1 /* $NetBSD: ip_output.c,v 1.153 2005/05/29 21:41:23 christos 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 1.153 2005/05/29 21:41:23 christos Exp $"); 102 103 #include "opt_pfil_hooks.h" 104 #include "opt_inet.h" 105 #include "opt_ipsec.h" 106 #include "opt_mrouting.h" 107 108 #include <sys/param.h> 109 #include <sys/malloc.h> 110 #include <sys/mbuf.h> 111 #include <sys/errno.h> 112 #include <sys/protosw.h> 113 #include <sys/socket.h> 114 #include <sys/socketvar.h> 115 #ifdef FAST_IPSEC 116 #include <sys/domain.h> 117 #endif 118 #include <sys/systm.h> 119 #include <sys/proc.h> 120 121 #include <net/if.h> 122 #include <net/route.h> 123 #include <net/pfil.h> 124 125 #include <netinet/in.h> 126 #include <netinet/in_systm.h> 127 #include <netinet/ip.h> 128 #include <netinet/in_pcb.h> 129 #include <netinet/in_var.h> 130 #include <netinet/ip_var.h> 131 #include <netinet/in_offload.h> 132 133 #ifdef MROUTING 134 #include <netinet/ip_mroute.h> 135 #endif 136 137 #include <machine/stdarg.h> 138 139 #ifdef IPSEC 140 #include <netinet6/ipsec.h> 141 #include <netkey/key.h> 142 #include <netkey/key_debug.h> 143 #ifdef IPSEC_NAT_T 144 #include <netinet/udp.h> 145 #endif 146 #endif /*IPSEC*/ 147 148 #ifdef FAST_IPSEC 149 #include <netipsec/ipsec.h> 150 #include <netipsec/key.h> 151 #include <netipsec/xform.h> 152 #endif /* FAST_IPSEC*/ 153 154 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 155 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 156 static void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *); 157 158 #ifdef PFIL_HOOKS 159 extern struct pfil_head inet_pfil_hook; /* XXX */ 160 #endif 161 162 int udp_do_loopback_cksum = 0; 163 int tcp_do_loopback_cksum = 0; 164 int ip_do_loopback_cksum = 0; 165 166 #define IN_NEED_CHECKSUM(ifp, csum_flags) \ 167 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 168 (((csum_flags) & M_CSUM_UDPv4) != 0 && udp_do_loopback_cksum) || \ 169 (((csum_flags) & M_CSUM_TCPv4) != 0 && tcp_do_loopback_cksum) || \ 170 (((csum_flags) & M_CSUM_IPv4) != 0 && ip_do_loopback_cksum))) 171 172 struct ip_tso_output_args { 173 struct ifnet *ifp; 174 struct sockaddr *sa; 175 struct rtentry *rt; 176 }; 177 178 static int ip_tso_output_callback(void *, struct mbuf *); 179 static int ip_tso_output(struct ifnet *, struct mbuf *, struct sockaddr *, 180 struct rtentry *); 181 182 static int 183 ip_tso_output_callback(void *vp, struct mbuf *m) 184 { 185 struct ip_tso_output_args *args = vp; 186 struct ifnet *ifp = args->ifp; 187 188 return (*ifp->if_output)(ifp, m, args->sa, args->rt); 189 } 190 191 static int 192 ip_tso_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa, 193 struct rtentry *rt) 194 { 195 struct ip_tso_output_args args; 196 197 args.ifp = ifp; 198 args.sa = sa; 199 args.rt = rt; 200 201 return tcp4_segment(m, ip_tso_output_callback, &args); 202 } 203 204 /* 205 * IP output. The packet in mbuf chain m contains a skeletal IP 206 * header (with len, off, ttl, proto, tos, src, dst). 207 * The mbuf chain containing the packet will be freed. 208 * The mbuf opt, if present, will not be freed. 209 */ 210 int 211 ip_output(struct mbuf *m0, ...) 212 { 213 struct ip *ip; 214 struct ifnet *ifp; 215 struct mbuf *m = m0; 216 int hlen = sizeof (struct ip); 217 int len, error = 0; 218 struct route iproute; 219 struct sockaddr_in *dst; 220 struct in_ifaddr *ia; 221 struct mbuf *opt; 222 struct route *ro; 223 int flags, sw_csum; 224 int *mtu_p; 225 u_long mtu; 226 struct ip_moptions *imo; 227 struct socket *so; 228 va_list ap; 229 #ifdef IPSEC 230 struct secpolicy *sp = NULL; 231 #ifdef IPSEC_NAT_T 232 int natt_frag = 0; 233 #endif 234 #endif /*IPSEC*/ 235 #ifdef FAST_IPSEC 236 struct inpcb *inp; 237 struct m_tag *mtag; 238 struct secpolicy *sp = NULL; 239 struct tdb_ident *tdbi; 240 int s; 241 #endif 242 u_int16_t ip_len; 243 244 len = 0; 245 va_start(ap, m0); 246 opt = va_arg(ap, struct mbuf *); 247 ro = va_arg(ap, struct route *); 248 flags = va_arg(ap, int); 249 imo = va_arg(ap, struct ip_moptions *); 250 so = va_arg(ap, struct socket *); 251 if (flags & IP_RETURNMTU) 252 mtu_p = va_arg(ap, int *); 253 else 254 mtu_p = NULL; 255 va_end(ap); 256 257 MCLAIM(m, &ip_tx_mowner); 258 #ifdef FAST_IPSEC 259 if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET) 260 inp = (struct inpcb *)so->so_pcb; 261 else 262 inp = NULL; 263 #endif /* FAST_IPSEC */ 264 265 #ifdef DIAGNOSTIC 266 if ((m->m_flags & M_PKTHDR) == 0) 267 panic("ip_output no HDR"); 268 #endif 269 if (opt) { 270 m = ip_insertoptions(m, opt, &len); 271 if (len >= sizeof(struct ip)) 272 hlen = len; 273 } 274 ip = mtod(m, struct ip *); 275 /* 276 * Fill in IP header. 