116 1117 /* 1118 * Insert IP options into preformed packet. 1119 * Adjust IP destination as required for IP source routing, 1120 * as indicated by a non-zero in_addr at the start of the options. 1121 */ 1122 static struct mbuf * 1123 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 1124 { 1125 struct ipoption *p = mtod(opt, struct ipoption *); 1126 struct mbuf *n; 1127 struct ip *ip = mtod(m, struct ip *); 1128 unsigned optlen; 1129 1130 optlen = opt->m_len - sizeof(p->ipopt_dst); 1131 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 1132 return (m); /* XXX should fail */ 1133 if (!in_nullhost(p->ipopt_dst)) 1134 ip->ip_dst = p->ipopt_dst; 1135 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 1136 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1137 if (n == 0) 1138 return (m); 1139 MCLAIM(n, m->m_owner); 1140 M_COPY_PKTHDR(n, m); 1141 m_tag_delete_chain(m, NULL); 1142 m->m_flags &= ~M_PKTHDR; 1143 m->m_len -= sizeof(struct ip); 1144 m->m_data += sizeof(struct ip); 1145 n->m_next = m; 1146 m = n; 1147 m->m_len = optlen + sizeof(struct ip); 1148 m->m_data += max_linkhdr; 1149 bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1150 } else { 1151 m->m_data -= optlen; 1152 m->m_len += optlen; 1153 memmove(mtod(m, caddr_t), ip, sizeof(struct ip)); 1154 } 1155 m->m_pkthdr.len += optlen; 1156 ip = mtod(m, struct ip *); 1157 bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen); 1158 *phlen = sizeof(struct ip) + optlen; 1159 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1160 return (m); 1161 } 1162 1163 /* 1164 * Copy options from ip to jp, 1165 * omitting those not copied during fragmentation. 1166 */ 1167 int 1168 ip_optcopy(struct ip *ip, struct ip *jp) 1169 { 1170 u_char *cp, *dp; 1171 int opt, optlen, cnt; 1172 1173 cp = (u_char *)(ip + 1); 1174 dp = (u_char *)(jp + 1); 1175 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1176 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1177 opt = cp[0]; 1178 if (opt == IPOPT_EOL) 1179 break; 1180 if (opt == IPOPT_NOP) { 1181 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1182 *dp++ = IPOPT_NOP; 1183 optlen = 1; 1184 continue; 1185 } 1186 #ifdef DIAGNOSTIC 1187 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1188 panic("malformed IPv4 option passed to ip_optcopy"); 1189 #endif 1190 optlen = cp[IPOPT_OLEN]; 1191 #ifdef DIAGNOSTIC 1192 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1193 panic("malformed IPv4 option passed to ip_optcopy"); 1194 #endif 1195 /* bogus lengths should have been caught by ip_dooptions */ 1196 if (optlen > cnt) 1197 optlen = cnt; 1198 if (IPOPT_COPIED(opt)) { 1199 bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen); 1200 dp += optlen; 1201 } 1202 } 1203 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1204 *dp++ = IPOPT_EOL; 1205 return (optlen); 1206 } 1207 1208 /* 1209 * IP socket option processing. 1210 */ 1211 int 1212 ip_ctloutput(int op, struct socket *so, int level, int optname, 1213 struct mbuf **mp) 1214 { 1215 struct inpcb *inp = sotoinpcb(so); 1216 struct mbuf *m = *mp; 1217 int optval = 0; 1218 int error = 0; 1219 #if defined(IPSEC) || defined(FAST_IPSEC) 1220 struct proc *p = curproc; /*XXX*/ 1221 #endif 1222 1223 if (level != IPPROTO_IP) { 1224 error = EINVAL; 1225 if (op == PRCO_SETOPT && *mp) 1226 (void) m_free(*mp); 1227 } else switch (op) { 1228 1229 case PRCO_SETOPT: 1230 switch (optname) { 1231 case IP_OPTIONS: 1232 #ifdef notyet 1233 case IP_RETOPTS: 1234 return (ip_pcbopts(optname, &inp->inp_options, m)); 