1 /* $NetBSD: ip_output.c,v 1.192 2008/02/06 03:20:51 matt 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 1.192 2008/02/06 03:20:51 matt 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 #include <sys/kauth.h> 116 #ifdef FAST_IPSEC 117 #include <sys/domain.h> 118 #endif 119 #include <sys/systm.h> 120 #include <sys/proc.h> 121 122 #include <net/if.h> 123 #include <net/route.h> 124 #include <net/pfil.h> 125 126 #include <netinet/in.h> 127 #include <netinet/in_systm.h> 128 #include <netinet/ip.h> 129 #include <netinet/in_pcb.h> 130 #include <netinet/in_var.h> 131 #include <netinet/ip_var.h> 132 #include <netinet/in_offload.h> 133 134 #ifdef MROUTING 135 #include <netinet/ip_mroute.h> 136 #endif 137 138 #include <machine/stdarg.h> 139 140 #ifdef IPSEC 141 #include <netinet6/ipsec.h> 142 #include <netkey/key.h> 143 #include <netkey/key_debug.h> 144 #endif /*IPSEC*/ 145 146 #ifdef FAST_IPSEC 147 #include <netipsec/ipsec.h> 148 #include <netipsec/key.h> 149 #include <netipsec/xform.h> 150 #endif /* FAST_IPSEC*/ 151 152 #ifdef IPSEC_NAT_T 153 #include <netinet/udp.h> 154 #endif 155 156 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 157 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 158 static void ip_mloopback(struct ifnet *, struct mbuf *, 159 const struct sockaddr_in *); 160 static int ip_getoptval(struct mbuf *, u_int8_t *, u_int); 161 162 #ifdef PFIL_HOOKS 163 extern struct pfil_head inet_pfil_hook; /* XXX */ 164 #endif 165 166 int ip_do_loopback_cksum = 0; 167 168 /* 169 * IP output. The packet in mbuf chain m contains a skeletal IP 170 * header (with len, off, ttl, proto, tos, src, dst). 171 * The mbuf chain containing the packet will be freed. 172 * The mbuf opt, if present, will not be freed. 173 */ 174 int 175 ip_output(struct mbuf *m0, ...) 176 { 177 struct rtentry *rt; 178 struct ip *ip; 179 struct ifnet *ifp; 180 struct mbuf *m = m0; 181 int hlen = sizeof (struct ip); 182 int len, error = 0; 183 struct route iproute; 184 const struct sockaddr_in *dst; 185 struct in_ifaddr *ia; 186 struct ifaddr *xifa; 187 struct mbuf *opt; 188 struct route *ro; 189 int flags, sw_csum; 190 int *mtu_p; 191 u_long mtu; 192 struct ip_moptions *imo; 193 struct socket *so; 194 va_list ap; 195 #ifdef IPSEC_NAT_T 196 int natt_frag = 0; 197 #endif 198 #ifdef IPSEC 199 struct secpolicy *sp = NULL; 200 #endif /*IPSEC*/ 201 #ifdef FAST_IPSEC 202 struct inpcb *inp; 203 struct secpolicy *sp = NULL; 204 int s; 205 #endif 206 u_int16_t ip_len; 207 union { 208 struct sockaddr dst; 209 struct sockaddr_in dst4; 210 } u; 211 struct sockaddr *rdst = &u.dst; /* real IP destination, as opposed 212 * to the nexthop 213 */ 214 215 len = 0; 216 va_start(ap, m0); 217 opt = va_arg(ap, struct mbuf *); 218 ro = va_arg(ap, struct route *); 219 flags = va_arg(ap, int); 220 imo = va_arg(ap, struct ip_moptions *); 221 so = va_arg(ap, struct socket *); 222 if (flags & IP_RETURNMTU) 223 mtu_p = va_arg(ap, int *); 224 else 225 mtu_p = NULL; 226 va_end(ap); 227 228 MCLAIM(m, &ip_tx_mowner); 229 #ifdef FAST_IPSEC 230 if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET) 231 inp = (struct inpcb *)so->so_pcb; 232 else 233 inp = NULL; 234 #endif /* FAST_IPSEC */ 235 236 #ifdef DIAGNOSTIC 237 if ((m->m_flags & M_PKTHDR) == 0) 238 panic("ip_output: no HDR"); 239 240 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) != 0) { 241 panic("ip_output: IPv6 checksum offload flags: %d", 242 m->m_pkthdr.csum_flags); 243 } 244 245 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) == 246 (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 247 panic("ip_output: conflicting checksum offload flags: %d", 248 m->m_pkthdr.csum_flags); 249 } 250 #endif 251 if (opt) { 252 m = ip_insertoptions(m, opt, &len); 253 if (len >= sizeof(struct ip)) 254 hlen = len; 255 } 256 ip = mtod(m, struct ip *); 257 /* 258 * Fill in IP header. 259 */ 260 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 261 ip->ip_v = IPVERSION; 262 ip->ip_off = htons(0); 263 /* ip->ip_id filled in after we find out source ia */ 264 ip->ip_hl = hlen >> 2; 265 ipstat.ips_localout++; 266 } else { 267 hlen = ip->ip_hl << 2; 268 } 269 /* 270 * Route packet. 271 */ 272 memset(&iproute, 0, sizeof(iproute)); 273 if (ro == NULL) 274 ro = &iproute; 275 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 276 dst = satocsin(rtcache_getdst(ro)); 277 /* 278 * If there is a cached route, 279 * check that it is to the same destination 280 * and is still up. If not, free it and try again. 281 * The address family should also be checked in case of sharing the 282 * cache with IPv6. 283 */ 284 if (dst == NULL) 285 ; 286 else if (dst->sin_family != AF_INET || 287 !in_hosteq(dst->sin_addr, ip->ip_dst)) 288 rtcache_free(ro); 289 290 if ((rt = rtcache_validate(ro)) == NULL && 291 (rt = rtcache_update(ro, 1)) == NULL) { 292 dst = &u.dst4; 293 rtcache_setdst(ro, &u.dst); 294 } 295 /* 296 * If routing to interface only, 297 * short circuit routing lookup. 298 */ 299 if (flags & IP_ROUTETOIF) { 300 if ((ia = ifatoia(ifa_ifwithladdr(sintocsa(dst)))) == NULL) { 301 ipstat.ips_noroute++; 302 error = ENETUNREACH; 303 goto bad; 304 } 305 ifp = ia->ia_ifp; 306 mtu = ifp->if_mtu; 307 ip->ip_ttl = 1; 308 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 309 ip->ip_dst.s_addr == INADDR_BROADCAST) && 310 imo != NULL && imo->imo_multicast_ifp != NULL) { 311 ifp = imo->imo_multicast_ifp; 312 mtu = ifp->if_mtu; 313 IFP_TO_IA(ifp, ia); 314 } else { 315 if (rt == NULL) 316 rt = rtcache_init(ro); 317 if (rt == NULL) { 318 ipstat.ips_noroute++; 319 error = EHOSTUNREACH; 320 goto bad; 321 } 322 ia = ifatoia(rt->rt_ifa); 323 ifp = rt->rt_ifp; 324 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 325 mtu = ifp->if_mtu; 326 rt->rt_use++; 327 if (rt->rt_flags & RTF_GATEWAY) 328 dst = satosin(rt->rt_gateway); 329 } 330 if (IN_MULTICAST(ip->ip_dst.s_addr) || 331 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 332 struct in_multi *inm; 333 