277 */ 278 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 279 ip->ip_v = IPVERSION; 280 ip->ip_off = htons(0); 281 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 282 ip->ip_id = ip_newid(); 283 } else { 284 285 /* 286 * TSO capable interfaces (typically?) increment 287 * ip_id for each segment. 288 * "allocate" enough ids here to increase the chance 289 * for them to be unique. 290 * 291 * note that the following calculation is not 292 * needed to be precise. wasting some ip_id is fine. 293 */ 294 295 unsigned int segsz = m->m_pkthdr.segsz; 296 unsigned int datasz = ntohs(ip->ip_len) - hlen; 297 unsigned int num = howmany(datasz, segsz); 298 299 ip->ip_id = ip_newid_range(num); 300 } 301 ip->ip_hl = hlen >> 2; 302 ipstat.ips_localout++; 303 } else { 304 hlen = ip->ip_hl << 2; 305 } 306 /* 307 * Route packet. 308 */ 309 if (ro == 0) { 310 ro = &iproute; 311 bzero((caddr_t)ro, sizeof (*ro)); 312 } 313 dst = satosin(&ro->ro_dst); 314 /* 315 * If there is a cached route, 316 * check that it is to the same destination 317 * and is still up. If not, free it and try again. 318 * The address family should also be checked in case of sharing the 319 * cache with IPv6. 320 */ 321 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 322 dst->sin_family != AF_INET || 323 !in_hosteq(dst->sin_addr, ip->ip_dst))) { 324 RTFREE(ro->ro_rt); 325 ro->ro_rt = (struct rtentry *)0; 326 } 327 if (ro->ro_rt == 0) { 328 bzero(dst, sizeof(*dst)); 329 dst->sin_family = AF_INET; 330 dst->sin_len = sizeof(*dst); 331 dst->sin_addr = ip->ip_dst; 332 } 333 /* 334 * If routing to interface only, 335 * short circuit routing lookup. 336 */ 337 if (flags & IP_ROUTETOIF) { 338 if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) { 339 ipstat.ips_noroute++; 340 error = ENETUNREACH; 341 goto bad; 342 } 343 ifp = ia->ia_ifp; 344 mtu = ifp->if_mtu; 345 ip->ip_ttl = 1; 346 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 347 ip->ip_dst.s_addr == INADDR_BROADCAST) && 348 imo != NULL && imo->imo_multicast_ifp != NULL) { 349 ifp = imo->imo_multicast_ifp; 350 mtu = ifp->if_mtu; 351 IFP_TO_IA(ifp, ia); 352 } else { 353 if (ro->ro_rt == 0) 354 rtalloc(ro); 355 if (ro->ro_rt == 0) { 356 ipstat.ips_noroute++; 357 error = EHOSTUNREACH; 358 goto bad; 359 } 360 ia = ifatoia(ro->ro_rt->rt_ifa); 361 ifp = ro->ro_rt->rt_ifp; 362 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 363 mtu = ifp->if_mtu; 364 ro->ro_rt->rt_use++; 365 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 366 dst = satosin(ro->ro_rt->rt_gateway); 367 } 368 if (IN_MULTICAST(ip->ip_dst.s_addr) || 369 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 370 struct in_multi *inm; 371 372 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 373 M_BCAST : M_MCAST; 374 /* 375 * IP destination address is multicast. Make sure "dst" 376 * still points to the address in "ro". (It may have been 377 * changed to point to a gateway address, above.) 378 */ 379 dst = satosin(&ro->ro_dst); 380 /* 381 * See if the caller provided any multicast options 382 */ 383 if (imo != NULL) 384 ip->ip_ttl = imo->imo_multicast_ttl; 385 else 386 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 387 388 /* 389 * if we don't know the outgoing ifp yet, we can't generate 390 * output 391 */ 392 if (!ifp) { 393 ipstat.ips_noroute++; 394 error = ENETUNREACH; 395 goto bad; 396 } 397 398 /* 399 * If the packet is multicast or broadcast, confirm that 400 * the outgoing interface can transmit it. 401 */ 402 if (((m->m_flags & M_MCAST) && 403 (ifp->if_flags & IFF_MULTICAST) == 0) || 404 ((m->m_flags & M_BCAST) && 405 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 406 ipstat.ips_noroute++; 407 error = ENETUNREACH; 408 goto bad; 409 } 410 /* 411 * If source address not specified yet, use an address 412 * of outgoing interface. 