1235 #else 1236 return (ip_pcbopts(&inp->inp_options, m)); 1237 #endif 1238 1239 case IP_TOS: 1240 case IP_TTL: 1241 case IP_RECVOPTS: 1242 case IP_RECVRETOPTS: 1243 case IP_RECVDSTADDR: 1244 case IP_RECVIF: 1245 if (m == NULL || m->m_len != sizeof(int)) 1246 error = EINVAL; 1247 else { 1248 optval = *mtod(m, int *); 1249 switch (optname) { 1250 1251 case IP_TOS: 1252 inp->inp_ip.ip_tos = optval; 1253 break; 1254 1255 case IP_TTL: 1256 inp->inp_ip.ip_ttl = optval; 1257 break; 1258 #define OPTSET(bit) \ 1259 if (optval) \ 1260 inp->inp_flags |= bit; \ 1261 else \ 1262 inp->inp_flags &= ~bit; 1263 1264 case IP_RECVOPTS: 1265 OPTSET(INP_RECVOPTS); 1266 break; 1267 1268 case IP_RECVRETOPTS: 1269 OPTSET(INP_RECVRETOPTS); 1270 break; 1271 1272 case IP_RECVDSTADDR: 1273 OPTSET(INP_RECVDSTADDR); 1274 break; 1275 1276 case IP_RECVIF: 1277 OPTSET(INP_RECVIF); 1278 break; 1279 } 1280 } 1281 break; 1282 #undef OPTSET 1283 1284 case IP_MULTICAST_IF: 1285 case IP_MULTICAST_TTL: 1286 case IP_MULTICAST_LOOP: 1287 case IP_ADD_MEMBERSHIP: 1288 case IP_DROP_MEMBERSHIP: 1289 error = ip_setmoptions(optname, &inp->inp_moptions, m); 1290 break; 1291 1292 case IP_PORTRANGE: 1293 if (m == 0 || m->m_len != sizeof(int)) 1294 error = EINVAL; 1295 else { 1296 optval = *mtod(m, int *); 1297 1298 switch (optval) { 1299 1300 case IP_PORTRANGE_DEFAULT: 1301 case IP_PORTRANGE_HIGH: 1302 inp->inp_flags &= ~(INP_LOWPORT); 1303 break; 1304 1305 case IP_PORTRANGE_LOW: 1306 inp->inp_flags |= INP_LOWPORT; 1307 break; 1308 1309 default: 1310 error = EINVAL; 1311 break; 1312 } 1313 } 1314 break; 1315 1316 #if defined(IPSEC) || defined(FAST_IPSEC) 1317 case IP_IPSEC_POLICY: 1318 { 1319 caddr_t req = NULL; 1320 size_t len = 0; 1321 int priv = 0; 1322 1323 #ifdef __NetBSD__ 1324 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1325 priv = 0; 1326 else 1327 priv = 1; 1328 #else 1329 priv = (in6p->in6p_socket->so_state & SS_PRIV); 1330 #endif 1331 if (m) { 1332 req = mtod(m, caddr_t); 1333 len = m->m_len; 1334 } 1335 error = ipsec4_set_policy(inp, optname, req, len, priv); 1336 break; 1337 } 1338 #endif /*IPSEC*/ 1339 1340 default: 1341 error = ENOPROTOOPT; 1342 break; 1343 } 1344 if (m) 1345 (void)m_free(m); 1346 break; 1347 1348 case PRCO_GETOPT: 1349 switch (optname) { 1350 case IP_OPTIONS: 1351 case IP_RETOPTS: 1352 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1353 MCLAIM(m, so->so_mowner); 1354 if (inp->inp_options) { 1355 m->m_len = inp->inp_options->m_len; 1356 bcopy(mtod(inp->inp_options, caddr_t), 1357 mtod(m, caddr_t), (unsigned)m->m_len); 1358 } else 1359 m->m_len = 0; 1360 break; 1361 1362 case IP_TOS: 1363 case IP_TTL: 1364 case IP_RECVOPTS: 1365 case IP_RECVRETOPTS: 1366 case IP_RECVDSTADDR: 1367 case IP_RECVIF: 1368 case IP_ERRORMTU: 1369 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1370 MCLAIM(m, so->so_mowner); 1371 m->m_len = sizeof(int); 1372 switch (optname) { 1373 1374 case IP_TOS: 1375 optval = inp->inp_ip.ip_tos; 1376 break; 1377 1378 case IP_TTL: 1379 optval = inp->inp_ip.ip_ttl; 1380 break; 1381 1382 case IP_ERRORMTU: 1383 optval = inp->inp_errormtu; 1384 break; 1385 1386 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1387 1388 case IP_RECVOPTS: 1389 optval = OPTBIT(INP_RECVOPTS); 1390 break; 1391 1392 case IP_RECVRETOPTS: 1393 optval = OPTBIT(INP_RECVRETOPTS); 1394 break; 1395 1396 case IP_RECVDSTADDR: 1397 optval = OPTBIT(INP_RECVDSTADDR); 1398 break; 1399 1400 case IP_RECVIF: 1401 optval = OPTBIT(INP_RECVIF); 1402 break; 