334 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 335 M_BCAST : M_MCAST; 336 /* 337 * See if the caller provided any multicast options 338 */ 339 if (imo != NULL) 340 ip->ip_ttl = imo->imo_multicast_ttl; 341 else 342 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 343 344 /* 345 * if we don't know the outgoing ifp yet, we can't generate 346 * output 347 */ 348 if (!ifp) { 349 ipstat.ips_noroute++; 350 error = ENETUNREACH; 351 goto bad; 352 } 353 354 /* 355 * If the packet is multicast or broadcast, confirm that 356 * the outgoing interface can transmit it. 357 */ 358 if (((m->m_flags & M_MCAST) && 359 (ifp->if_flags & IFF_MULTICAST) == 0) || 360 ((m->m_flags & M_BCAST) && 361 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 362 ipstat.ips_noroute++; 363 error = ENETUNREACH; 364 goto bad; 365 } 366 /* 367 * If source address not specified yet, use an address 368 * of outgoing interface. 369 */ 370 if (in_nullhost(ip->ip_src)) { 371 struct in_ifaddr *xia; 372 373 IFP_TO_IA(ifp, xia); 374 if (!xia) { 375 error = EADDRNOTAVAIL; 376 goto bad; 377 } 378 xifa = &xia->ia_ifa; 379 if (xifa->ifa_getifa != NULL) { 380 xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 381 } 382 ip->ip_src = xia->ia_addr.sin_addr; 383 } 384 385 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 386 if (inm != NULL && 387 (imo == NULL || imo->imo_multicast_loop)) { 388 /* 389 * If we belong to the destination multicast group 390 * on the outgoing interface, and the caller did not 391 * forbid loopback, loop back a copy. 392 */ 393 ip_mloopback(ifp, m, &u.dst4); 394 } 395 #ifdef MROUTING 396 else { 397 /* 398 * If we are acting as a multicast router, perform 399 * multicast forwarding as if the packet had just 400 * arrived on the interface to which we are about 401 * to send. The multicast forwarding function 402 * recursively calls this function, using the 403 * IP_FORWARDING flag to prevent infinite recursion. 404 * 405 * Multicasts that are looped back by ip_mloopback(), 406 * above, will be forwarded by the ip_input() routine, 407 * if necessary. 408 */ 409 extern struct socket *ip_mrouter; 410 411 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 412 if (ip_mforward(m, ifp) != 0) { 413 m_freem(m); 414 goto done; 415 } 416 } 417 } 418 #endif 419 /* 420 * Multicasts with a time-to-live of zero may be looped- 421 * back, above, but must not be transmitted on a network. 422 * Also, multicasts addressed to the loopback interface 423 * are not sent -- the above call to ip_mloopback() will 424 * loop back a copy if this host actually belongs to the 425 * destination group on the loopback interface. 426 */ 427 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 428 m_freem(m); 429 goto done; 430 } 431 432 goto sendit; 433 } 434 /* 435 * If source address not specified yet, use address 436 * of outgoing interface. 437 */ 438 if (in_nullhost(ip->ip_src)) { 439 xifa = &ia->ia_ifa; 440 if (xifa->ifa_getifa != NULL) 441 ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 442 ip->ip_src = ia->ia_addr.sin_addr; 443 } 444 445 /* 446 * packets with Class-D address as source are not valid per 447 * RFC 1112 448 */ 449 if (IN_MULTICAST(ip->ip_src.s_addr)) { 450 ipstat.ips_odropped++; 451 error = EADDRNOTAVAIL; 452 goto bad; 453 } 454 455 /* 456 * Look for broadcast address and 457 * and verify user is allowed to send 458 * such a packet. 459 */ 460 if (in_broadcast(dst->sin_addr, ifp)) { 461 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 462 error = EADDRNOTAVAIL; 463 goto bad; 464 } 465 if ((flags & IP_ALLOWBROADCAST) == 0) { 466 error = EACCES; 467 goto bad; 468 } 469 /* don't allow broadcast messages to be fragmented */ 470 if (ntohs(ip->ip_len) > ifp->if_mtu) { 471 error = EMSGSIZE; 472 goto bad; 473 } 474 m->m_flags |= M_BCAST; 475 } else 476 m->m_flags &= ~M_BCAST; 477 478 sendit: 479 if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) { 480 if (m->m_pkthdr.len < IP_MINFRAGSIZE) { 481 ip->ip_id = 0; 482 } else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 483 ip->ip_id = ip_newid(ia); 484 } else { 485 486 /* 487 * TSO capable interfaces (typically?) increment 488 * ip_id for each segment. 489 * "allocate" enough ids here to increase the chance 490 * for them to be unique. 491 * 492 * note that the following calculation is not 493 * needed to be precise. wasting some ip_id is fine. 494 */ 495 496 unsigned int segsz = m->m_pkthdr.segsz; 497 unsigned int datasz = ntohs(ip->ip_len) - hlen; 498 unsigned int num = howmany(datasz, segsz); 499 500 ip->ip_id = ip_newid_range(ia, num); 501 } 502 } 503 /* 504 * If we're doing Path MTU Discovery, we need to set DF unless 505 * the route's MTU is locked. 506 */ 507 if ((flags & IP_MTUDISC) != 0 && rt != NULL && 508 (rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 509 ip->ip_off |= htons(IP_DF); 510 511 /* Remember the current ip_len */ 512 ip_len = ntohs(ip->ip_len); 513 514 #ifdef IPSEC 515 /* get SP for this packet */ 516 if (so == NULL) 517 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 518 flags, &error); 519 else { 520 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 521 IPSEC_DIR_OUTBOUND)) 522 goto skip_ipsec; 523 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 524 } 525 526 if (sp == NULL) { 527 ipsecstat.out_inval++; 528 goto bad; 529 } 530 531 error = 0; 532 533 /* check policy */ 534 switch (sp->policy) { 535 case IPSEC_POLICY_DISCARD: 536 /* 537 * This packet is just discarded. 538 */ 539 ipsecstat.out_polvio++; 540 goto bad; 541 542 case IPSEC_POLICY_BYPASS: 543 case IPSEC_POLICY_NONE: 544 /* no need to do IPsec. */ 545 goto skip_ipsec; 546 547 case IPSEC_POLICY_IPSEC: 548 if (sp->req == NULL) { 549 /* XXX should be panic ? */ 550 printf("ip_output: No IPsec request specified.\n"); 551 error = EINVAL; 552 goto bad; 553 } 554 break; 555 556 case IPSEC_POLICY_ENTRUST: 557 default: 558 printf("ip_output: Invalid policy found. %d\n", sp->policy); 559 } 560 561 #ifdef IPSEC_NAT_T 562 /* 563 * NAT-T ESP fragmentation: don't do IPSec processing now, 564 * we'll do it on each fragmented packet. 