413 */ 414 if (in_nullhost(ip->ip_src)) { 415 struct in_ifaddr *xia; 416 417 IFP_TO_IA(ifp, xia); 418 if (!xia) { 419 error = EADDRNOTAVAIL; 420 goto bad; 421 } 422 ip->ip_src = xia->ia_addr.sin_addr; 423 } 424 425 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 426 if (inm != NULL && 427 (imo == NULL || imo->imo_multicast_loop)) { 428 /* 429 * If we belong to the destination multicast group 430 * on the outgoing interface, and the caller did not 431 * forbid loopback, loop back a copy. 432 */ 433 ip_mloopback(ifp, m, dst); 434 } 435 #ifdef MROUTING 436 else { 437 /* 438 * If we are acting as a multicast router, perform 439 * multicast forwarding as if the packet had just 440 * arrived on the interface to which we are about 441 * to send. The multicast forwarding function 442 * recursively calls this function, using the 443 * IP_FORWARDING flag to prevent infinite recursion. 444 * 445 * Multicasts that are looped back by ip_mloopback(), 446 * above, will be forwarded by the ip_input() routine, 447 * if necessary. 448 */ 449 extern struct socket *ip_mrouter; 450 451 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 452 if (ip_mforward(m, ifp) != 0) { 453 m_freem(m); 454 goto done; 455 } 456 } 457 } 458 #endif 459 /* 460 * Multicasts with a time-to-live of zero may be looped- 461 * back, above, but must not be transmitted on a network. 462 * Also, multicasts addressed to the loopback interface 463 * are not sent -- the above call to ip_mloopback() will 464 * loop back a copy if this host actually belongs to the 465 * destination group on the loopback interface. 466 */ 467 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 468 m_freem(m); 469 goto done; 470 } 471 472 goto sendit; 473 } 474 #ifndef notdef 475 /* 476 * If source address not specified yet, use address 477 * of outgoing interface. 478 */ 479 if (in_nullhost(ip->ip_src)) 480 ip->ip_src = ia->ia_addr.sin_addr; 481 #endif 482 483 /* 484 * packets with Class-D address as source are not valid per 485 * RFC 1112 486 */ 487 if (IN_MULTICAST(ip->ip_src.s_addr)) { 488 ipstat.ips_odropped++; 489 error = EADDRNOTAVAIL; 490 goto bad; 491 } 492 493 /* 494 * Look for broadcast address and 495 * and verify user is allowed to send 496 * such a packet. 497 */ 498 if (in_broadcast(dst->sin_addr, ifp)) { 499 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 500 error = EADDRNOTAVAIL; 501 goto bad; 502 } 503 if ((flags & IP_ALLOWBROADCAST) == 0) { 504 error = EACCES; 505 goto bad; 506 } 507 /* don't allow broadcast messages to be fragmented */ 508 if (ntohs(ip->ip_len) > ifp->if_mtu) { 509 error = EMSGSIZE; 510 goto bad; 511 } 512 m->m_flags |= M_BCAST; 513 } else 514 m->m_flags &= ~M_BCAST; 515 516 sendit: 517 /* 518 * If we're doing Path MTU Discovery, we need to set DF unless 519 * the route's MTU is locked. 520 */ 521 if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL && 522 (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 523 ip->ip_off |= htons(IP_DF); 524 525 /* Remember the current ip_len */ 526 ip_len = ntohs(ip->ip_len); 527 528 #ifdef IPSEC 529 /* get SP for this packet */ 530 if (so == NULL) 531 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 532 flags, &error); 533 else { 534 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 535 IPSEC_DIR_OUTBOUND)) 536 goto skip_ipsec; 537 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 538 } 539 540 if (sp == NULL) { 541 ipsecstat.out_inval++; 542 goto bad; 543 } 544 545 error = 0; 546 547 /* check policy */ 548 switch (sp->policy) { 549 case IPSEC_POLICY_DISCARD: 550 /* 551 * This packet is just discarded. 552 */ 553 ipsecstat.out_polvio++; 554 goto bad; 555 556 case IPSEC_POLICY_BYPASS: 557 case IPSEC_POLICY_NONE: 558 /* no need to do IPsec. */ 559 goto skip_ipsec; 560 561 case IPSEC_POLICY_IPSEC: 562 if (sp->req == NULL) { 563 /* XXX should be panic ? */ 564 printf("ip_output: No IPsec request specified.\n"); 565 error = EINVAL; 566 goto bad; 567 } 568 break; 569 570 case IPSEC_POLICY_ENTRUST: 571 default: 572 printf("ip_output: Invalid policy found. %d\n", sp->policy); 573 } 574 575 #ifdef IPSEC_NAT_T 576 /* 577 * NAT-T ESP fragmentation: don't do IPSec processing now, 578 * we'll do it on each fragmented packet. 