1403 } 1404 *mtod(m, int *) = optval; 1405 break; 1406 1407 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */ 1408 /* XXX: code broken */ 1409 case IP_IPSEC_POLICY: 1410 { 1411 caddr_t req = NULL; 1412 size_t len = 0; 1413 1414 if (m) { 1415 req = mtod(m, caddr_t); 1416 len = m->m_len; 1417 } 1418 error = ipsec4_get_policy(inp, req, len, mp); 1419 break; 1420 } 1421 #endif /*IPSEC*/ 1422 1423 case IP_MULTICAST_IF: 1424 case IP_MULTICAST_TTL: 1425 case IP_MULTICAST_LOOP: 1426 case IP_ADD_MEMBERSHIP: 1427 case IP_DROP_MEMBERSHIP: 1428 error = ip_getmoptions(optname, inp->inp_moptions, mp); 1429 if (*mp) 1430 MCLAIM(*mp, so->so_mowner); 1431 break; 1432 1433 case IP_PORTRANGE: 1434 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1435 MCLAIM(m, so->so_mowner); 1436 m->m_len = sizeof(int); 1437 1438 if (inp->inp_flags & INP_LOWPORT) 1439 optval = IP_PORTRANGE_LOW; 1440 else 1441 optval = IP_PORTRANGE_DEFAULT; 1442 1443 *mtod(m, int *) = optval; 1444 break; 1445 1446 default: 1447 error = ENOPROTOOPT; 1448 break; 1449 } 1450 break; 1451 } 1452 return (error); 1453 } 1454 1455 /* 1456 * Set up IP options in pcb for insertion in output packets. 1457 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1458 * with destination address if source routed. 1459 */ 1460 int 1461 #ifdef notyet 1462 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m) 1463 #else 1464 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m) 1465 #endif 1466 { 1467 int cnt, optlen; 1468 u_char *cp; 1469 u_char opt; 1470 1471 /* turn off any old options */ 1472 if (*pcbopt) 1473 (void)m_free(*pcbopt); 1474 *pcbopt = 0; 1475 if (m == (struct mbuf *)0 || m->m_len == 0) { 1476 /* 1477 * Only turning off any previous options. 1478 */ 1479 if (m) 1480 (void)m_free(m); 1481 return (0); 1482 } 1483 1484 #ifndef __vax__ 1485 if (m->m_len % sizeof(int32_t)) 1486 goto bad; 1487 #endif 1488 /* 1489 * IP first-hop destination address will be stored before 1490 * actual options; move other options back 1491 * and clear it when none present. 1492 */ 1493 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1494 goto bad; 1495 cnt = m->m_len; 1496 m->m_len += sizeof(struct in_addr); 1497 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1498 memmove(cp, mtod(m, caddr_t), (unsigned)cnt); 1499 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1500 1501 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1502 opt = cp[IPOPT_OPTVAL]; 1503 if (opt == IPOPT_EOL) 1504 break; 1505 if (opt == IPOPT_NOP) 1506 optlen = 1; 1507 else { 1508 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1509 goto bad; 1510 optlen = cp[IPOPT_OLEN]; 1511 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1512 goto bad; 1513 } 1514 switch (opt) { 1515 1516 default: 1517 break; 1518 1519 case IPOPT_LSRR: 1520 case IPOPT_SSRR: 1521 /* 1522 * user process specifies route as: 1523 * ->A->B->C->D 1524 * D must be our final destination (but we can't 1525 * check that since we may not have connected yet). 1526 * A is first hop destination, which doesn't appear in 1527 * actual IP option, but is stored before the options. 1528 */ 1529 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1530 goto bad; 1531 m->m_len -= sizeof(struct in_addr); 1532 cnt -= sizeof(struct in_addr); 1533 optlen -= sizeof(struct in_addr); 1534 cp[IPOPT_OLEN] = optlen; 1535 /* 1536 * Move first hop before start of options. 1537 */ 1538 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1539 sizeof(struct in_addr)); 1540 /* 1541 * Then copy rest of options back 1542 * to close up the deleted entry. 