1 /* $NetBSD: ip_output.c,v 1.196 2008/04/28 20:24:09 martin 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 * 49 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 50 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 51 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 52 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 53 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 54 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 55 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 56 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 57 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 58 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 59 * POSSIBILITY OF SUCH DAMAGE. 60 */ 61 62 /* 63 * Copyright (c) 1982, 1986, 1988, 1990, 1993 64 * The Regents of the University of California. All rights reserved. 65 * 66 * Redistribution and use in source and binary forms, with or without 67 * modification, are permitted provided that the following conditions 68 * are met: 69 * 1. Redistributions of source code must retain the above copyright 70 * notice, this list of conditions and the following disclaimer. 71 * 2. Redistributions in binary form must reproduce the above copyright 72 * notice, this list of conditions and the following disclaimer in the 73 * documentation and/or other materials provided with the distribution. 74 * 3. Neither the name of the University nor the names of its contributors 75 * may be used to endorse or promote products derived from this software 76 * without specific prior written permission. 77 * 78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 88 * SUCH DAMAGE. 89 * 90 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 91 */ 92 93 #include <sys/cdefs.h> 94 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.196 2008/04/28 20:24:09 martin Exp $"); 95 96 #include "opt_pfil_hooks.h" 97 #include "opt_inet.h" 98 #include "opt_ipsec.h" 99 #include "opt_mrouting.h" 100 101 #include <sys/param.h> 102 #include <sys/malloc.h> 103 #include <sys/mbuf.h> 104 #include <sys/errno.h> 105 #include <sys/protosw.h> 106 #include <sys/socket.h> 107 #include <sys/socketvar.h> 108 #include <sys/kauth.h> 109 #ifdef FAST_IPSEC 110 #include <sys/domain.h> 111 #endif 112 #include <sys/systm.h> 113 #include <sys/proc.h> 114 115 #include <net/if.h> 116 #include <net/route.h> 117 #include <net/pfil.h> 118 119 #include <netinet/in.h> 120 #include <netinet/in_systm.h> 121 #include <netinet/ip.h> 122 #include <netinet/in_pcb.h> 123 #include <netinet/in_var.h> 124 #include <netinet/ip_var.h> 125 #include <netinet/ip_private.h> 126 #include <netinet/in_offload.h> 127 128 #ifdef MROUTING 129 #include <netinet/ip_mroute.h> 130 #endif 131 132 #include <machine/stdarg.h> 133 134 #ifdef IPSEC 135 #include <netinet6/ipsec.h> 136 #include <netinet6/ipsec_private.h> 137 #include <netkey/key.h> 138 #include <netkey/key_debug.h> 139 #endif /*IPSEC*/ 140 141 #ifdef FAST_IPSEC 142 #include <netipsec/ipsec.h> 143 #include <netipsec/key.h> 144 #include <netipsec/xform.h> 145 #endif /* FAST_IPSEC*/ 146 147 #ifdef IPSEC_NAT_T 148 #include <netinet/udp.h> 149 #endif 150 151 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 152 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 153 static void ip_mloopback(struct ifnet *, struct mbuf *, 154 const struct sockaddr_in *); 155 static int ip_getoptval(struct mbuf *, u_int8_t *, u_int); 156 157 #ifdef PFIL_HOOKS 158 extern struct pfil_head inet_pfil_hook; /* XXX */ 159 #endif 160 161 int ip_do_loopback_cksum = 0; 162 163 /* 164 * IP output. The packet in mbuf chain m contains a skeletal IP 165 * header (with len, off, ttl, proto, tos, src, dst). 166 * The mbuf chain containing the packet will be freed. 167 * The mbuf opt, if present, will not be freed. 168 */ 169 int 170 ip_output(struct mbuf *m0, ...) 171 { 172 struct rtentry *rt; 173 struct ip *ip; 174 struct ifnet *ifp; 175 struct mbuf *m = m0; 176 int hlen = sizeof (struct ip); 177 int len, error = 0; 178 struct route iproute; 179 const struct sockaddr_in *dst; 180 struct in_ifaddr *ia; 181 struct ifaddr *xifa; 182 struct mbuf *opt; 183 struct route *ro; 184 int flags, sw_csum; 185 int *mtu_p; 186 u_long mtu; 187 struct ip_moptions *imo; 188 struct socket *so; 189 va_list ap; 190 #ifdef IPSEC_NAT_T 191 int natt_frag = 0; 192 #endif 193 #ifdef IPSEC 194 struct secpolicy *sp = NULL; 195 #endif /*IPSEC*/ 196 #ifdef FAST_IPSEC 197 struct inpcb *inp; 198 struct secpolicy *sp = NULL; 199 int s; 200 #endif 201 u_int16_t ip_len; 202 union { 203 struct sockaddr dst; 204 struct sockaddr_in dst4; 205 } u; 206 struct sockaddr *rdst = &u.dst; /* real IP destination, as opposed 207 * to the nexthop 208 */ 209 210 len = 0; 211 va_start(ap, m0); 212 opt = va_arg(ap, struct mbuf *); 213 ro = va_arg(ap, struct route *); 214 flags = va_arg(ap, int); 215 imo = va_arg(ap, struct ip_moptions *); 216 so = va_arg(ap, struct socket *); 217 if (flags & IP_RETURNMTU) 218 mtu_p = va_arg(ap, int *); 219 else 220 mtu_p = NULL; 221 va_end(ap); 222 223 MCLAIM(m, &ip_tx_mowner); 224 #ifdef FAST_IPSEC 225 if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET) 226 inp = (struct inpcb *)so->so_pcb; 227 else 228 inp = NULL; 229 #endif /* FAST_IPSEC */ 230 231 #ifdef DIAGNOSTIC 232 if ((m->m_flags & M_PKTHDR) == 0) 233 panic("ip_output: no HDR"); 234 235 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) != 0) { 236 panic("ip_output: IPv6 checksum offload flags: %d", 237 m->m_pkthdr.csum_flags); 238 } 239 240 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) == 241 (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 242 panic("ip_output: conflicting checksum offload flags: %d", 243 m->m_pkthdr.csum_flags); 244 } 245 #endif 246 if (opt) { 247 m = ip_insertoptions(m, opt, &len); 248 if (len >= sizeof(struct ip)) 249 hlen = len; 250 } 251 ip = mtod(m, struct ip *); 252 /* 253 * Fill in IP header. 254 */ 255 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 256 ip->ip_v = IPVERSION; 257 ip->ip_off = htons(0); 258 /* ip->ip_id filled in after we find out source ia */ 259 ip->ip_hl = hlen >> 2; 260 IP_STATINC(IP_STAT_LOCALOUT); 261 } else { 262 hlen = ip->ip_hl << 2; 263 } 264 /* 265 * Route packet. 