579 */ 580 if (sp->req->sav && 581 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 582 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 583 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 584 natt_frag = 1; 585 mtu = sp->req->sav->esp_frag; 586 goto skip_ipsec; 587 } 588 } 589 #endif /* IPSEC_NAT_T */ 590 591 /* 592 * ipsec4_output() expects ip_len and ip_off in network 593 * order. They have been set to network order above. 594 */ 595 596 { 597 struct ipsec_output_state state; 598 bzero(&state, sizeof(state)); 599 state.m = m; 600 if (flags & IP_ROUTETOIF) { 601 state.ro = &iproute; 602 bzero(&iproute, sizeof(iproute)); 603 } else 604 state.ro = ro; 605 state.dst = (struct sockaddr *)dst; 606 607 /* 608 * We can't defer the checksum of payload data if 609 * we're about to encrypt/authenticate it. 610 * 611 * XXX When we support crypto offloading functions of 612 * XXX network interfaces, we need to reconsider this, 613 * XXX since it's likely that they'll support checksumming, 614 * XXX as well. 615 */ 616 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 617 in_delayed_cksum(m); 618 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 619 } 620 621 error = ipsec4_output(&state, sp, flags); 622 623 m = state.m; 624 if (flags & IP_ROUTETOIF) { 625 /* 626 * if we have tunnel mode SA, we may need to ignore 627 * IP_ROUTETOIF. 628 */ 629 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 630 flags &= ~IP_ROUTETOIF; 631 ro = state.ro; 632 } 633 } else 634 ro = state.ro; 635 dst = (struct sockaddr_in *)state.dst; 636 if (error) { 637 /* mbuf is already reclaimed in ipsec4_output. */ 638 m0 = NULL; 639 switch (error) { 640 case EHOSTUNREACH: 641 case ENETUNREACH: 642 case EMSGSIZE: 643 case ENOBUFS: 644 case ENOMEM: 645 break; 646 default: 647 printf("ip4_output (ipsec): error code %d\n", error); 648 /*fall through*/ 649 case ENOENT: 650 /* don't show these error codes to the user */ 651 error = 0; 652 break; 653 } 654 goto bad; 655 } 656 657 /* be sure to update variables that are affected by ipsec4_output() */ 658 ip = mtod(m, struct ip *); 659 hlen = ip->ip_hl << 2; 660 ip_len = ntohs(ip->ip_len); 661 662 if (ro->ro_rt == NULL) { 663 if ((flags & IP_ROUTETOIF) == 0) { 664 printf("ip_output: " 665 "can't update route after IPsec processing\n"); 666 error = EHOSTUNREACH; /*XXX*/ 667 goto bad; 668 } 669 } else { 670 /* nobody uses ia beyond here */ 671 if (state.encap) { 672 ifp = ro->ro_rt->rt_ifp; 673 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 674 mtu = ifp->if_mtu; 675 } 676 } 677 } 678 skip_ipsec: 679 #endif /*IPSEC*/ 680 #ifdef FAST_IPSEC 681 /* 682 * Check the security policy (SP) for the packet and, if 683 * required, do IPsec-related processing. There are two 684 * cases here; the first time a packet is sent through 685 * it will be untagged and handled by ipsec4_checkpolicy. 686 * If the packet is resubmitted to ip_output (e.g. after 687 * AH, ESP, etc. processing), there will be a tag to bypass 688 * the lookup and related policy checking. 689 */ 690 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 691 s = splsoftnet(); 692 if (mtag != NULL) { 693 tdbi = (struct tdb_ident *)(mtag + 1); 694 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 695 if (sp == NULL) 696 error = -EINVAL; /* force silent drop */ 697 m_tag_delete(m, mtag); 698 } else { 699 if (inp != NULL && 700 IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND)) 701 goto spd_done; 702 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 703 &error, inp); 704 } 705 /* 706 * There are four return cases: 707 * sp != NULL apply IPsec policy 708 * sp == NULL, error == 0 no IPsec handling needed 709 * sp == NULL, error == -EINVAL discard packet w/o error 710 * sp == NULL, error != 0 discard packet, report error 711 */ 712 if (sp != NULL) { 713 /* Loop detection, check if ipsec processing already done */ 714 IPSEC_ASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 715 for (mtag = m_tag_first(m); mtag != NULL; 716 mtag = m_tag_next(m, mtag)) { 717 #ifdef MTAG_ABI_COMPAT 718 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 719 continue; 720 #endif 721 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 722 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 723 continue; 724 /* 725 * Check if policy has an SA associated with it. 726 * This can happen when an SP has yet to acquire 727 * an SA; e.g. on first reference. If it occurs, 728 * then we let ipsec4_process_packet do its thing. 