1543 */ 1544 (void)memmove(&cp[IPOPT_OFFSET+1], 1545 &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr), 1546 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1547 break; 1548 } 1549 } 1550 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1551 goto bad; 1552 *pcbopt = m; 1553 return (0); 1554 1555 bad: 1556 (void)m_free(m); 1557 return (EINVAL); 1558 } 1559 1560 /* 1561 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1562 */ 1563 static struct ifnet * 1564 ip_multicast_if(struct in_addr *a, int *ifindexp) 1565 { 1566 int ifindex; 1567 struct ifnet *ifp = NULL; 1568 struct in_ifaddr *ia; 1569 1570 if (ifindexp) 1571 *ifindexp = 0; 1572 if (ntohl(a->s_addr) >> 24 == 0) { 1573 ifindex = ntohl(a->s_addr) & 0xffffff; 1574 if (ifindex < 0 || if_indexlim <= ifindex) 1575 return NULL; 1576 ifp = ifindex2ifnet[ifindex]; 1577 if (!ifp) 1578 return NULL; 1579 if (ifindexp) 1580 *ifindexp = ifindex; 1581 } else { 1582 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1583 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1584 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1585 ifp = ia->ia_ifp; 1586 break; 1587 } 1588 } 1589 } 1590 return ifp; 1591 } 1592 1593 static int 1594 ip_getoptval(struct mbuf *m, u_int8_t *val, u_int maxval) 1595 { 1596 u_int tval; 1597 1598 if (m == NULL) 1599 return EINVAL; 1600 1601 switch (m->m_len) { 1602 case sizeof(u_char): 1603 tval = *(mtod(m, u_char *)); 1604 break; 1605 case sizeof(u_int): 1606 tval = *(mtod(m, u_int *)); 1607 break; 1608 default: 1609 return EINVAL; 1610 } 1611 1612 if (tval > maxval) 1613 return EINVAL; 1614 1615 *val = tval; 1616 return 0; 1617 } 1618 1619 /* 1620 * Set the IP multicast options in response to user setsockopt(). 1621 */ 1622 int 1623 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m) 1624 { 1625 int error = 0; 1626 int i; 1627 struct in_addr addr; 1628 struct ip_mreq *mreq; 1629 struct ifnet *ifp; 1630 struct ip_moptions *imo = *imop; 1631 struct route ro; 1632 struct sockaddr_in *dst; 1633 int ifindex; 1634 1635 if (imo == NULL) { 1636 /* 1637 * No multicast option buffer attached to the pcb; 1638 * allocate one and initialize to default values. 1639 */ 1640 imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS, 1641 M_WAITOK); 1642 1643 if (imo == NULL) 1644 return (ENOBUFS); 1645 *imop = imo; 1646 imo->imo_multicast_ifp = NULL; 1647 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1648 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1649 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1650 imo->imo_num_memberships = 0; 1651 } 1652 1653 switch (optname) { 1654 1655 case IP_MULTICAST_IF: 1656 /* 1657 * Select the interface for outgoing multicast packets. 1658 */ 1659 if (m == NULL || m->m_len != sizeof(struct in_addr)) { 1660 error = EINVAL; 1661 break; 1662 } 1663 addr = *(mtod(m, struct in_addr *)); 1664 /* 1665 * INADDR_ANY is used to remove a previous selection. 1666 * When no interface is selected, a default one is 1667 * chosen every time a multicast packet is sent. 1668 */ 1669 if (in_nullhost(addr)) { 1670 imo->imo_multicast_ifp = NULL; 1671 break; 1672 } 1673 /* 1674 * The selected interface is identified by its local 1675 * IP address. Find the interface and confirm that 1676 * it supports multicasting. 