266 */ 267 memset(&iproute, 0, sizeof(iproute)); 268 if (ro == NULL) 269 ro = &iproute; 270 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 271 dst = satocsin(rtcache_getdst(ro)); 272 /* 273 * If there is a cached route, 274 * check that it is to the same destination 275 * and is still up. If not, free it and try again. 276 * The address family should also be checked in case of sharing the 277 * cache with IPv6. 278 */ 279 if (dst == NULL) 280 ; 281 else if (dst->sin_family != AF_INET || 282 !in_hosteq(dst->sin_addr, ip->ip_dst)) 283 rtcache_free(ro); 284 285 if ((rt = rtcache_validate(ro)) == NULL && 286 (rt = rtcache_update(ro, 1)) == NULL) { 287 dst = &u.dst4; 288 rtcache_setdst(ro, &u.dst); 289 } 290 /* 291 * If routing to interface only, 292 * short circuit routing lookup. 293 */ 294 if (flags & IP_ROUTETOIF) { 295 if ((ia = ifatoia(ifa_ifwithladdr(sintocsa(dst)))) == NULL) { 296 IP_STATINC(IP_STAT_NOROUTE); 297 error = ENETUNREACH; 298 goto bad; 299 } 300 ifp = ia->ia_ifp; 301 mtu = ifp->if_mtu; 302 ip->ip_ttl = 1; 303 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 304 ip->ip_dst.s_addr == INADDR_BROADCAST) && 305 imo != NULL && imo->imo_multicast_ifp != NULL) { 306 ifp = imo->imo_multicast_ifp; 307 mtu = ifp->if_mtu; 308 IFP_TO_IA(ifp, ia); 309 } else { 310 if (rt == NULL) 311 rt = rtcache_init(ro); 312 if (rt == NULL) { 313 IP_STATINC(IP_STAT_NOROUTE); 314 error = EHOSTUNREACH; 315 goto bad; 316 } 317 ia = ifatoia(rt->rt_ifa); 318 ifp = rt->rt_ifp; 319 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 320 mtu = ifp->if_mtu; 321 rt->rt_use++; 322 if (rt->rt_flags & RTF_GATEWAY) 323 dst = satosin(rt->rt_gateway); 324 } 325 if (IN_MULTICAST(ip->ip_dst.s_addr) || 326 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 327 struct in_multi *inm; 328 329 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 330 M_BCAST : M_MCAST; 331 /* 332 * See if the caller provided any multicast options 333 */ 334 if (imo != NULL) 335 ip->ip_ttl = imo->imo_multicast_ttl; 336 else 337 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 338 339 /* 340 * if we don't know the outgoing ifp yet, we can't generate 341 * output 342 */ 343 if (!ifp) { 344 IP_STATINC(IP_STAT_NOROUTE); 345 error = ENETUNREACH; 346 goto bad; 347 } 348 349 /* 350 * If the packet is multicast or broadcast, confirm that 351 * the outgoing interface can transmit it. 352 */ 353 if (((m->m_flags & M_MCAST) && 354 (ifp->if_flags & IFF_MULTICAST) == 0) || 355 ((m->m_flags & M_BCAST) && 356 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 357 IP_STATINC(IP_STAT_NOROUTE); 358 error = ENETUNREACH; 359 goto bad; 360 } 361 /* 362 * If source address not specified yet, use an address 363 * of outgoing interface. 364 */ 365 if (in_nullhost(ip->ip_src)) { 366 struct in_ifaddr *xia; 367 368 IFP_TO_IA(ifp, xia); 369 if (!xia) { 370 error = EADDRNOTAVAIL; 371 goto bad; 372 } 373 xifa = &xia->ia_ifa; 374 if (xifa->ifa_getifa != NULL) { 375 xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 376 } 377 ip->ip_src = xia->ia_addr.sin_addr; 378 } 379 380 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 381 if (inm != NULL && 382 (imo == NULL || imo->imo_multicast_loop)) { 383 /* 384 * If we belong to the destination multicast group 385 * on the outgoing interface, and the caller did not 386 * forbid loopback, loop back a copy. 387 */ 388 ip_mloopback(ifp, m, &u.dst4); 389 } 390 #ifdef MROUTING 391 else { 392 /* 393 * If we are acting as a multicast router, perform 394 * multicast forwarding as if the packet had just 395 * arrived on the interface to which we are about 396 * to send. The multicast forwarding function 397 * recursively calls this function, using the 398 * IP_FORWARDING flag to prevent infinite recursion. 399 * 400 * Multicasts that are looped back by ip_mloopback(), 401 * above, will be forwarded by the ip_input() routine, 402 * if necessary. 403 */ 404 extern struct socket *ip_mrouter; 405 406 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 407 if (ip_mforward(m, ifp) != 0) { 408 m_freem(m); 409 goto done; 410 } 411 } 412 } 413 #endif 414 /* 415 * Multicasts with a time-to-live of zero may be looped- 416 * back, above, but must not be transmitted on a network. 417 * Also, multicasts addressed to the loopback interface 418 * are not sent -- the above call to ip_mloopback() will 419 * loop back a copy if this host actually belongs to the 420 * destination group on the loopback interface. 421 */ 422 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 423 m_freem(m); 424 goto done; 425 } 426 427 goto sendit; 428 } 429 /* 430 * If source address not specified yet, use address 431 * of outgoing interface. 432 */ 433 if (in_nullhost(ip->ip_src)) { 434 xifa = &ia->ia_ifa; 435 if (xifa->ifa_getifa != NULL) 436 ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 437 ip->ip_src = ia->ia_addr.sin_addr; 438 } 439 440 /* 441 * packets with Class-D address as source are not valid per 442 * RFC 1112 443 */ 444 if (IN_MULTICAST(ip->ip_src.s_addr)) { 445 IP_STATINC(IP_STAT_ODROPPED); 446 error = EADDRNOTAVAIL; 447 goto bad; 448 } 449 450 /* 451 * Look for broadcast address and 452 * and verify user is allowed to send 453 * such a packet. 454 */ 455 if (in_broadcast(dst->sin_addr, ifp)) { 456 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 457 error = EADDRNOTAVAIL; 458 goto bad; 459 } 460 if ((flags & IP_ALLOWBROADCAST) == 0) { 461 error = EACCES; 462 goto bad; 463 } 464 /* don't allow broadcast messages to be fragmented */ 465 if (ntohs(ip->ip_len) > ifp->if_mtu) { 466 error = EMSGSIZE; 467 goto bad; 468 } 469 m->m_flags |= M_BCAST; 470 } else 471 m->m_flags &= ~M_BCAST; 472 473 sendit: 474 if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) { 475 if (m->m_pkthdr.len < IP_MINFRAGSIZE) { 476 ip->ip_id = 0; 477 } else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 478 ip->ip_id = ip_newid(ia); 479 } else { 480 481 /* 482 * TSO capable interfaces (typically?) increment 483 * ip_id for each segment. 484 * "allocate" enough ids here to increase the chance 485 * for them to be unique. 