565 */ 566 if (sp->req->sav && 567 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 568 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 569 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 570 natt_frag = 1; 571 mtu = sp->req->sav->esp_frag; 572 goto skip_ipsec; 573 } 574 } 575 #endif /* IPSEC_NAT_T */ 576 577 /* 578 * ipsec4_output() expects ip_len and ip_off in network 579 * order. They have been set to network order above. 580 */ 581 582 { 583 struct ipsec_output_state state; 584 bzero(&state, sizeof(state)); 585 state.m = m; 586 if (flags & IP_ROUTETOIF) { 587 state.ro = &iproute; 588 memset(&iproute, 0, sizeof(iproute)); 589 } else 590 state.ro = ro; 591 state.dst = sintocsa(dst); 592 593 /* 594 * We can't defer the checksum of payload data if 595 * we're about to encrypt/authenticate it. 596 * 597 * XXX When we support crypto offloading functions of 598 * XXX network interfaces, we need to reconsider this, 599 * XXX since it's likely that they'll support checksumming, 600 * XXX as well. 601 */ 602 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 603 in_delayed_cksum(m); 604 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 605 } 606 607 error = ipsec4_output(&state, sp, flags); 608 609 m = state.m; 610 if (flags & IP_ROUTETOIF) { 611 /* 612 * if we have tunnel mode SA, we may need to ignore 613 * IP_ROUTETOIF. 614 */ 615 if (state.ro != &iproute || 616 rtcache_validate(state.ro) != NULL) { 617 flags &= ~IP_ROUTETOIF; 618 ro = state.ro; 619 } 620 } else 621 ro = state.ro; 622 dst = satocsin(state.dst); 623 if (error) { 624 /* mbuf is already reclaimed in ipsec4_output. */ 625 m0 = NULL; 626 switch (error) { 627 case EHOSTUNREACH: 628 case ENETUNREACH: 629 case EMSGSIZE: 630 case ENOBUFS: 631 case ENOMEM: 632 break; 633 default: 634 printf("ip4_output (ipsec): error code %d\n", error); 635 /*fall through*/ 636 case ENOENT: 637 /* don't show these error codes to the user */ 638 error = 0; 639 break; 640 } 641 goto bad; 642 } 643 644 /* be sure to update variables that are affected by ipsec4_output() */ 645 ip = mtod(m, struct ip *); 646 hlen = ip->ip_hl << 2; 647 ip_len = ntohs(ip->ip_len); 648 649 if ((rt = rtcache_validate(ro)) == NULL) { 650 if ((flags & IP_ROUTETOIF) == 0) { 651 printf("ip_output: " 652 "can't update route after IPsec processing\n"); 653 error = EHOSTUNREACH; /*XXX*/ 654 goto bad; 655 } 656 } else { 657 /* nobody uses ia beyond here */ 658 if (state.encap) { 659 ifp = rt->rt_ifp; 660 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 661 mtu = ifp->if_mtu; 662 } 663 } 664 } 665 skip_ipsec: 666 #endif /*IPSEC*/ 667 #ifdef FAST_IPSEC 668 /* 669 * Check the security policy (SP) for the packet and, if 670 * required, do IPsec-related processing. There are two 671 * cases here; the first time a packet is sent through 672 * it will be untagged and handled by ipsec4_checkpolicy. 673 * If the packet is resubmitted to ip_output (e.g. after 674 * AH, ESP, etc. processing), there will be a tag to bypass 675 * the lookup and related policy checking. 676 */ 677 if (!ipsec_outdone(m)) { 678 s = splsoftnet(); 679 if (inp != NULL && 680 IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND)) 681 goto spd_done; 682 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 683 &error, inp); 684 /* 685 * There are four return cases: 686 * sp != NULL apply IPsec policy 687 * sp == NULL, error == 0 no IPsec handling needed 688 * sp == NULL, error == -EINVAL discard packet w/o error 689 * sp == NULL, error != 0 discard packet, report error 690 */ 691 if (sp != NULL) { 692 #ifdef IPSEC_NAT_T 693 /* 694 * NAT-T ESP fragmentation: don't do IPSec processing now, 695 * we'll do it on each fragmented packet. 696 */ 697 if (sp->req->sav && 698 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 699 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 700 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 701 natt_frag = 1; 702 mtu = sp->req->sav->esp_frag; 703 splx(s); 704 goto spd_done; 705 } 706 } 707 #endif /* IPSEC_NAT_T */ 708 709 /* 710 * Do delayed checksums now because we send before 711 * this is done in the normal processing path. 712 */ 713 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 714 in_delayed_cksum(m); 715 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 716 } 717 718 #ifdef __FreeBSD__ 719 ip->ip_len = htons(ip->ip_len); 720 ip->ip_off = htons(ip->ip_off); 721 #endif 722 723 /* NB: callee frees mbuf */ 724 error = ipsec4_process_packet(m, sp->req, flags, 0); 725 /* 726 * Preserve KAME behaviour: ENOENT can be returned 727 * when an SA acquire is in progress. Don't propagate 728 * this to user-level; it confuses applications. 729 * 730 * XXX this will go away when the SADB is redone. 731 */ 732 if (error == ENOENT) 733 error = 0; 734 splx(s); 735 goto done; 736 } else { 737 splx(s); 738 739 if (error != 0) { 740 /* 741 * Hack: -EINVAL is used to signal that a packet 742 * should be silently discarded. This is typically 743 * because we asked key management for an SA and 744 * it was delayed (e.g. kicked up to IKE). 745 */ 746 if (error == -EINVAL) 747 error = 0; 748 goto bad; 749 } else { 750 /* No IPsec processing for this packet. */ 751 } 752 } 753 } 754 spd_done: 755 #endif /* FAST_IPSEC */ 756 757 #ifdef PFIL_HOOKS 758 /* 759 * Run through list of hooks for output packets. 760 */ 761 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 762 goto done; 763 if (m == NULL) 764 goto done; 765 766 ip = mtod(m, struct ip *); 767 hlen = ip->ip_hl << 2; 768 ip_len = ntohs(ip->ip_len); 769 #endif /* PFIL_HOOKS */ 770 771 m->m_pkthdr.csum_data |= hlen << 16; 772 773 #if IFA_STATS 774 /* 775 * search for the source address structure to 776 * maintain output statistics. 