1677 */ 1678 ifp = ip_multicast_if(&addr, &ifindex); 1679 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1680 error = EADDRNOTAVAIL; 1681 break; 1682 } 1683 imo->imo_multicast_ifp = ifp; 1684 if (ifindex) 1685 imo->imo_multicast_addr = addr; 1686 else 1687 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1688 break; 1689 1690 case IP_MULTICAST_TTL: 1691 /* 1692 * Set the IP time-to-live for outgoing multicast packets. 1693 */ 1694 error = ip_getoptval(m, &imo->imo_multicast_ttl, MAXTTL); 1695 break; 1696 1697 case IP_MULTICAST_LOOP: 1698 /* 1699 * Set the loopback flag for outgoing multicast packets. 1700 * Must be zero or one. 1701 */ 1702 error = ip_getoptval(m, &imo->imo_multicast_loop, 1); 1703 break; 1704 1705 case IP_ADD_MEMBERSHIP: 1706 /* 1707 * Add a multicast group membership. 1708 * Group must be a valid IP multicast address. 1709 */ 1710 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1711 error = EINVAL; 1712 break; 1713 } 1714 mreq = mtod(m, struct ip_mreq *); 1715 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1716 error = EINVAL; 1717 break; 1718 } 1719 /* 1720 * If no interface address was provided, use the interface of 1721 * the route to the given multicast address. 1722 */ 1723 if (in_nullhost(mreq->imr_interface)) { 1724 bzero((caddr_t)&ro, sizeof(ro)); 1725 ro.ro_rt = NULL; 1726 dst = satosin(&ro.ro_dst); 1727 dst->sin_len = sizeof(*dst); 1728 dst->sin_family = AF_INET; 1729 dst->sin_addr = mreq->imr_multiaddr; 1730 rtalloc(&ro); 1731 if (ro.ro_rt == NULL) { 1732 error = EADDRNOTAVAIL; 1733 break; 1734 } 1735 ifp = ro.ro_rt->rt_ifp; 1736 rtfree(ro.ro_rt); 1737 } else { 1738 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1739 } 1740 /* 1741 * See if we found an interface, and confirm that it 1742 * supports multicast. 1743 */ 1744 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1745 error = EADDRNOTAVAIL; 1746 break; 1747 } 1748 /* 1749 * See if the membership already exists or if all the 1750 * membership slots are full. 1751 */ 1752 for (i = 0; i < imo->imo_num_memberships; ++i) { 1753 if (imo->imo_membership[i]->inm_ifp == ifp && 1754 in_hosteq(imo->imo_membership[i]->inm_addr, 1755 mreq->imr_multiaddr)) 1756 break; 1757 } 1758 if (i < imo->imo_num_memberships) { 1759 error = EADDRINUSE; 1760 break; 1761 } 1762 if (i == IP_MAX_MEMBERSHIPS) { 1763 error = ETOOMANYREFS; 1764 break; 1765 } 1766 /* 1767 * Everything looks good; add a new record to the multicast 1768 * address list for the given interface. 1769 */ 1770 if ((imo->imo_membership[i] = 1771 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1772 error = ENOBUFS; 1773 break; 1774 } 1775 ++imo->imo_num_memberships; 1776 break; 1777 1778 case IP_DROP_MEMBERSHIP: 1779 /* 1780 * Drop a multicast group membership. 1781 * Group must be a valid IP multicast address. 1782 */ 1783 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1784 error = EINVAL; 1785 break; 1786 } 1787 mreq = mtod(m, struct ip_mreq *); 1788 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1789 error = EINVAL; 1790 break; 1791 } 1792 /* 1793 * If an interface address was specified, get a pointer 1794 * to its ifnet structure. 1795 */ 1796 if (in_nullhost(mreq->imr_interface)) 1797 ifp = NULL; 1798 else { 1799 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1800 if (ifp == NULL) { 1801 error = EADDRNOTAVAIL; 1802 break; 1803 } 1804 } 1805 /* 1806 * Find the membership in the membership array. 1807 */ 1808 for (i = 0; i < imo->imo_num_memberships; ++i) { 1809 if ((ifp == NULL || 1810 imo->imo_membership[i]->inm_ifp == ifp) && 1811 in_hosteq(imo->imo_membership[i]->inm_addr, 1812 mreq->imr_multiaddr)) 1813 break; 1814 } 1815 if (i == imo->imo_num_memberships) { 1816 error = EADDRNOTAVAIL; 1817 break; 1818 } 1819 /* 1820 * Give up the multicast address record to which the 1821 * membership points. 