486 * 487 * note that the following calculation is not 488 * needed to be precise. wasting some ip_id is fine. 489 */ 490 491 unsigned int segsz = m->m_pkthdr.segsz; 492 unsigned int datasz = ntohs(ip->ip_len) - hlen; 493 unsigned int num = howmany(datasz, segsz); 494 495 ip->ip_id = ip_newid_range(ia, num); 496 } 497 } 498 /* 499 * If we're doing Path MTU Discovery, we need to set DF unless 500 * the route's MTU is locked. 501 */ 502 if ((flags & IP_MTUDISC) != 0 && rt != NULL && 503 (rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 504 ip->ip_off |= htons(IP_DF); 505 506 /* Remember the current ip_len */ 507 ip_len = ntohs(ip->ip_len); 508 509 #ifdef IPSEC 510 /* get SP for this packet */ 511 if (so == NULL) 512 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 513 flags, &error); 514 else { 515 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 516 IPSEC_DIR_OUTBOUND)) 517 goto skip_ipsec; 518 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 519 } 520 521 if (sp == NULL) { 522 IPSEC_STATINC(IPSEC_STAT_IN_INVAL); 523 goto bad; 524 } 525 526 error = 0; 527 528 /* check policy */ 529 switch (sp->policy) { 530 case IPSEC_POLICY_DISCARD: 531 /* 532 * This packet is just discarded. 533 */ 534 IPSEC_STATINC(IPSEC_STAT_OUT_POLVIO); 535 goto bad; 536 537 case IPSEC_POLICY_BYPASS: 538 case IPSEC_POLICY_NONE: 539 /* no need to do IPsec. */ 540 goto skip_ipsec; 541 542 case IPSEC_POLICY_IPSEC: 543 if (sp->req == NULL) { 544 /* XXX should be panic ? */ 545 printf("ip_output: No IPsec request specified.\n"); 546 error = EINVAL; 547 goto bad; 548 } 549 break; 550 551 case IPSEC_POLICY_ENTRUST: 552 default: 553 printf("ip_output: Invalid policy found. %d\n", sp->policy); 554 } 555 556 #ifdef IPSEC_NAT_T 557 /* 558 * NAT-T ESP fragmentation: don't do IPSec processing now, 559 * we'll do it on each fragmented packet. 560 */ 561 if (sp->req->sav && 562 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 563 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 564 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 565 natt_frag = 1; 566 mtu = sp->req->sav->esp_frag; 567 goto skip_ipsec; 568 } 569 } 570 #endif /* IPSEC_NAT_T */ 571 572 /* 573 * ipsec4_output() expects ip_len and ip_off in network 574 * order. They have been set to network order above. 575 */ 576 577 { 578 struct ipsec_output_state state; 579 bzero(&state, sizeof(state)); 580 state.m = m; 581 if (flags & IP_ROUTETOIF) { 582 state.ro = &iproute; 583 memset(&iproute, 0, sizeof(iproute)); 584 } else 585 state.ro = ro; 586 state.dst = sintocsa(dst); 587 588 /* 589 * We can't defer the checksum of payload data if 590 * we're about to encrypt/authenticate it. 591 * 592 * XXX When we support crypto offloading functions of 593 * XXX network interfaces, we need to reconsider this, 594 * XXX since it's likely that they'll support checksumming, 595 * XXX as well. 596 */ 597 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 598 in_delayed_cksum(m); 599 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 600 } 601 602 error = ipsec4_output(&state, sp, flags); 603 604 m = state.m; 605 if (flags & IP_ROUTETOIF) { 606 /* 607 * if we have tunnel mode SA, we may need to ignore 608 * IP_ROUTETOIF. 609 */ 610 if (state.ro != &iproute || 611 rtcache_validate(state.ro) != NULL) { 612 flags &= ~IP_ROUTETOIF; 613 ro = state.ro; 614 } 615 } else 616 ro = state.ro; 617 dst = satocsin(state.dst); 618 if (error) { 619 /* mbuf is already reclaimed in ipsec4_output. */ 620 m0 = NULL; 621 switch (error) { 622 case EHOSTUNREACH: 623 case ENETUNREACH: 624 case EMSGSIZE: 625 case ENOBUFS: 626 case ENOMEM: 627 break; 628 default: 629 printf("ip4_output (ipsec): error code %d\n", error); 630 /*fall through*/ 631 case ENOENT: 632 /* don't show these error codes to the user */ 633 error = 0; 634 break; 635 } 636 goto bad; 637 } 638 639 /* be sure to update variables that are affected by ipsec4_output() */ 640 ip = mtod(m, struct ip *); 641 hlen = ip->ip_hl << 2; 642 ip_len = ntohs(ip->ip_len); 643 644 if ((rt = rtcache_validate(ro)) == NULL) { 645 if ((flags & IP_ROUTETOIF) == 0) { 646 printf("ip_output: " 647 "can't update route after IPsec processing\n"); 648 error = EHOSTUNREACH; /*XXX*/ 649 goto bad; 650 } 651 } else { 652 /* nobody uses ia beyond here */ 653 if (state.encap) { 654 ifp = rt->rt_ifp; 655 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 656 mtu = ifp->if_mtu; 657 } 658 } 659 } 660 skip_ipsec: 661 #endif /*IPSEC*/ 662 #ifdef FAST_IPSEC 663 /* 664 * Check the security policy (SP) for the packet and, if 665 * required, do IPsec-related processing. There are two 666 * cases here; the first time a packet is sent through 667 * it will be untagged and handled by ipsec4_checkpolicy. 668 * If the packet is resubmitted to ip_output (e.g. after 669 * AH, ESP, etc. processing), there will be a tag to bypass 670 * the lookup and related policy checking. 671 */ 672 if (!ipsec_outdone(m)) { 673 s = splsoftnet(); 674 if (inp != NULL && 675 IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND)) 676 goto spd_done; 677 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 678 &error, inp); 679 /* 680 * There are four return cases: 681 * sp != NULL apply IPsec policy 682 * sp == NULL, error == 0 no IPsec handling needed 683 * sp == NULL, error == -EINVAL discard packet w/o error 684 * sp == NULL, error != 0 discard packet, report error 685 */ 686 if (sp != NULL) { 687 #ifdef IPSEC_NAT_T 688 /* 689 * NAT-T ESP fragmentation: don't do IPSec processing now, 690 * we'll do it on each fragmented packet. 691 */ 692 if (sp->req->sav && 693 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 694 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 695 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 696 natt_frag = 1; 697 mtu = sp->req->sav->esp_frag; 698 splx(s); 699 goto spd_done; 700 } 701 } 702 #endif /* IPSEC_NAT_T */ 703 704 /* 705 * Do delayed checksums now because we send before 706 * this is done in the normal processing path. 707 */ 708 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 709 in_delayed_cksum(m); 710 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 711 } 712 713 #ifdef __FreeBSD__ 714 ip->ip_len = htons(ip->ip_len); 715 ip->ip_off = htons(ip->ip_off); 716 #endif 717 718 /* NB: callee frees mbuf */ 719 error = ipsec4_process_packet(m, sp->req, flags, 0); 720 /* 721 * Preserve KAME behaviour: ENOENT can be returned 722 * when an SA acquire is in progress. Don't propagate 723 * this to user-level; it confuses applications. 