777 */ 778 INADDR_TO_IA(ip->ip_src, ia); 779 #endif 780 781 /* Maybe skip checksums on loopback interfaces. */ 782 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 783 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 784 } 785 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 786 /* 787 * If small enough for mtu of path, or if using TCP segmentation 788 * offload, can just send directly. 789 */ 790 if (ip_len <= mtu || 791 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 792 #if IFA_STATS 793 if (ia) 794 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 795 #endif 796 /* 797 * Always initialize the sum to 0! Some HW assisted 798 * checksumming requires this. 799 */ 800 ip->ip_sum = 0; 801 802 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 803 /* 804 * Perform any checksums that the hardware can't do 805 * for us. 806 * 807 * XXX Does any hardware require the {th,uh}_sum 808 * XXX fields to be 0? 809 */ 810 if (sw_csum & M_CSUM_IPv4) { 811 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 812 ip->ip_sum = in_cksum(m, hlen); 813 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 814 } 815 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 816 if (IN_NEED_CHECKSUM(ifp, 817 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 818 in_delayed_cksum(m); 819 } 820 m->m_pkthdr.csum_flags &= 821 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 822 } 823 } 824 825 #ifdef IPSEC 826 /* clean ipsec history once it goes out of the node */ 827 ipsec_delaux(m); 828 #endif 829 830 if (__predict_true( 831 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 || 832 (ifp->if_capenable & IFCAP_TSOv4) != 0)) { 833 error = 834 (*ifp->if_output)(ifp, m, 835 (m->m_flags & M_MCAST) ? 836 sintocsa(rdst) : sintocsa(dst), 837 rt); 838 } else { 839 error = 840 ip_tso_output(ifp, m, 841 (m->m_flags & M_MCAST) ? 842 sintocsa(rdst) : sintocsa(dst), 843 rt); 844 } 845 goto done; 846 } 847 848 /* 849 * We can't use HW checksumming if we're about to 850 * to fragment the packet. 851 * 852 * XXX Some hardware can do this. 853 */ 854 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 855 if (IN_NEED_CHECKSUM(ifp, 856 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 857 in_delayed_cksum(m); 858 } 859 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 860 } 861 862 /* 863 * Too large for interface; fragment if possible. 864 * Must be able to put at least 8 bytes per fragment. 865 */ 866 if (ntohs(ip->ip_off) & IP_DF) { 867 if (flags & IP_RETURNMTU) 868 *mtu_p = mtu; 869 error = EMSGSIZE; 870 ipstat.ips_cantfrag++; 871 goto bad; 872 } 873 874 error = ip_fragment(m, ifp, mtu); 875 if (error) { 876 m = NULL; 877 goto bad; 878 } 879 880 for (; m; m = m0) { 881 m0 = m->m_nextpkt; 882 m->m_nextpkt = 0; 883 if (error == 0) { 884 #if IFA_STATS 885 if (ia) 886 ia->ia_ifa.ifa_data.ifad_outbytes += 887 ntohs(ip->ip_len); 888 #endif 889 #ifdef IPSEC 890 /* clean ipsec history once it goes out of the node */ 891 ipsec_delaux(m); 892 #endif /* IPSEC */ 893 894 #ifdef IPSEC_NAT_T 895 /* 896 * If we get there, the packet has not been handeld by 897 * IPSec whereas it should have. Now that it has been 898 * fragmented, re-inject it in ip_output so that IPsec 899 * processing can occur. 900 */ 901 if (natt_frag) { 902 error = ip_output(m, opt, 903 ro, flags, imo, so, mtu_p); 904 } else 905 #endif /* IPSEC_NAT_T */ 906 { 907 KASSERT((m->m_pkthdr.csum_flags & 908 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 909 error = (*ifp->if_output)(ifp, m, 910 (m->m_flags & M_MCAST) ? 911 sintocsa(rdst) : sintocsa(dst), 912 rt); 913 } 914 } else 915 m_freem(m); 916 } 917 918 if (error == 0) 919 ipstat.ips_fragmented++; 920 done: 921 rtcache_free(&iproute); 922 923 #ifdef IPSEC 924 if (sp != NULL) { 925 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 926 printf("DP ip_output call free SP:%p\n", sp)); 927 key_freesp(sp); 928 } 929 #endif /* IPSEC */ 930 #ifdef FAST_IPSEC 931 if (sp != NULL) 932 KEY_FREESP(&sp); 933 #endif /* FAST_IPSEC */ 934 935 return (error); 936 bad: 937 m_freem(m); 938 goto done; 939 } 940 941 int 942 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 943 { 944 struct ip *ip, *mhip; 945 struct mbuf *m0; 946 int len, hlen, off; 947 int mhlen, firstlen; 948 struct mbuf **mnext; 949 int sw_csum = m->m_pkthdr.csum_flags; 950 int fragments = 0; 951 int s; 952 int error = 0; 953 954 ip = mtod(m, struct ip *); 955 hlen = ip->ip_hl << 2; 956 if (ifp != NULL) 957 sw_csum &= ~ifp->if_csum_flags_tx; 958 959 len = (mtu - hlen) &~ 7; 960 if (len < 8) { 961 m_freem(m); 962 return (EMSGSIZE); 963 } 964 965 firstlen = len; 966 mnext = &m->m_nextpkt; 967 968 /* 969 * Loop through length of segment after first fragment, 970 * make new header and copy data of each part and link onto chain. 971 */ 972 m0 = m; 973 mhlen = sizeof (struct ip); 974 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 975 MGETHDR(m, M_DONTWAIT, MT_HEADER); 976 if (m == 0) { 977 error = ENOBUFS; 978 ipstat.ips_odropped++; 979 goto sendorfree; 980 } 981 MCLAIM(m, m0->m_owner); 982 *mnext = m; 983 mnext = &m->m_nextpkt; 984 m->m_data += max_linkhdr; 985 mhip = mtod(m, struct ip *); 986 *mhip = *ip; 987 /* we must inherit MCAST and BCAST flags */ 988 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 989 if (hlen > sizeof (struct ip)) { 990 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 991 mhip->ip_hl = mhlen >> 2; 992 } 993 m->m_len = mhlen; 994 mhip->ip_off = ((off - hlen) >> 3) + 995 (ntohs(ip->ip_off) & ~IP_MF); 996 if (ip->ip_off & htons(IP_MF)) 997 mhip->ip_off |= IP_MF; 998 if (off + len >= ntohs(ip->ip_len)) 999 len = ntohs(ip->ip_len) - off; 1000 else 1001 mhip->ip_off |= IP_MF; 1002 HTONS(mhip->ip_off); 1003 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 1004 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 1005 if (m->m_next == 0) { 1006 error = ENOBUFS; /* ??? */ 1007 ipstat.ips_odropped++; 1008 goto sendorfree; 1009 } 1010 m->m_pkthdr.len = mhlen + len; 1011 m->m_pkthdr.rcvif = (struct ifnet *)0; 1012 mhip->ip_sum = 0; 1013 if (sw_csum & M_CSUM_IPv4) { 1014 mhip->ip_sum = in_cksum(m, mhlen); 1015 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 1016 } else { 1017 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 1018 m->m_pkthdr.csum_data |= mhlen << 16; 1019 } 1020 ipstat.ips_ofragments++; 1021 fragments++; 1022 } 1023 /* 1024 * Update first fragment by trimming what's been copied out 1025 * and updating header, then send each fragment (in order). 