1822 */ 1823 in_delmulti(imo->imo_membership[i]); 1824 /* 1825 * Remove the gap in the membership array. 1826 */ 1827 for (++i; i < imo->imo_num_memberships; ++i) 1828 imo->imo_membership[i-1] = imo->imo_membership[i]; 1829 --imo->imo_num_memberships; 1830 break; 1831 1832 default: 1833 error = EOPNOTSUPP; 1834 break; 1835 } 1836 1837 /* 1838 * If all options have default values, no need to keep the mbuf. 1839 */ 1840 if (imo->imo_multicast_ifp == NULL && 1841 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1842 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1843 imo->imo_num_memberships == 0) { 1844 free(*imop, M_IPMOPTS); 1845 *imop = NULL; 1846 } 1847 1848 return (error); 1849 } 1850 1851 /* 1852 * Return the IP multicast options in response to user getsockopt(). 1853 */ 1854 int 1855 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp) 1856 { 1857 u_char *ttl; 1858 u_char *loop; 1859 struct in_addr *addr; 1860 struct in_ifaddr *ia; 1861 1862 *mp = m_get(M_WAIT, MT_SOOPTS); 1863 1864 switch (optname) { 1865 1866 case IP_MULTICAST_IF: 1867 addr = mtod(*mp, struct in_addr *); 1868 (*mp)->m_len = sizeof(struct in_addr); 1869 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1870 *addr = zeroin_addr; 1871 else if (imo->imo_multicast_addr.s_addr) { 1872 /* return the value user has set */ 1873 *addr = imo->imo_multicast_addr; 1874 } else { 1875 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1876 *addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1877 } 1878 return (0); 1879 1880 case IP_MULTICAST_TTL: 1881 ttl = mtod(*mp, u_char *); 1882 (*mp)->m_len = 1; 1883 *ttl = imo ? imo->imo_multicast_ttl 1884 : IP_DEFAULT_MULTICAST_TTL; 1885 return (0); 1886 1887 case IP_MULTICAST_LOOP: 1888 loop = mtod(*mp, u_char *); 1889 (*mp)->m_len = 1; 1890 *loop = imo ? imo->imo_multicast_loop 1891 : IP_DEFAULT_MULTICAST_LOOP; 1892 return (0); 1893 1894 default: 1895 return (EOPNOTSUPP); 1896 } 1897 } 1898 1899 /* 1900 * Discard the IP multicast options. 1901 */ 1902 void 1903 ip_freemoptions(struct ip_moptions *imo) 1904 { 1905 int i; 1906 1907 if (imo != NULL) { 1908 for (i = 0; i < imo->imo_num_memberships; ++i) 1909 in_delmulti(imo->imo_membership[i]); 1910 free(imo, M_IPMOPTS); 1911 } 1912 } 1913 1914 /* 1915 * Routine called from ip_output() to loop back a copy of an IP multicast 1916 * packet to the input queue of a specified interface. Note that this 1917 * calls the output routine of the loopback "driver", but with an interface 1918 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1919 */ 1920 static void 1921 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst) 1922 { 1923 struct ip *ip; 1924 struct mbuf *copym; 1925 1926 copym = m_copy(m, 0, M_COPYALL); 1927 if (copym != NULL 1928 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1929 copym = m_pullup(copym, sizeof(struct ip)); 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 1937 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1938 in_delayed_cksum(copym); 1939 copym->m_pkthdr.csum_flags &= 1940 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1941 } 1942 1943 ip->ip_sum = 0; 1944 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1945 (void) looutput(ifp, copym, sintosa(dst), NULL); 1946 } 1947 } 1948
Indexes created Wed Oct 01 12:09:54 GMT 2025