724 * 725 * XXX this will go away when the SADB is redone. 726 */ 727 if (error == ENOENT) 728 error = 0; 729 splx(s); 730 goto done; 731 } else { 732 splx(s); 733 734 if (error != 0) { 735 /* 736 * Hack: -EINVAL is used to signal that a packet 737 * should be silently discarded. This is typically 738 * because we asked key management for an SA and 739 * it was delayed (e.g. kicked up to IKE). 740 */ 741 if (error == -EINVAL) 742 error = 0; 743 goto bad; 744 } else { 745 /* No IPsec processing for this packet. */ 746 } 747 } 748 } 749 spd_done: 750 #endif /* FAST_IPSEC */ 751 752 #ifdef PFIL_HOOKS 753 /* 754 * Run through list of hooks for output packets. 755 */ 756 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 757 goto done; 758 if (m == NULL) 759 goto done; 760 761 ip = mtod(m, struct ip *); 762 hlen = ip->ip_hl << 2; 763 ip_len = ntohs(ip->ip_len); 764 #endif /* PFIL_HOOKS */ 765 766 m->m_pkthdr.csum_data |= hlen << 16; 767 768 #if IFA_STATS 769 /* 770 * search for the source address structure to 771 * maintain output statistics. 772 */ 773 INADDR_TO_IA(ip->ip_src, ia); 774 #endif 775 776 /* Maybe skip checksums on loopback interfaces. */ 777 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 778 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 779 } 780 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 781 /* 782 * If small enough for mtu of path, or if using TCP segmentation 783 * offload, can just send directly. 784 */ 785 if (ip_len <= mtu || 786 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 787 #if IFA_STATS 788 if (ia) 789 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 790 #endif 791 /* 792 * Always initialize the sum to 0! Some HW assisted 793 * checksumming requires this. 794 */ 795 ip->ip_sum = 0; 796 797 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 798 /* 799 * Perform any checksums that the hardware can't do 800 * for us. 801 * 802 * XXX Does any hardware require the {th,uh}_sum 803 * XXX fields to be 0? 804 */ 805 if (sw_csum & M_CSUM_IPv4) { 806 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 807 ip->ip_sum = in_cksum(m, hlen); 808 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 809 } 810 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 811 if (IN_NEED_CHECKSUM(ifp, 812 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 813 in_delayed_cksum(m); 814 } 815 m->m_pkthdr.csum_flags &= 816 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 817 } 818 } 819 820 #ifdef IPSEC 821 /* clean ipsec history once it goes out of the node */ 822 ipsec_delaux(m); 823 #endif 824 825 if (__predict_true( 826 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 || 827 (ifp->if_capenable & IFCAP_TSOv4) != 0)) { 828 error = 829 (*ifp->if_output)(ifp, m, 830 (m->m_flags & M_MCAST) ? 831 sintocsa(rdst) : sintocsa(dst), 832 rt); 833 } else { 834 error = 835 ip_tso_output(ifp, m, 836 (m->m_flags & M_MCAST) ? 837 sintocsa(rdst) : sintocsa(dst), 838 rt); 839 } 840 goto done; 841 } 842 843 /* 844 * We can't use HW checksumming if we're about to 845 * to fragment the packet. 846 * 847 * XXX Some hardware can do this. 848 */ 849 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 850 if (IN_NEED_CHECKSUM(ifp, 851 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 852 in_delayed_cksum(m); 853 } 854 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 855 } 856 857 /* 858 * Too large for interface; fragment if possible. 859 * Must be able to put at least 8 bytes per fragment. 860 */ 861 if (ntohs(ip->ip_off) & IP_DF) { 862 if (flags & IP_RETURNMTU) 863 *mtu_p = mtu; 864 error = EMSGSIZE; 865 IP_STATINC(IP_STAT_CANTFRAG); 866 goto bad; 867 } 868 869 error = ip_fragment(m, ifp, mtu); 870 if (error) { 871 m = NULL; 872 goto bad; 873 } 874 875 for (; m; m = m0) { 876 m0 = m->m_nextpkt; 877 m->m_nextpkt = 0; 878 if (error == 0) { 879 #if IFA_STATS 880 if (ia) 881 ia->ia_ifa.ifa_data.ifad_outbytes += 882 ntohs(ip->ip_len); 883 #endif 884 #ifdef IPSEC 885 /* clean ipsec history once it goes out of the node */ 886 ipsec_delaux(m); 887 #endif /* IPSEC */ 888 889 #ifdef IPSEC_NAT_T 890 /* 891 * If we get there, the packet has not been handeld by 892 * IPSec whereas it should have. Now that it has been 893 * fragmented, re-inject it in ip_output so that IPsec 894 * processing can occur. 895 */ 896 if (natt_frag) { 897 error = ip_output(m, opt, 898 ro, flags, imo, so, mtu_p); 899 } else 900 #endif /* IPSEC_NAT_T */ 901 { 902 KASSERT((m->m_pkthdr.csum_flags & 903 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 904 error = (*ifp->if_output)(ifp, m, 905 (m->m_flags & M_MCAST) ? 906 sintocsa(rdst) : sintocsa(dst), 907 rt); 908 } 909 } else 910 m_freem(m); 911 } 912 913 if (error == 0) 914 IP_STATINC(IP_STAT_FRAGMENTED); 915 done: 916 rtcache_free(&iproute); 917 918 #ifdef IPSEC 919 if (sp != NULL) { 920 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 921 printf("DP ip_output call free SP:%p\n", sp)); 922 key_freesp(sp); 923 } 924 #endif /* IPSEC */ 925 #ifdef FAST_IPSEC 926 if (sp != NULL) 927 KEY_FREESP(&sp); 928 #endif /* FAST_IPSEC */ 929 930 return (error); 931 bad: 932 m_freem(m); 933 goto done; 934 } 935 936 int 937 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 938 { 939 struct ip *ip, *mhip; 940 struct mbuf *m0; 941 int len, hlen, off; 942 int mhlen, firstlen; 943 struct mbuf **mnext; 944 int sw_csum = m->m_pkthdr.csum_flags; 945 int fragments = 0; 946 int s; 947 int error = 0; 948 949 ip = mtod(m, struct ip *); 950 hlen = ip->ip_hl << 2; 951 if (ifp != NULL) 952 sw_csum &= ~ifp->if_csum_flags_tx; 953 954 len = (mtu - hlen) &~ 7; 955 if (len < 8) { 956 m_freem(m); 957 return (EMSGSIZE); 958 } 959 960 firstlen = len; 961 mnext = &m->m_nextpkt; 962 963 /* 964 * Loop through length of segment after first fragment, 965 * make new header and copy data of each part and link onto chain. 