1026 */ 1027 m = m0; 1028 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 1029 m->m_pkthdr.len = hlen + firstlen; 1030 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 1031 ip->ip_off |= htons(IP_MF); 1032 ip->ip_sum = 0; 1033 if (sw_csum & M_CSUM_IPv4) { 1034 ip->ip_sum = in_cksum(m, hlen); 1035 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 1036 } else { 1037 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 1038 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 1039 sizeof(struct ip)); 1040 } 1041 sendorfree: 1042 /* 1043 * If there is no room for all the fragments, don't queue 1044 * any of them. 1045 */ 1046 if (ifp != NULL) { 1047 s = splnet(); 1048 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 1049 error == 0) { 1050 error = ENOBUFS; 1051 ipstat.ips_odropped++; 1052 IFQ_INC_DROPS(&ifp->if_snd); 1053 } 1054 splx(s); 1055 } 1056 if (error) { 1057 for (m = m0; m; m = m0) { 1058 m0 = m->m_nextpkt; 1059 m->m_nextpkt = NULL; 1060 m_freem(m); 1061 } 1062 } 1063 return (error); 1064 } 1065 1066 /* 1067 * Process a delayed payload checksum calculation. 1068 */ 1069 void 1070 in_delayed_cksum(struct mbuf *m) 1071 { 1072 struct ip *ip; 1073 u_int16_t csum, offset; 1074 1075 ip = mtod(m, struct ip *); 1076 offset = ip->ip_hl << 2; 1077 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 1078 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 1079 csum = 0xffff; 1080 1081 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 1082 1083 if ((offset + sizeof(u_int16_t)) > m->m_len) { 1084 /* This happen when ip options were inserted 1085 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 1086 m->m_len, offset, ip->ip_p); 1087 */ 1088 m_copyback(m, offset, sizeof(csum), (void *) &csum); 1089 } else 1090 *(u_int16_t *)(mtod(m, char *) + offset) = csum; 1091 } 1092 1093 /* 1094 * Determine the maximum length of the options to be inserted; 1095 * we would far rather allocate too much space rather than too little. 1096 */ 1097 1098 u_int 1099 ip_optlen(struct inpcb *inp) 1100 { 1101 struct mbuf *m = inp->inp_options; 1102 1103 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 1104 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 1105 else 1106 return 0; 1107 } 1108 1109 1110 /* 1111 * Insert IP options into preformed packet. 1112 * Adjust IP destination as required for IP source routing, 1113 * as indicated by a non-zero in_addr at the start of the options. 1114 */ 1115 static struct mbuf * 1116 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 1117 { 1118 struct ipoption *p = mtod(opt, struct ipoption *); 1119 struct mbuf *n; 1120 struct ip *ip = mtod(m, struct ip *); 1121 unsigned optlen; 1122 1123 optlen = opt->m_len - sizeof(p->ipopt_dst); 1124 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 1125 return (m); /* XXX should fail */ 1126 if (!in_nullhost(p->ipopt_dst)) 1127 ip->ip_dst = p->ipopt_dst; 1128 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 1129 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1130 if (n == 0) 1131 return (m); 1132 MCLAIM(n, m->m_owner); 1133 M_MOVE_PKTHDR(n, m); 1134 m->m_len -= sizeof(struct ip); 1135 m->m_data += sizeof(struct ip); 1136 n->m_next = m; 1137 m = n; 1138 m->m_len = optlen + sizeof(struct ip); 1139 m->m_data += max_linkhdr; 1140 bcopy((void *)ip, mtod(m, void *), sizeof(struct ip)); 1141 } else { 1142 m->m_data -= optlen; 1143 m->m_len += optlen; 1144 memmove(mtod(m, void *), ip, sizeof(struct ip)); 1145 } 1146 m->m_pkthdr.len += optlen; 1147 ip = mtod(m, struct ip *); 1148 bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen); 1149 *phlen = sizeof(struct ip) + optlen; 1150 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1151 return (m); 1152 } 1153 1154 /* 1155 * Copy options from ip to jp, 1156 * omitting those not copied during fragmentation. 1157 */ 1158 int 1159 ip_optcopy(struct ip *ip, struct ip *jp) 1160 { 1161 u_char *cp, *dp; 1162 int opt, optlen, cnt; 1163 1164 cp = (u_char *)(ip + 1); 1165 dp = (u_char *)(jp + 1); 1166 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1167 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1168 opt = cp[0]; 1169 if (opt == IPOPT_EOL) 1170 break; 1171 if (opt == IPOPT_NOP) { 1172 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1173 *dp++ = IPOPT_NOP; 1174 optlen = 1; 1175 continue; 1176 } 1177 #ifdef DIAGNOSTIC 1178 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1179 panic("malformed IPv4 option passed to ip_optcopy"); 1180 #endif 1181 optlen = cp[IPOPT_OLEN]; 1182 #ifdef DIAGNOSTIC 1183 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1184 panic("malformed IPv4 option passed to ip_optcopy"); 1185 #endif 1186 /* bogus lengths should have been caught by ip_dooptions */ 1187 if (optlen > cnt) 1188 optlen = cnt; 1189 if (IPOPT_COPIED(opt)) { 1190 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 1191 dp += optlen; 1192 } 1193 } 1194 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1195 *dp++ = IPOPT_EOL; 1196 return (optlen); 1197 } 1198 1199 /* 1200 * IP socket option processing. 