729 */ 730 if (sp->req->sav == NULL) 731 break; 732 tdbi = (struct tdb_ident *)(mtag + 1); 733 if (tdbi->spi == sp->req->sav->spi && 734 tdbi->proto == sp->req->sav->sah->saidx.proto && 735 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 736 sizeof (union sockaddr_union)) == 0) { 737 /* 738 * No IPsec processing is needed, free 739 * reference to SP. 740 * 741 * NB: null pointer to avoid free at 742 * done: below. 743 */ 744 KEY_FREESP(&sp), sp = NULL; 745 splx(s); 746 goto spd_done; 747 } 748 } 749 750 /* 751 * Do delayed checksums now because we send before 752 * this is done in the normal processing path. 753 */ 754 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 755 in_delayed_cksum(m); 756 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 757 } 758 759 #ifdef __FreeBSD__ 760 ip->ip_len = htons(ip->ip_len); 761 ip->ip_off = htons(ip->ip_off); 762 #endif 763 764 /* NB: callee frees mbuf */ 765 error = ipsec4_process_packet(m, sp->req, flags, 0); 766 /* 767 * Preserve KAME behaviour: ENOENT can be returned 768 * when an SA acquire is in progress. Don't propagate 769 * this to user-level; it confuses applications. 770 * 771 * XXX this will go away when the SADB is redone. 772 */ 773 if (error == ENOENT) 774 error = 0; 775 splx(s); 776 goto done; 777 } else { 778 splx(s); 779 780 if (error != 0) { 781 /* 782 * Hack: -EINVAL is used to signal that a packet 783 * should be silently discarded. This is typically 784 * because we asked key management for an SA and 785 * it was delayed (e.g. kicked up to IKE). 786 */ 787 if (error == -EINVAL) 788 error = 0; 789 goto bad; 790 } else { 791 /* No IPsec processing for this packet. */ 792 } 793 #ifdef notyet 794 /* 795 * If deferred crypto processing is needed, check that 796 * the interface supports it. 797 */ 798 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 799 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 800 /* notify IPsec to do its own crypto */ 801 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 802 error = EHOSTUNREACH; 803 goto bad; 804 } 805 #endif 806 } 807 spd_done: 808 #endif /* FAST_IPSEC */ 809 810 #ifdef PFIL_HOOKS 811 /* 812 * Run through list of hooks for output packets. 813 */ 814 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 815 goto done; 816 if (m == NULL) 817 goto done; 818 819 ip = mtod(m, struct ip *); 820 hlen = ip->ip_hl << 2; 821 #endif /* PFIL_HOOKS */ 822 823 m->m_pkthdr.csum_data |= hlen << 16; 824 825 #if IFA_STATS 826 /* 827 * search for the source address structure to 828 * maintain output statistics. 829 */ 830 INADDR_TO_IA(ip->ip_src, ia); 831 #endif 832 833 /* Maybe skip checksums on loopback interfaces. */ 834 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 835 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 836 } 837 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 838 /* 839 * If small enough for mtu of path, or if using TCP segmentation 840 * offload, can just send directly. 841 */ 842 if (ip_len <= mtu || 843 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 844 #if IFA_STATS 845 if (ia) 846 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 847 #endif 848 /* 849 * Always initialize the sum to 0! Some HW assisted 850 * checksumming requires this. 851 */ 852 ip->ip_sum = 0; 853 854 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 855 /* 856 * Perform any checksums that the hardware can't do 857 * for us. 858 * 859 * XXX Does any hardware require the {th,uh}_sum 860 * XXX fields to be 0? 861 */ 862 if (sw_csum & M_CSUM_IPv4) { 863 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 864 ip->ip_sum = in_cksum(m, hlen); 865 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 866 } 867 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 868 if (IN_NEED_CHECKSUM(ifp, 869 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 870 in_delayed_cksum(m); 871 } 872 m->m_pkthdr.csum_flags &= 873 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 874 } 875 } 876 877 #ifdef IPSEC 878 /* clean ipsec history once it goes out of the node */ 879 ipsec_delaux(m); 880 #endif 881 882 if (__predict_true( 883 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 || 884 (ifp->if_capenable & IFCAP_TSOv4) != 0)) { 885 error = 886 (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt); 887 } else { 888 error = 889 ip_tso_output(ifp, m, sintosa(dst), ro->ro_rt); 890 } 891 goto done; 892 } 893 894 /* 895 * We can't use HW checksumming if we're about to 896 * to fragment the packet. 