966 */ 967 m0 = m; 968 mhlen = sizeof (struct ip); 969 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 970 MGETHDR(m, M_DONTWAIT, MT_HEADER); 971 if (m == 0) { 972 error = ENOBUFS; 973 IP_STATINC(IP_STAT_ODROPPED); 974 goto sendorfree; 975 } 976 MCLAIM(m, m0->m_owner); 977 *mnext = m; 978 mnext = &m->m_nextpkt; 979 m->m_data += max_linkhdr; 980 mhip = mtod(m, struct ip *); 981 *mhip = *ip; 982 /* we must inherit MCAST and BCAST flags */ 983 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 984 if (hlen > sizeof (struct ip)) { 985 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 986 mhip->ip_hl = mhlen >> 2; 987 } 988 m->m_len = mhlen; 989 mhip->ip_off = ((off - hlen) >> 3) + 990 (ntohs(ip->ip_off) & ~IP_MF); 991 if (ip->ip_off & htons(IP_MF)) 992 mhip->ip_off |= IP_MF; 993 if (off + len >= ntohs(ip->ip_len)) 994 len = ntohs(ip->ip_len) - off; 995 else 996 mhip->ip_off |= IP_MF; 997 HTONS(mhip->ip_off); 998 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 999 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 1000 if (m->m_next == 0) { 1001 error = ENOBUFS; /* ??? */ 1002 IP_STATINC(IP_STAT_ODROPPED); 1003 goto sendorfree; 1004 } 1005 m->m_pkthdr.len = mhlen + len; 1006 m->m_pkthdr.rcvif = (struct ifnet *)0; 1007 mhip->ip_sum = 0; 1008 if (sw_csum & M_CSUM_IPv4) { 1009 mhip->ip_sum = in_cksum(m, mhlen); 1010 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 1011 } else { 1012 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 1013 m->m_pkthdr.csum_data |= mhlen << 16; 1014 } 1015 IP_STATINC(IP_STAT_OFRAGMENTS); 1016 fragments++; 1017 } 1018 /* 1019 * Update first fragment by trimming what's been copied out 1020 * and updating header, then send each fragment (in order). 1021 */ 1022 m = m0; 1023 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 1024 m->m_pkthdr.len = hlen + firstlen; 1025 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 1026 ip->ip_off |= htons(IP_MF); 1027 ip->ip_sum = 0; 1028 if (sw_csum & M_CSUM_IPv4) { 1029 ip->ip_sum = in_cksum(m, hlen); 1030 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 1031 } else { 1032 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 1033 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 1034 sizeof(struct ip)); 1035 } 1036 sendorfree: 1037 /* 1038 * If there is no room for all the fragments, don't queue 1039 * any of them. 1040 */ 1041 if (ifp != NULL) { 1042 s = splnet(); 1043 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 1044 error == 0) { 1045 error = ENOBUFS; 1046 IP_STATINC(IP_STAT_ODROPPED); 1047 IFQ_INC_DROPS(&ifp->if_snd); 1048 } 1049 splx(s); 1050 } 1051 if (error) { 1052 for (m = m0; m; m = m0) { 1053 m0 = m->m_nextpkt; 1054 m->m_nextpkt = NULL; 1055 m_freem(m); 1056 } 1057 } 1058 return (error); 1059 } 1060 1061 /* 1062 * Process a delayed payload checksum calculation. 1063 */ 1064 void 1065 in_delayed_cksum(struct mbuf *m) 1066 { 1067 struct ip *ip; 1068 u_int16_t csum, offset; 1069 1070 ip = mtod(m, struct ip *); 1071 offset = ip->ip_hl << 2; 1072 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 1073 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 1074 csum = 0xffff; 1075 1076 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 1077 1078 if ((offset + sizeof(u_int16_t)) > m->m_len) { 1079 /* This happen when ip options were inserted 1080 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 1081 m->m_len, offset, ip->ip_p); 1082 */ 1083 m_copyback(m, offset, sizeof(csum), (void *) &csum); 1084 } else 1085 *(u_int16_t *)(mtod(m, char *) + offset) = csum; 1086 } 1087 1088 /* 1089 * Determine the maximum length of the options to be inserted; 1090 * we would far rather allocate too much space rather than too little. 1091 */ 1092 1093 u_int 1094 ip_optlen(struct inpcb *inp) 1095 { 1096 struct mbuf *m = inp->inp_options; 1097 1098 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 1099 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 1100 else 1101 return 0; 1102 } 1103 1104 1105 /* 1106 * Insert IP options into preformed packet. 1107 * Adjust IP destination as required for IP source routing, 1108 * as indicated by a non-zero in_addr at the start of the options. 1109 */ 1110 static struct mbuf * 1111 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 1112 { 1113 struct ipoption *p = mtod(opt, struct ipoption *); 1114 struct mbuf *n; 1115 struct ip *ip = mtod(m, struct ip *); 1116 unsigned optlen; 1117 1118 optlen = opt->m_len - sizeof(p->ipopt_dst); 1119 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 1120 return (m); /* XXX should fail */ 1121 if (!in_nullhost(p->ipopt_dst)) 1122 ip->ip_dst = p->ipopt_dst; 1123 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 1124 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1125 if (n == 0) 1126 return (m); 1127 MCLAIM(n, m->m_owner); 1128 M_MOVE_PKTHDR(n, m); 1129 m->m_len -= sizeof(struct ip); 1130 m->m_data += sizeof(struct ip); 1131 n->m_next = m; 1132 m = n; 1133 m->m_len = optlen + sizeof(struct ip); 1134 m->m_data += max_linkhdr; 1135 bcopy((void *)ip, mtod(m, void *), sizeof(struct ip)); 1136 } else { 1137 m->m_data -= optlen; 1138 m->m_len += optlen; 1139 memmove(mtod(m, void *), ip, sizeof(struct ip)); 1140 } 1141 m->m_pkthdr.len += optlen; 1142 ip = mtod(m, struct ip *); 1143 bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen); 1144 *phlen = sizeof(struct ip) + optlen; 1145 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1146 return (m); 1147 } 1148 1149 /* 1150 * Copy options from ip to jp, 1151 * omitting those not copied during fragmentation. 1152 */ 1153 int 1154 ip_optcopy(struct ip *ip, struct ip *jp) 1155 { 1156 u_char *cp, *dp; 1157 int opt, optlen, cnt; 1158 1159 cp = (u_char *)(ip + 1); 1160 dp = (u_char *)(jp + 1); 1161 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1162 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1163 opt = cp[0]; 1164 if (opt == IPOPT_EOL) 1165 break; 1166 if (opt == IPOPT_NOP) { 1167 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1168 *dp++ = IPOPT_NOP; 1169 optlen = 1; 1170 continue; 1171 } 1172 #ifdef DIAGNOSTIC 1173 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1174 panic("malformed IPv4 option passed to ip_optcopy"); 1175 #endif 1176 optlen = cp[IPOPT_OLEN]; 1177 #ifdef DIAGNOSTIC 1178 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1179 panic("malformed IPv4 option passed to ip_optcopy"); 1180 #endif 1181 /* bogus lengths should have been caught by ip_dooptions */ 1182 if (optlen > cnt) 1183 optlen = cnt; 1184 if (IPOPT_COPIED(opt)) { 1185 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 1186 dp += optlen; 1187 } 1188 } 1189 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1190 *dp++ = IPOPT_EOL; 1191 return (optlen); 1192 } 1193 1194 /* 1195 * IP socket option processing. 