1201 */ 1202 int 1203 ip_ctloutput(int op, struct socket *so, int level, int optname, 1204 struct mbuf **mp) 1205 { 1206 struct inpcb *inp = sotoinpcb(so); 1207 struct mbuf *m = *mp; 1208 int optval = 0; 1209 int error = 0; 1210 #if defined(IPSEC) || defined(FAST_IPSEC) 1211 struct lwp *l = curlwp; /*XXX*/ 1212 #endif 1213 1214 if (level != IPPROTO_IP) { 1215 if (op == PRCO_SETOPT && *mp) 1216 (void) m_free(*mp); 1217 if (level == SOL_SOCKET && optname == SO_NOHEADER) 1218 return 0; 1219 return ENOPROTOOPT; 1220 } 1221 1222 switch (op) { 1223 1224 case PRCO_SETOPT: 1225 switch (optname) { 1226 case IP_OPTIONS: 1227 #ifdef notyet 1228 case IP_RETOPTS: 1229 return (ip_pcbopts(optname, &inp->inp_options, m)); 1230 #else 1231 return (ip_pcbopts(&inp->inp_options, m)); 1232 #endif 1233 1234 case IP_TOS: 1235 case IP_TTL: 1236 case IP_RECVOPTS: 1237 case IP_RECVRETOPTS: 1238 case IP_RECVDSTADDR: 1239 case IP_RECVIF: 1240 if (m == NULL || m->m_len != sizeof(int)) 1241 error = EINVAL; 1242 else { 1243 optval = *mtod(m, int *); 1244 switch (optname) { 1245 1246 case IP_TOS: 1247 inp->inp_ip.ip_tos = optval; 1248 break; 1249 1250 case IP_TTL: 1251 inp->inp_ip.ip_ttl = optval; 1252 break; 1253 #define OPTSET(bit) \ 1254 if (optval) \ 1255 inp->inp_flags |= bit; \ 1256 else \ 1257 inp->inp_flags &= ~bit; 1258 1259 case IP_RECVOPTS: 1260 OPTSET(INP_RECVOPTS); 1261 break; 1262 1263 case IP_RECVRETOPTS: 1264 OPTSET(INP_RECVRETOPTS); 1265 break; 1266 1267 case IP_RECVDSTADDR: 1268 OPTSET(INP_RECVDSTADDR); 1269 break; 1270 1271 case IP_RECVIF: 1272 OPTSET(INP_RECVIF); 1273 break; 1274 } 1275 } 1276 break; 1277 #undef OPTSET 1278 1279 case IP_MULTICAST_IF: 1280 case IP_MULTICAST_TTL: 1281 case IP_MULTICAST_LOOP: 1282 case IP_ADD_MEMBERSHIP: 1283 case IP_DROP_MEMBERSHIP: 1284 error = ip_setmoptions(optname, &inp->inp_moptions, m); 1285 break; 1286 1287 case IP_PORTRANGE: 1288 if (m == 0 || m->m_len != sizeof(int)) 1289 error = EINVAL; 1290 else { 1291 optval = *mtod(m, int *); 1292 1293 switch (optval) { 1294 1295 case IP_PORTRANGE_DEFAULT: 1296 case IP_PORTRANGE_HIGH: 1297 inp->inp_flags &= ~(INP_LOWPORT); 1298 break; 1299 1300 case IP_PORTRANGE_LOW: 1301 inp->inp_flags |= INP_LOWPORT; 1302 break; 1303 1304 default: 1305 error = EINVAL; 1306 break; 1307 } 1308 } 1309 break; 1310 1311 #if defined(IPSEC) || defined(FAST_IPSEC) 1312 case IP_IPSEC_POLICY: 1313 { 1314 void *req = NULL; 1315 size_t len = 0; 1316 int priv = 0; 1317 1318 #ifdef __NetBSD__ 1319 if (l == 0 || kauth_authorize_generic(l->l_cred, 1320 KAUTH_GENERIC_ISSUSER, NULL)) 1321 priv = 0; 1322 else 1323 priv = 1; 1324 #else 1325 priv = (in6p->in6p_socket->so_state & SS_PRIV); 1326 #endif 1327 if (m) { 1328 req = mtod(m, void *); 1329 len = m->m_len; 1330 } 1331 error = ipsec4_set_policy(inp, optname, req, len, priv); 1332 break; 1333 } 1334 #endif /*IPSEC*/ 1335 1336 default: 1337 error = ENOPROTOOPT; 1338 break; 1339 } 1340 if (m) 1341 (void)m_free(m); 1342 break; 1343 1344 case PRCO_GETOPT: 1345 switch (optname) { 1346 case IP_OPTIONS: 1347 case IP_RETOPTS: 1348 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1349 MCLAIM(m, so->so_mowner); 1350 if (inp->inp_options) { 1351 m->m_len = inp->inp_options->m_len; 1352 bcopy(mtod(inp->inp_options, void *), 1353 mtod(m, void *), (unsigned)m->m_len); 1354 } else 1355 m->m_len = 0; 1356 break; 1357 1358 case IP_TOS: 1359 case IP_TTL: 1360 case IP_RECVOPTS: 1361 case IP_RECVRETOPTS: 1362 case IP_RECVDSTADDR: 1363 case IP_RECVIF: 1364 case IP_ERRORMTU: 1365 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1366 MCLAIM(m, so->so_mowner); 1367 m->m_len = sizeof(int); 1368 switch (optname) { 1369 1370 case IP_TOS: 1371 optval = inp->inp_ip.ip_tos; 1372 break; 1373 1374 case IP_TTL: 1375 optval = inp->inp_ip.ip_ttl; 1376 break; 1377 1378 case IP_ERRORMTU: 1379 optval = inp->inp_errormtu; 1380 break; 1381 1382 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1383 1384 case IP_RECVOPTS: 1385 optval = OPTBIT(INP_RECVOPTS); 1386 break; 1387 1388 case IP_RECVRETOPTS: 1389 optval = OPTBIT(INP_RECVRETOPTS); 1390 break; 1391 1392 case IP_RECVDSTADDR: 1393 optval = OPTBIT(INP_RECVDSTADDR); 1394 break; 1395 1396 case IP_RECVIF: 1397 optval = OPTBIT(INP_RECVIF); 1398 break; 1399 } 1400 *mtod(m, int *) = optval; 1401 break; 1402 1403 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */ 1404 /* XXX: code broken */ 1405 case IP_IPSEC_POLICY: 1406 { 1407 void *req = NULL; 1408 size_t len = 0; 1409 1410 if (m) { 1411 req = mtod(m, void *); 1412 len = m->m_len; 1413 } 1414 error = ipsec4_get_policy(inp, req, len, mp); 1415 break; 1416 } 1417 #endif /*IPSEC*/ 1418 1419 case IP_MULTICAST_IF: 1420 case IP_MULTICAST_TTL: 1421 case IP_MULTICAST_LOOP: 1422 case IP_ADD_MEMBERSHIP: 1423 case IP_DROP_MEMBERSHIP: 1424 error = ip_getmoptions(optname, inp->inp_moptions, mp); 1425 if (*mp) 1426 MCLAIM(*mp, so->so_mowner); 1427 break; 1428 1429 case IP_PORTRANGE: 1430 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1431 MCLAIM(m, so->so_mowner); 1432 m->m_len = sizeof(int); 1433 1434 if (inp->inp_flags & INP_LOWPORT) 1435 optval = IP_PORTRANGE_LOW; 1436 else 1437 optval = IP_PORTRANGE_DEFAULT; 1438 1439 *mtod(m, int *) = optval; 1440 break; 1441 1442 default: 1443 error = ENOPROTOOPT; 1444 break; 1445 } 1446 break; 1447 } 1448 return (error); 1449 } 1450 1451 /* 1452 * Set up IP options in pcb for insertion in output packets. 1453 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1454 * with destination address if source routed. 1455 */ 1456 int 1457 #ifdef notyet 1458 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m) 1459 #else 1460 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m) 1461 #endif 1462 { 1463 int cnt, optlen; 1464 u_char *cp; 1465 u_char opt; 1466 1467 /* turn off any old options */ 1468 if (*pcbopt) 1469 (void)m_free(*pcbopt); 1470 *pcbopt = 0; 1471 if (m == (struct mbuf *)0 || m->m_len == 0) { 1472 /* 1473 * Only turning off any previous options. 1474 */ 1475 if (m) 1476 (void)m_free(m); 1477 return (0); 1478 } 1479 1480 #ifndef __vax__ 1481 if (m->m_len % sizeof(int32_t)) 1482 goto bad; 1483 #endif 1484 /* 1485 * IP first-hop destination address will be stored before 1486 * actual options; move other options back 1487 * and clear it when none present. 