897 * 898 * XXX Some hardware can do this. 899 */ 900 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 901 if (IN_NEED_CHECKSUM(ifp, 902 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 903 in_delayed_cksum(m); 904 } 905 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 906 } 907 908 /* 909 * Too large for interface; fragment if possible. 910 * Must be able to put at least 8 bytes per fragment. 911 */ 912 if (ntohs(ip->ip_off) & IP_DF) { 913 if (flags & IP_RETURNMTU) 914 *mtu_p = mtu; 915 error = EMSGSIZE; 916 ipstat.ips_cantfrag++; 917 goto bad; 918 } 919 920 error = ip_fragment(m, ifp, mtu); 921 if (error) { 922 m = NULL; 923 goto bad; 924 } 925 926 for (; m; m = m0) { 927 m0 = m->m_nextpkt; 928 m->m_nextpkt = 0; 929 if (error == 0) { 930 #if IFA_STATS 931 if (ia) 932 ia->ia_ifa.ifa_data.ifad_outbytes += 933 ntohs(ip->ip_len); 934 #endif 935 #ifdef IPSEC 936 /* clean ipsec history once it goes out of the node */ 937 ipsec_delaux(m); 938 939 #ifdef IPSEC_NAT_T 940 /* 941 * If we get there, the packet has not been handeld by 942 * IPSec whereas it should have. Now that it has been 943 * fragmented, re-inject it in ip_output so that IPsec 944 * processing can occur. 945 */ 946 if (natt_frag) { 947 error = ip_output(m, opt, 948 ro, flags, imo, so, mtu_p); 949 } else 950 #endif /* IPSEC_NAT_T */ 951 #endif /* IPSEC */ 952 { 953 KASSERT((m->m_pkthdr.csum_flags & 954 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 955 error = (*ifp->if_output)(ifp, m, sintosa(dst), 956 ro->ro_rt); 957 } 958 } else 959 m_freem(m); 960 } 961 962 if (error == 0) 963 ipstat.ips_fragmented++; 964 done: 965 if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) { 966 RTFREE(ro->ro_rt); 967 ro->ro_rt = 0; 968 } 969 970 #ifdef IPSEC 971 if (sp != NULL) { 972 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 973 printf("DP ip_output call free SP:%p\n", sp)); 974 key_freesp(sp); 975 } 976 #endif /* IPSEC */ 977 #ifdef FAST_IPSEC 978 if (sp != NULL) 979 KEY_FREESP(&sp); 980 #endif /* FAST_IPSEC */ 981 982 return (error); 983 bad: 984 m_freem(m); 985 goto done; 986 } 987 988 int 989 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 990 { 991 struct ip *ip, *mhip; 992 struct mbuf *m0; 993 int len, hlen, off; 994 int mhlen, firstlen; 995 struct mbuf **mnext; 996 int sw_csum = m->m_pkthdr.csum_flags; 997 int fragments = 0; 998 int s; 999 int error = 0; 1000 1001 ip = mtod(m, struct ip *); 1002 hlen = ip->ip_hl << 2; 1003 if (ifp != NULL) 1004 sw_csum &= ~ifp->if_csum_flags_tx; 1005 1006 len = (mtu - hlen) &~ 7; 1007 if (len < 8) { 1008 m_freem(m); 1009 return (EMSGSIZE); 1010 } 1011 1012 firstlen = len; 1013 mnext = &m->m_nextpkt; 1014 1015 /* 1016 * Loop through length of segment after first fragment, 1017 * make new header and copy data of each part and link onto chain. 1018 */ 1019 m0 = m; 1020 mhlen = sizeof (struct ip); 1021 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 1022 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1023 if (m == 0) { 1024 error = ENOBUFS; 1025 ipstat.ips_odropped++; 1026 goto sendorfree; 1027 } 1028 MCLAIM(m, m0->m_owner); 1029 *mnext = m; 1030 mnext = &m->m_nextpkt; 1031 m->m_data += max_linkhdr; 1032 mhip = mtod(m, struct ip *); 1033 *mhip = *ip; 1034 /* we must inherit MCAST and BCAST flags */ 1035 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 1036 if (hlen > sizeof (struct ip)) { 1037 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 1038 mhip->ip_hl = mhlen >> 2; 1039 } 1040 m->m_len = mhlen; 1041 mhip->ip_off = ((off - hlen) >> 3) + 1042 (ntohs(ip->ip_off) & ~IP_MF); 1043 if (ip->ip_off & htons(IP_MF)) 1044 mhip->ip_off |= IP_MF; 1045 if (off + len >= ntohs(ip->ip_len)) 1046 len = ntohs(ip->ip_len) - off; 1047 else 1048 mhip->ip_off |= IP_MF; 1049 HTONS(mhip->ip_off); 1050 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 1051 m->m_next = m_copy(m0, off, len); 1052 if (m->m_next == 0) { 1053 error = ENOBUFS; /* ??? */ 1054 ipstat.ips_odropped++; 1055 goto sendorfree; 1056 } 1057 m->m_pkthdr.len = mhlen + len; 1058 m->m_pkthdr.rcvif = (struct ifnet *)0; 1059 mhip->ip_sum = 0; 1060 if (sw_csum & M_CSUM_IPv4) { 1061 mhip->ip_sum = in_cksum(m, mhlen); 1062 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 1063 } else { 1064 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 1065 m->m_pkthdr.csum_data |= mhlen << 16; 1066 } 1067 ipstat.ips_ofragments++; 1068 fragments++; 1069 } 1070 /* 1071 * Update first fragment by trimming what's been copied out 1072 * and updating header, then send each fragment (in order). 1073 */ 1074 m = m0; 1075 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 1076 m->m_pkthdr.len = hlen + firstlen; 1077 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 1078 ip->ip_off |= htons(IP_MF); 1079 ip->ip_sum = 0; 1080 if (sw_csum & M_CSUM_IPv4) { 1081 ip->ip_sum = in_cksum(m, hlen); 1082 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 1083 } else { 1084 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 1085 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 1086 sizeof(struct ip)); 1087 } 1088 sendorfree: 1089 /* 1090 * If there is no room for all the fragments, don't queue 1091 * any of them. 1092 */ 1093 if (ifp != NULL) { 1094 s = splnet(); 1095 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 1096 error == 0) { 1097 error = ENOBUFS; 1098 ipstat.ips_odropped++; 1099 IFQ_INC_DROPS(&ifp->if_snd); 1100 } 1101 splx(s); 1102 } 1103 if (error) { 1104 for (m = m0; m; m = m0) { 1105 m0 = m->m_nextpkt; 1106 m->m_nextpkt = NULL; 1107 m_freem(m); 1108 } 1109 } 1110 return (error); 1111 } 1112 1113 /* 1114 * Process a delayed payload checksum calculation. 1115 */ 1116 void 1117 in_delayed_cksum(struct mbuf *m) 1118 { 1119 struct ip *ip; 1120 u_int16_t csum, offset; 1121 1122 ip = mtod(m, struct ip *); 1123 offset = ip->ip_hl << 2; 1124 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 1125 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 1126 csum = 0xffff; 1127 1128 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 1129 1130 if ((offset + sizeof(u_int16_t)) > m->m_len) { 1131 /* This happen when ip options were inserted 1132 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 1133 m->m_len, offset, ip->ip_p); 1134 */ 1135 m_copyback(m, offset, sizeof(csum), (caddr_t) &csum); 1136 } else 1137 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum; 1138 } 1139 1140 /* 1141 * Determine the maximum length of the options to be inserted; 1142 * we would far rather allocate too much space rather than too little. 1143 */ 1144 1145 u_int 1146 ip_optlen(struct inpcb *inp) 1147 { 1148 struct mbuf *m = inp->inp_options; 1149 1150 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 1151 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 1152 else 1153 return 0; 1154 } 1155 1156 1157 /* 1158 * Insert IP options into preformed packet. 1159 * Adjust IP destination as required for IP source routing, 1160 * as indicated by a non-zero in_addr at the start of the options. 1161 */ 1162 static struct mbuf * 1163 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 1164 { 1165 struct ipoption *p = mtod(opt, struct ipoption *); 1166 struct mbuf *n; 1167 struct ip *ip = mtod(m, struct ip *); 1168 unsigned optlen; 1169 1170 optlen = opt->m_len - sizeof(p->ipopt_dst); 1171 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 1172 return (m); /* XXX should fail */ 1173 if (!in_nullhost(p->ipopt_dst)) 1174 ip->ip_dst = p->ipopt_dst; 1175 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 1176 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1177 if (n == 0) 1178 return (m); 1179 MCLAIM(n, m->m_owner); 1180 M_COPY_PKTHDR(n, m); 1181 m_tag_delete_chain(m, NULL); 1182 m->m_flags &= ~M_PKTHDR; 1183 m->m_len -= sizeof(struct ip); 1184 m->m_data += sizeof(struct ip); 1185 n->m_next = m; 1186 m = n; 1187 m->m_len = optlen + sizeof(struct ip); 1188 m->m_data += max_linkhdr; 1189 bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1190 } else { 1191 m->m_data -= optlen; 1192 m->m_len += optlen; 1193 memmove(mtod(m, caddr_t), ip, sizeof(struct ip)); 1194 } 1195 m->m_pkthdr.len += optlen; 1196 ip = mtod(m, struct ip *); 1197 bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen); 1198 *phlen = sizeof(struct ip) + optlen; 1199 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1200 return (m); 1201 } 1202 1203 /* 1204 * Copy options from ip to jp, 1205 * omitting those not copied during fragmentation. 