1196 */ 1197 int 1198 ip_ctloutput(int op, struct socket *so, int level, int optname, 1199 struct mbuf **mp) 1200 { 1201 struct inpcb *inp = sotoinpcb(so); 1202 struct mbuf *m = *mp; 1203 int optval = 0; 1204 int error = 0; 1205 #if defined(IPSEC) || defined(FAST_IPSEC) 1206 struct lwp *l = curlwp; /*XXX*/ 1207 #endif 1208 1209 if (level != IPPROTO_IP) { 1210 if (op == PRCO_SETOPT && *mp) 1211 (void) m_free(*mp); 1212 if (level == SOL_SOCKET && optname == SO_NOHEADER) 1213 return 0; 1214 return ENOPROTOOPT; 1215 } 1216 1217 switch (op) { 1218 1219 case PRCO_SETOPT: 1220 switch (optname) { 1221 case IP_OPTIONS: 1222 #ifdef notyet 1223 case IP_RETOPTS: 1224 return (ip_pcbopts(optname, &inp->inp_options, m)); 1225 #else 1226 return (ip_pcbopts(&inp->inp_options, m)); 1227 #endif 1228 1229 case IP_TOS: 1230 case IP_TTL: 1231 case IP_RECVOPTS: 1232 case IP_RECVRETOPTS: 1233 case IP_RECVDSTADDR: 1234 case IP_RECVIF: 1235 if (m == NULL || m->m_len != sizeof(int)) 1236 error = EINVAL; 1237 else { 1238 optval = *mtod(m, int *); 1239 switch (optname) { 1240 1241 case IP_TOS: 1242 inp->inp_ip.ip_tos = optval; 1243 break; 1244 1245 case IP_TTL: 1246 inp->inp_ip.ip_ttl = optval; 1247 break; 1248 #define OPTSET(bit) \ 1249 if (optval) \ 1250 inp->inp_flags |= bit; \ 1251 else \ 1252 inp->inp_flags &= ~bit; 1253 1254 case IP_RECVOPTS: 1255 OPTSET(INP_RECVOPTS); 1256 break; 1257 1258 case IP_RECVRETOPTS: 1259 OPTSET(INP_RECVRETOPTS); 1260 break; 1261 1262 case IP_RECVDSTADDR: 1263 OPTSET(INP_RECVDSTADDR); 1264 break; 1265 1266 case IP_RECVIF: 1267 OPTSET(INP_RECVIF); 1268 break; 1269 } 1270 } 1271 break; 1272 #undef OPTSET 1273 1274 case IP_MULTICAST_IF: 1275 case IP_MULTICAST_TTL: 1276 case IP_MULTICAST_LOOP: 1277 case IP_ADD_MEMBERSHIP: 1278 case IP_DROP_MEMBERSHIP: 1279 error = ip_setmoptions(optname, &inp->inp_moptions, m); 1280 break; 1281 1282 case IP_PORTRANGE: 1283 if (m == 0 || m->m_len != sizeof(int)) 1284 error = EINVAL; 1285 else { 1286 optval = *mtod(m, int *); 1287 1288 switch (optval) { 1289 1290 case IP_PORTRANGE_DEFAULT: 1291 case IP_PORTRANGE_HIGH: 1292 inp->inp_flags &= ~(INP_LOWPORT); 1293 break; 1294 1295 case IP_PORTRANGE_LOW: 1296 inp->inp_flags |= INP_LOWPORT; 1297 break; 1298 1299 default: 1300 error = EINVAL; 1301 break; 1302 } 1303 } 1304 break; 1305 1306 #if defined(IPSEC) || defined(FAST_IPSEC) 1307 case IP_IPSEC_POLICY: 1308 { 1309 void *req = NULL; 1310 size_t len = 0; 1311 int priv = 0; 1312 1313 #ifdef __NetBSD__ 1314 if (l == 0 || kauth_authorize_generic(l->l_cred, 1315 KAUTH_GENERIC_ISSUSER, NULL)) 1316 priv = 0; 1317 else 1318 priv = 1; 1319 #else 1320 priv = (in6p->in6p_socket->so_state & SS_PRIV); 1321 #endif 1322 if (m) { 1323 req = mtod(m, void *); 1324 len = m->m_len; 1325 } 1326 error = ipsec4_set_policy(inp, optname, req, len, priv); 1327 break; 1328 } 1329 #endif /*IPSEC*/ 1330 1331 default: 1332 error = ENOPROTOOPT; 1333 break; 1334 } 1335 if (m) 1336 (void)m_free(m); 1337 break; 1338 1339 case PRCO_GETOPT: 1340 switch (optname) { 1341 case IP_OPTIONS: 1342 case IP_RETOPTS: 1343 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1344 MCLAIM(m, so->so_mowner); 1345 if (inp->inp_options) { 1346 m->m_len = inp->inp_options->m_len; 1347 bcopy(mtod(inp->inp_options, void *), 1348 mtod(m, void *), (unsigned)m->m_len); 1349 } else 1350 m->m_len = 0; 1351 break; 1352 1353 case IP_TOS: 1354 case IP_TTL: 1355 case IP_RECVOPTS: 1356 case IP_RECVRETOPTS: 1357 case IP_RECVDSTADDR: 1358 case IP_RECVIF: 1359 case IP_ERRORMTU: 1360 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1361 MCLAIM(m, so->so_mowner); 1362 m->m_len = sizeof(int); 1363 switch (optname) { 1364 1365 case IP_TOS: 1366 optval = inp->inp_ip.ip_tos; 1367 break; 1368 1369 case IP_TTL: 1370 optval = inp->inp_ip.ip_ttl; 1371 break; 1372 1373 case IP_ERRORMTU: 1374 optval = inp->inp_errormtu; 1375 break; 1376 1377 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1378 1379 case IP_RECVOPTS: 1380 optval = OPTBIT(INP_RECVOPTS); 1381 break; 1382 1383 case IP_RECVRETOPTS: 1384 optval = OPTBIT(INP_RECVRETOPTS); 1385 break; 1386 1387 case IP_RECVDSTADDR: 1388 optval = OPTBIT(INP_RECVDSTADDR); 1389 break; 1390 1391 case IP_RECVIF: 1392 optval = OPTBIT(INP_RECVIF); 1393 break; 1394 } 1395 *mtod(m, int *) = optval; 1396 break; 1397 1398 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */ 1399 /* XXX: code broken */ 1400 case IP_IPSEC_POLICY: 1401 { 1402 void *req = NULL; 1403 size_t len = 0; 1404 1405 if (m) { 1406 req = mtod(m, void *); 1407 len = m->m_len; 1408 } 1409 error = ipsec4_get_policy(inp, req, len, mp); 1410 break; 1411 } 1412 #endif /*IPSEC*/ 1413 1414 case IP_MULTICAST_IF: 1415 case IP_MULTICAST_TTL: 1416 case IP_MULTICAST_LOOP: 1417 case IP_ADD_MEMBERSHIP: 1418 case IP_DROP_MEMBERSHIP: 1419 error = ip_getmoptions(optname, inp->inp_moptions, mp); 1420 if (*mp) 1421 MCLAIM(*mp, so->so_mowner); 1422 break; 1423 1424 case IP_PORTRANGE: 1425 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1426 MCLAIM(m, so->so_mowner); 1427 m->m_len = sizeof(int); 1428 1429 if (inp->inp_flags & INP_LOWPORT) 1430 optval = IP_PORTRANGE_LOW; 1431 else 1432 optval = IP_PORTRANGE_DEFAULT; 1433 1434 *mtod(m, int *) = optval; 1435 break; 1436 1437 default: 1438 error = ENOPROTOOPT; 1439 break; 1440 } 1441 break; 1442 } 1443 return (error); 1444 } 1445 1446 /* 1447 * Set up IP options in pcb for insertion in output packets. 1448 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1449 * with destination address if source routed. 1450 */ 1451 int 1452 #ifdef notyet 1453 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m) 1454 #else 1455 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m) 1456 #endif 1457 { 1458 int cnt, optlen; 1459 u_char *cp; 1460 u_char opt; 1461 1462 /* turn off any old options */ 1463 if (*pcbopt) 1464 (void)m_free(*pcbopt); 1465 *pcbopt = 0; 1466 if (m == (struct mbuf *)0 || m->m_len == 0) { 1467 /* 1468 * Only turning off any previous options. 