1488 */ 1489 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1490 goto bad; 1491 cnt = m->m_len; 1492 m->m_len += sizeof(struct in_addr); 1493 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1494 memmove(cp, mtod(m, void *), (unsigned)cnt); 1495 bzero(mtod(m, void *), sizeof(struct in_addr)); 1496 1497 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1498 opt = cp[IPOPT_OPTVAL]; 1499 if (opt == IPOPT_EOL) 1500 break; 1501 if (opt == IPOPT_NOP) 1502 optlen = 1; 1503 else { 1504 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1505 goto bad; 1506 optlen = cp[IPOPT_OLEN]; 1507 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1508 goto bad; 1509 } 1510 switch (opt) { 1511 1512 default: 1513 break; 1514 1515 case IPOPT_LSRR: 1516 case IPOPT_SSRR: 1517 /* 1518 * user process specifies route as: 1519 * ->A->B->C->D 1520 * D must be our final destination (but we can't 1521 * check that since we may not have connected yet). 1522 * A is first hop destination, which doesn't appear in 1523 * actual IP option, but is stored before the options. 1524 */ 1525 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1526 goto bad; 1527 m->m_len -= sizeof(struct in_addr); 1528 cnt -= sizeof(struct in_addr); 1529 optlen -= sizeof(struct in_addr); 1530 cp[IPOPT_OLEN] = optlen; 1531 /* 1532 * Move first hop before start of options. 1533 */ 1534 bcopy((void *)&cp[IPOPT_OFFSET+1], mtod(m, void *), 1535 sizeof(struct in_addr)); 1536 /* 1537 * Then copy rest of options back 1538 * to close up the deleted entry. 1539 */ 1540 (void)memmove(&cp[IPOPT_OFFSET+1], 1541 &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr), 1542 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1543 break; 1544 } 1545 } 1546 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1547 goto bad; 1548 *pcbopt = m; 1549 return (0); 1550 1551 bad: 1552 (void)m_free(m); 1553 return (EINVAL); 1554 } 1555 1556 /* 1557 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1558 */ 1559 static struct ifnet * 1560 ip_multicast_if(struct in_addr *a, int *ifindexp) 1561 { 1562 int ifindex; 1563 struct ifnet *ifp = NULL; 1564 struct in_ifaddr *ia; 1565 1566 if (ifindexp) 1567 *ifindexp = 0; 1568 if (ntohl(a->s_addr) >> 24 == 0) { 1569 ifindex = ntohl(a->s_addr) & 0xffffff; 1570 if (ifindex < 0 || if_indexlim <= ifindex) 1571 return NULL; 1572 ifp = ifindex2ifnet[ifindex]; 1573 if (!ifp) 1574 return NULL; 1575 if (ifindexp) 1576 *ifindexp = ifindex; 1577 } else { 1578 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1579 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1580 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1581 ifp = ia->ia_ifp; 1582 break; 1583 } 1584 } 1585 } 1586 return ifp; 1587 } 1588 1589 static int 1590 ip_getoptval(struct mbuf *m, u_int8_t *val, u_int maxval) 1591 { 1592 u_int tval; 1593 1594 if (m == NULL) 1595 return EINVAL; 1596 1597 switch (m->m_len) { 1598 case sizeof(u_char): 1599 tval = *(mtod(m, u_char *)); 1600 break; 1601 case sizeof(u_int): 1602 tval = *(mtod(m, u_int *)); 1603 break; 1604 default: 1605 return EINVAL; 1606 } 1607 1608 if (tval > maxval) 1609 return EINVAL; 1610 1611 *val = tval; 1612 return 0; 1613 } 1614 1615 /* 1616 * Set the IP multicast options in response to user setsockopt(). 1617 */ 1618 int 1619 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m) 1620 { 1621 int error = 0; 1622 int i; 1623 struct in_addr addr; 1624 struct ip_mreq *mreq; 1625 struct ifnet *ifp; 1626 struct ip_moptions *imo = *imop; 1627 int ifindex; 1628 1629 if (imo == NULL) { 1630 /* 1631 * No multicast option buffer attached to the pcb; 1632 * allocate one and initialize to default values. 1633 */ 1634 imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS, 1635 M_WAITOK); 1636 1637 if (imo == NULL) 1638 return (ENOBUFS); 1639 *imop = imo; 1640 imo->imo_multicast_ifp = NULL; 1641 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1642 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1643 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1644 imo->imo_num_memberships = 0; 1645 } 1646 1647 switch (optname) { 1648 1649 case IP_MULTICAST_IF: 1650 /* 1651 * Select the interface for outgoing multicast packets. 1652 */ 1653 if (m == NULL || m->m_len != sizeof(struct in_addr)) { 1654 error = EINVAL; 1655 break; 1656 } 1657 addr = *(mtod(m, struct in_addr *)); 1658 /* 1659 * INADDR_ANY is used to remove a previous selection. 1660 * When no interface is selected, a default one is 1661 * chosen every time a multicast packet is sent. 1662 */ 1663 if (in_nullhost(addr)) { 1664 imo->imo_multicast_ifp = NULL; 1665 break; 1666 } 1667 /* 1668 * The selected interface is identified by its local 1669 * IP address. Find the interface and confirm that 1670 * it supports multicasting. 1671 */ 1672 ifp = ip_multicast_if(&addr, &ifindex); 1673 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1674 error = EADDRNOTAVAIL; 1675 break; 1676 } 1677 imo->imo_multicast_ifp = ifp; 1678 if (ifindex) 1679 imo->imo_multicast_addr = addr; 1680 else 1681 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1682 break; 1683 1684 case IP_MULTICAST_TTL: 1685 /* 1686 * Set the IP time-to-live for outgoing multicast packets. 1687 */ 1688 error = ip_getoptval(m, &imo->imo_multicast_ttl, MAXTTL); 1689 break; 1690 1691 case IP_MULTICAST_LOOP: 1692 /* 1693 * Set the loopback flag for outgoing multicast packets. 1694 * Must be zero or one. 1695 */ 1696 error = ip_getoptval(m, &imo->imo_multicast_loop, 1); 1697 break; 1698 1699 case IP_ADD_MEMBERSHIP: 1700 /* 1701 * Add a multicast group membership. 1702 * Group must be a valid IP multicast address. 1703 */ 1704 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1705 error = EINVAL; 1706 break; 1707 } 1708 mreq = mtod(m, struct ip_mreq *); 1709 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1710 error = EINVAL; 1711 break; 1712 } 1713 /* 1714 * If no interface address was provided, use the interface of 1715 * the route to the given multicast address. 