1469 */ 1470 if (m) 1471 (void)m_free(m); 1472 return (0); 1473 } 1474 1475 #ifndef __vax__ 1476 if (m->m_len % sizeof(int32_t)) 1477 goto bad; 1478 #endif 1479 /* 1480 * IP first-hop destination address will be stored before 1481 * actual options; move other options back 1482 * and clear it when none present. 1483 */ 1484 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1485 goto bad; 1486 cnt = m->m_len; 1487 m->m_len += sizeof(struct in_addr); 1488 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1489 memmove(cp, mtod(m, void *), (unsigned)cnt); 1490 bzero(mtod(m, void *), sizeof(struct in_addr)); 1491 1492 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1493 opt = cp[IPOPT_OPTVAL]; 1494 if (opt == IPOPT_EOL) 1495 break; 1496 if (opt == IPOPT_NOP) 1497 optlen = 1; 1498 else { 1499 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1500 goto bad; 1501 optlen = cp[IPOPT_OLEN]; 1502 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1503 goto bad; 1504 } 1505 switch (opt) { 1506 1507 default: 1508 break; 1509 1510 case IPOPT_LSRR: 1511 case IPOPT_SSRR: 1512 /* 1513 * user process specifies route as: 1514 * ->A->B->C->D 1515 * D must be our final destination (but we can't 1516 * check that since we may not have connected yet). 1517 * A is first hop destination, which doesn't appear in 1518 * actual IP option, but is stored before the options. 1519 */ 1520 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1521 goto bad; 1522 m->m_len -= sizeof(struct in_addr); 1523 cnt -= sizeof(struct in_addr); 1524 optlen -= sizeof(struct in_addr); 1525 cp[IPOPT_OLEN] = optlen; 1526 /* 1527 * Move first hop before start of options. 1528 */ 1529 bcopy((void *)&cp[IPOPT_OFFSET+1], mtod(m, void *), 1530 sizeof(struct in_addr)); 1531 /* 1532 * Then copy rest of options back 1533 * to close up the deleted entry. 1534 */ 1535 (void)memmove(&cp[IPOPT_OFFSET+1], 1536 &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr), 1537 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1538 break; 1539 } 1540 } 1541 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1542 goto bad; 1543 *pcbopt = m; 1544 return (0); 1545 1546 bad: 1547 (void)m_free(m); 1548 return (EINVAL); 1549 } 1550 1551 /* 1552 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1553 */ 1554 static struct ifnet * 1555 ip_multicast_if(struct in_addr *a, int *ifindexp) 1556 { 1557 int ifindex; 1558 struct ifnet *ifp = NULL; 1559 struct in_ifaddr *ia; 1560 1561 if (ifindexp) 1562 *ifindexp = 0; 1563 if (ntohl(a->s_addr) >> 24 == 0) { 1564 ifindex = ntohl(a->s_addr) & 0xffffff; 1565 if (ifindex < 0 || if_indexlim <= ifindex) 1566 return NULL; 1567 ifp = ifindex2ifnet[ifindex]; 1568 if (!ifp) 1569 return NULL; 1570 if (ifindexp) 1571 *ifindexp = ifindex; 1572 } else { 1573 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1574 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1575 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1576 ifp = ia->ia_ifp; 1577 break; 1578 } 1579 } 1580 } 1581 return ifp; 1582 } 1583 1584 static int 1585 ip_getoptval(struct mbuf *m, u_int8_t *val, u_int maxval) 1586 { 1587 u_int tval; 1588 1589 if (m == NULL) 1590 return EINVAL; 1591 1592 switch (m->m_len) { 1593 case sizeof(u_char): 1594 tval = *(mtod(m, u_char *)); 1595 break; 1596 case sizeof(u_int): 1597 tval = *(mtod(m, u_int *)); 1598 break; 1599 default: 1600 return EINVAL; 1601 } 1602 1603 if (tval > maxval) 1604 return EINVAL; 1605 1606 *val = tval; 1607 return 0; 1608 } 1609 1610 /* 1611 * Set the IP multicast options in response to user setsockopt(). 1612 */ 1613 int 1614 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m) 1615 { 1616 int error = 0; 1617 int i; 1618 struct in_addr addr; 1619 struct ip_mreq *mreq; 1620 struct ifnet *ifp; 1621 struct ip_moptions *imo = *imop; 1622 int ifindex; 1623 1624 if (imo == NULL) { 1625 /* 1626 * No multicast option buffer attached to the pcb; 1627 * allocate one and initialize to default values. 1628 */ 1629 imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS, 1630 M_WAITOK); 1631 1632 if (imo == NULL) 1633 return (ENOBUFS); 1634 *imop = imo; 1635 imo->imo_multicast_ifp = NULL; 1636 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1637 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1638 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1639 imo->imo_num_memberships = 0; 1640 } 1641 1642 switch (optname) { 1643 1644 case IP_MULTICAST_IF: 1645 /* 1646 * Select the interface for outgoing multicast packets. 1647 */ 1648 if (m == NULL || m->m_len != sizeof(struct in_addr)) { 1649 error = EINVAL; 1650 break; 1651 } 1652 addr = *(mtod(m, struct in_addr *)); 1653 /* 1654 * INADDR_ANY is used to remove a previous selection. 1655 * When no interface is selected, a default one is 1656 * chosen every time a multicast packet is sent. 1657 */ 1658 if (in_nullhost(addr)) { 1659 imo->imo_multicast_ifp = NULL; 1660 break; 1661 } 1662 /* 1663 * The selected interface is identified by its local 1664 * IP address. Find the interface and confirm that 1665 * it supports multicasting. 1666 */ 1667 ifp = ip_multicast_if(&addr, &ifindex); 1668 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1669 error = EADDRNOTAVAIL; 1670 break; 1671 } 1672 imo->imo_multicast_ifp = ifp; 1673 if (ifindex) 1674 imo->imo_multicast_addr = addr; 1675 else 1676 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1677 break; 1678 1679 case IP_MULTICAST_TTL: 1680 /* 1681 * Set the IP time-to-live for outgoing multicast packets. 1682 */ 1683 error = ip_getoptval(m, &imo->imo_multicast_ttl, MAXTTL); 1684 break; 1685 1686 case IP_MULTICAST_LOOP: 1687 /* 1688 * Set the loopback flag for outgoing multicast packets. 1689 * Must be zero or one. 1690 */ 1691 error = ip_getoptval(m, &imo->imo_multicast_loop, 1); 1692 break; 1693 1694 case IP_ADD_MEMBERSHIP: 1695 /* 1696 * Add a multicast group membership. 1697 * Group must be a valid IP multicast address. 1698 */ 1699 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1700 error = EINVAL; 1701 break; 1702 } 1703 mreq = mtod(m, struct ip_mreq *); 1704 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1705 error = EINVAL; 1706 break; 1707 } 1708 /* 1709 * If no interface address was provided, use the interface of 1710 * the route to the given multicast address. 1711 */ 1712 if (in_nullhost(mreq->imr_interface)) { 1713 struct rtentry *rt; 1714 union { 1715 struct sockaddr dst; 1716 struct sockaddr_in dst4; 1717 } u; 1718 struct route ro; 1719 1720 memset(&ro, 0, sizeof(ro)); 1721 1722 sockaddr_in_init(&u.dst4, &mreq->imr_multiaddr, 0); 1723 rtcache_setdst(&ro, &u.dst); 1724 ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp 1725 : NULL; 1726 rtcache_free(&ro); 1727 } else { 1728 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1729 } 1730 /* 1731 * See if we found an interface, and confirm that it 1732 * supports multicast. 1733 */ 1734 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1735 error = EADDRNOTAVAIL; 1736 break; 1737 } 1738 /* 1739 * See if the membership already exists or if all the 1740 * membership slots are full. 1741 */ 1742 for (i = 0; i < imo->imo_num_memberships; ++i) { 1743 if (imo->imo_membership[i]->inm_ifp == ifp && 1744 in_hosteq(imo->imo_membership[i]->inm_addr, 1745 mreq->imr_multiaddr)) 1746 break; 1747 } 1748 if (i < imo->imo_num_memberships) { 1749 error = EADDRINUSE; 1750 break; 1751 } 1752 if (i == IP_MAX_MEMBERSHIPS) { 1753 error = ETOOMANYREFS; 1754 break; 1755 } 1756 /* 1757 * Everything looks good; add a new record to the multicast 1758 * address list for the given interface. 1759 */ 1760 if ((imo->imo_membership[i] = 1761 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1762 error = ENOBUFS; 1763 break; 1764 } 1765 ++imo->imo_num_memberships; 1766 break; 1767 1768 case IP_DROP_MEMBERSHIP: 1769 /* 1770 * Drop a multicast group membership. 1771 * Group must be a valid IP multicast address. 1772 */ 1773 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1774 error = EINVAL; 1775 break; 1776 } 1777 mreq = mtod(m, struct ip_mreq *); 1778 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1779 error = EINVAL; 1780 break; 1781 } 1782 /* 1783 * If an interface address was specified, get a pointer 1784 * to its ifnet structure. 1785 */ 1786 if (in_nullhost(mreq->imr_interface)) 1787 ifp = NULL; 1788 else { 1789 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1790 if (ifp == NULL) { 1791 error = EADDRNOTAVAIL; 1792 break; 1793 } 1794 } 1795 /* 1796 * Find the membership in the membership array. 1797 */ 1798 for (i = 0; i < imo->imo_num_memberships; ++i) { 1799 if ((ifp == NULL || 1800 imo->imo_membership[i]->inm_ifp == ifp) && 1801 in_hosteq(imo->imo_membership[i]->inm_addr, 1802 mreq->imr_multiaddr)) 1803 break; 1804 } 1805 if (i == imo->imo_num_memberships) { 1806 error = EADDRNOTAVAIL; 1807 break; 1808 } 1809 /* 1810 * Give up the multicast address record to which the 1811 * membership points. 1812 */ 1813 in_delmulti(imo->imo_membership[i]); 1814 /* 1815 * Remove the gap in the membership array. 1816 */ 1817 for (++i; i < imo->imo_num_memberships; ++i) 1818 imo->imo_membership[i-1] = imo->imo_membership[i]; 1819 --imo->imo_num_memberships; 1820 break; 1821 1822 default: 1823 error = EOPNOTSUPP; 1824 break; 1825 } 1826 1827 /* 1828 * If all options have default values, no need to keep the mbuf. 1829 */ 1830 if (imo->imo_multicast_ifp == NULL && 1831 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1832 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1833 imo->imo_num_memberships == 0) { 1834 free(*imop, M_IPMOPTS); 1835 *imop = NULL; 1836 } 1837 1838 return (error); 1839 } 1840 1841 /* 1842 * Return the IP multicast options in response to user getsockopt(). 1843 */ 1844 int 1845 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp) 1846 { 1847 u_char *ttl; 1848 u_char *loop; 1849 struct in_addr *addr; 1850 struct in_ifaddr *ia; 1851 1852 *mp = m_get(M_WAIT, MT_SOOPTS); 1853 1854 switch (optname) { 1855 1856 case IP_MULTICAST_IF: 1857 addr = mtod(*mp, struct in_addr *); 1858 (*mp)->m_len = sizeof(struct in_addr); 1859 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1860 *addr = zeroin_addr; 1861 else if (imo->imo_multicast_addr.s_addr) { 1862 /* return the value user has set */ 1863 *addr = imo->imo_multicast_addr; 1864 } else { 1865 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1866 *addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1867 } 1868 return (0); 1869 1870 case IP_MULTICAST_TTL: 1871 ttl = mtod(*mp, u_char *); 1872 (*mp)->m_len = 1; 1873 *ttl = imo ? imo->imo_multicast_ttl 1874 : IP_DEFAULT_MULTICAST_TTL; 1875 return (0); 1876 1877 case IP_MULTICAST_LOOP: 1878 loop = mtod(*mp, u_char *); 1879 (*mp)->m_len = 1; 1880 *loop = imo ? imo->imo_multicast_loop 1881 : IP_DEFAULT_MULTICAST_LOOP; 1882 return (0); 1883 1884 default: 1885 return (EOPNOTSUPP); 1886 } 1887 } 1888 1889 /* 1890 * Discard the IP multicast options. 1891 */ 1892 void 1893 ip_freemoptions(struct ip_moptions *imo) 1894 { 1895 int i; 1896 1897 if (imo != NULL) { 1898 for (i = 0; i < imo->imo_num_memberships; ++i) 1899 in_delmulti(imo->imo_membership[i]); 1900 free(imo, M_IPMOPTS); 1901 } 1902 } 1903 1904 /* 1905 * Routine called from ip_output() to loop back a copy of an IP multicast 1906 * packet to the input queue of a specified interface. Note that this 1907 * calls the output routine of the loopback "driver", but with an interface 1908 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1909 */ 1910 static void 1911 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 1912 { 1913 struct ip *ip; 1914 struct mbuf *copym; 1915 1916 copym = m_copypacket(m, M_DONTWAIT); 1917 if (copym != NULL 1918 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1919 copym = m_pullup(copym, sizeof(struct ip)); 1920 if (copym == NULL) 1921 return; 1922 /* 1923 * We don't bother to fragment if the IP length is greater 1924 * than the interface's MTU. Can this possibly matter? 1925 */ 1926 ip = mtod(copym, struct ip *); 1927 1928 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1929 in_delayed_cksum(copym); 1930 copym->m_pkthdr.csum_flags &= 1931 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1932 } 1933 1934 ip->ip_sum = 0; 1935 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1936 (void)looutput(ifp, copym, sintocsa(dst), NULL); 1937 } 1938