1716 */ 1717 if (in_nullhost(mreq->imr_interface)) { 1718 struct rtentry *rt; 1719 union { 1720 struct sockaddr dst; 1721 struct sockaddr_in dst4; 1722 } u; 1723 struct route ro; 1724 1725 memset(&ro, 0, sizeof(ro)); 1726 1727 sockaddr_in_init(&u.dst4, &mreq->imr_multiaddr, 0); 1728 rtcache_setdst(&ro, &u.dst); 1729 ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp 1730 : NULL; 1731 rtcache_free(&ro); 1732 } else { 1733 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1734 } 1735 /* 1736 * See if we found an interface, and confirm that it 1737 * supports multicast. 1738 */ 1739 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1740 error = EADDRNOTAVAIL; 1741 break; 1742 } 1743 /* 1744 * See if the membership already exists or if all the 1745 * membership slots are full. 1746 */ 1747 for (i = 0; i < imo->imo_num_memberships; ++i) { 1748 if (imo->imo_membership[i]->inm_ifp == ifp && 1749 in_hosteq(imo->imo_membership[i]->inm_addr, 1750 mreq->imr_multiaddr)) 1751 break; 1752 } 1753 if (i < imo->imo_num_memberships) { 1754 error = EADDRINUSE; 1755 break; 1756 } 1757 if (i == IP_MAX_MEMBERSHIPS) { 1758 error = ETOOMANYREFS; 1759 break; 1760 } 1761 /* 1762 * Everything looks good; add a new record to the multicast 1763 * address list for the given interface. 1764 */ 1765 if ((imo->imo_membership[i] = 1766 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1767 error = ENOBUFS; 1768 break; 1769 } 1770 ++imo->imo_num_memberships; 1771 break; 1772 1773 case IP_DROP_MEMBERSHIP: 1774 /* 1775 * Drop a multicast group membership. 1776 * Group must be a valid IP multicast address. 1777 */ 1778 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1779 error = EINVAL; 1780 break; 1781 } 1782 mreq = mtod(m, struct ip_mreq *); 1783 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1784 error = EINVAL; 1785 break; 1786 } 1787 /* 1788 * If an interface address was specified, get a pointer 1789 * to its ifnet structure. 1790 */ 1791 if (in_nullhost(mreq->imr_interface)) 1792 ifp = NULL; 1793 else { 1794 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1795 if (ifp == NULL) { 1796 error = EADDRNOTAVAIL; 1797 break; 1798 } 1799 } 1800 /* 1801 * Find the membership in the membership array. 1802 */ 1803 for (i = 0; i < imo->imo_num_memberships; ++i) { 1804 if ((ifp == NULL || 1805 imo->imo_membership[i]->inm_ifp == ifp) && 1806 in_hosteq(imo->imo_membership[i]->inm_addr, 1807 mreq->imr_multiaddr)) 1808 break; 1809 } 1810 if (i == imo->imo_num_memberships) { 1811 error = EADDRNOTAVAIL; 1812 break; 1813 } 1814 /* 1815 * Give up the multicast address record to which the 1816 * membership points. 1817 */ 1818 in_delmulti(imo->imo_membership[i]); 1819 /* 1820 * Remove the gap in the membership array. 1821 */ 1822 for (++i; i < imo->imo_num_memberships; ++i) 1823 imo->imo_membership[i-1] = imo->imo_membership[i]; 1824 --imo->imo_num_memberships; 1825 break; 1826 1827 default: 1828 error = EOPNOTSUPP; 1829 break; 1830 } 1831 1832 /* 1833 * If all options have default values, no need to keep the mbuf. 1834 */ 1835 if (imo->imo_multicast_ifp == NULL && 1836 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1837 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1838 imo->imo_num_memberships == 0) { 1839 free(*imop, M_IPMOPTS); 1840 *imop = NULL; 1841 } 1842 1843 return (error); 1844 } 1845 1846 /* 1847 * Return the IP multicast options in response to user getsockopt(). 1848 */ 1849 int 1850 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp) 1851 { 1852 u_char *ttl; 1853 u_char *loop; 1854 struct in_addr *addr; 1855 struct in_ifaddr *ia; 1856 1857 *mp = m_get(M_WAIT, MT_SOOPTS); 1858 1859 switch (optname) { 1860 1861 case IP_MULTICAST_IF: 1862 addr = mtod(*mp, struct in_addr *); 1863 (*mp)->m_len = sizeof(struct in_addr); 1864 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1865 *addr = zeroin_addr; 1866 else if (imo->imo_multicast_addr.s_addr) { 1867 /* return the value user has set */ 1868 *addr = imo->imo_multicast_addr; 1869 } else { 1870 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1871 *addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1872 } 1873 return (0); 1874 1875 case IP_MULTICAST_TTL: 1876 ttl = mtod(*mp, u_char *); 1877 (*mp)->m_len = 1; 1878 *ttl = imo ? imo->imo_multicast_ttl 1879 : IP_DEFAULT_MULTICAST_TTL; 1880 return (0); 1881 1882 case IP_MULTICAST_LOOP: 1883 loop = mtod(*mp, u_char *); 1884 (*mp)->m_len = 1; 1885 *loop = imo ? imo->imo_multicast_loop 1886 : IP_DEFAULT_MULTICAST_LOOP; 1887 return (0); 1888 1889 default: 1890 return (EOPNOTSUPP); 1891 } 1892 } 1893 1894 /* 1895 * Discard the IP multicast options. 1896 */ 1897 void 1898 ip_freemoptions(struct ip_moptions *imo) 1899 { 1900 int i; 1901 1902 if (imo != NULL) { 1903 for (i = 0; i < imo->imo_num_memberships; ++i) 1904 in_delmulti(imo->imo_membership[i]); 1905 free(imo, M_IPMOPTS); 1906 } 1907 } 1908 1909 /* 1910 * Routine called from ip_output() to loop back a copy of an IP multicast 1911 * packet to the input queue of a specified interface. Note that this 1912 * calls the output routine of the loopback "driver", but with an interface 1913 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1914 */ 1915 static void 1916 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 1917 { 1918 struct ip *ip; 1919 struct mbuf *copym; 1920 1921 copym = m_copypacket(m, M_DONTWAIT); 1922 if (copym != NULL 1923 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1924 copym = m_pullup(copym, sizeof(struct ip)); 1925 if (copym == NULL) 1926 return; 1927 /* 1928 * We don't bother to fragment if the IP length is greater 1929 * than the interface's MTU. Can this possibly matter? 1930 */ 1931 ip = mtod(copym, struct ip *); 1932 1933 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1934 in_delayed_cksum(copym); 1935 copym->m_pkthdr.csum_flags &= 1936 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1937 } 1938 1939 ip->ip_sum = 0; 1940 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1941 (void)looutput(ifp, copym, sintocsa(dst), NULL); 1942 } 1943