1206 */ 1207 int 1208 ip_optcopy(struct ip *ip, struct ip *jp) 1209 { 1210 u_char *cp, *dp; 1211 int opt, optlen, cnt; 1212 1213 cp = (u_char *)(ip + 1); 1214 dp = (u_char *)(jp + 1); 1215 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1216 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1217 opt = cp[0]; 1218 if (opt == IPOPT_EOL) 1219 break; 1220 if (opt == IPOPT_NOP) { 1221 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1222 *dp++ = IPOPT_NOP; 1223 optlen = 1; 1224 continue; 1225 } 1226 #ifdef DIAGNOSTIC 1227 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1228 panic("malformed IPv4 option passed to ip_optcopy"); 1229 #endif 1230 optlen = cp[IPOPT_OLEN]; 1231 #ifdef DIAGNOSTIC 1232 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1233 panic("malformed IPv4 option passed to ip_optcopy"); 1234 #endif 1235 /* bogus lengths should have been caught by ip_dooptions */ 1236 if (optlen > cnt) 1237 optlen = cnt; 1238 if (IPOPT_COPIED(opt)) { 1239 bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen); 1240 dp += optlen; 1241 } 1242 } 1243 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1244 *dp++ = IPOPT_EOL; 1245 return (optlen); 1246 } 1247 1248 /* 1249 * IP socket option processing. 1250 */ 1251 int 1252 ip_ctloutput(int op, struct socket *so, int level, int optname, 1253 struct mbuf **mp) 1254 { 1255 struct inpcb *inp = sotoinpcb(so); 1256 struct mbuf *m = *mp; 1257 int optval = 0; 1258 int error = 0; 1259 #if defined(IPSEC) || defined(FAST_IPSEC) 1260 struct proc *p = curproc; /*XXX*/ 1261 #endif 1262 1263 if (level != IPPROTO_IP) { 1264 error = EINVAL; 1265 if (op == PRCO_SETOPT && *mp) 1266 (void) m_free(*mp); 1267 } else switch (op) { 1268 1269 case PRCO_SETOPT: 1270 switch (optname) { 1271 case IP_OPTIONS: 1272 #ifdef notyet 1273 case IP_RETOPTS: 1274 return (ip_pcbopts(optname, &inp->inp_options, m)); 1275 #else 1276 return (ip_pcbopts(&inp->inp_options, m)); 1277 #endif 1278 1279 case IP_TOS: 1280 case IP_TTL: 1281 case IP_RECVOPTS: 1282 case IP_RECVRETOPTS: 1283 case IP_RECVDSTADDR: 1284 case IP_RECVIF: 1285 if (m == NULL || m->m_len != sizeof(int)) 1286 error = EINVAL; 1287 else { 1288 optval = *mtod(m, int *); 1289 switch (optname) { 1290 1291 case IP_TOS: 1292 inp->inp_ip.ip_tos = optval; 1293 break; 1294 1295 case IP_TTL: 1296 inp->inp_ip.ip_ttl = optval; 1297 break; 1298 #define OPTSET(bit) \ 1299 if (optval) \ 1300 inp->inp_flags |= bit; \ 1301 else \ 1302 inp->inp_flags &= ~bit; 1303 1304 case IP_RECVOPTS: 1305 OPTSET(INP_RECVOPTS); 1306 break; 1307 1308 case IP_RECVRETOPTS: 1309 OPTSET(INP_RECVRETOPTS); 1310 break; 1311 1312 case IP_RECVDSTADDR: 1313 OPTSET(INP_RECVDSTADDR); 1314 break; 1315 1316 case IP_RECVIF: 1317 OPTSET(INP_RECVIF); 1318 break; 1319 } 1320 } 1321 break; 1322 #undef OPTSET 1323 1324 case IP_MULTICAST_IF: 1325 case IP_MULTICAST_TTL: 1326 case IP_MULTICAST_LOOP: 1327 case IP_ADD_MEMBERSHIP: 1328 case IP_DROP_MEMBERSHIP: 1329 error = ip_setmoptions(optname, &inp->inp_moptions, m); 1330 break; 1331 1332 case IP_PORTRANGE: 1333 if (m == 0 || m->m_len != sizeof(int)) 1334 error = EINVAL; 1335 else { 1336 optval = *mtod(m, int *); 1337 1338 switch (optval) { 1339 1340 case IP_PORTRANGE_DEFAULT: 1341 case IP_PORTRANGE_HIGH: 1342 inp->inp_flags &= ~(INP_LOWPORT); 1343 break; 1344 1345 case IP_PORTRANGE_LOW: 1346 inp->inp_flags |= INP_LOWPORT; 1347 break; 1348 1349 default: 1350 error = EINVAL; 1351 break; 1352 } 1353 } 1354 break; 1355 1356 #if defined(IPSEC) || defined(FAST_IPSEC) 1357 case IP_IPSEC_POLICY: 1358 { 1359 caddr_t req = NULL; 1360 size_t len = 0; 1361 int priv = 0; 1362 1363 #ifdef __NetBSD__ 1364 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1365 priv = 0; 1366 else 1367 priv = 1; 1368 #else 1369 priv = (in6p->in6p_socket->so_state & SS_PRIV); 1370 #endif 1371 if (m) { 1372 req = mtod(m, caddr_t); 1373 len = m->m_len; 1374 } 1375 error = ipsec4_set_policy(inp, optname, req, len, priv); 1376 break; 1377 } 1378 #endif /*IPSEC*/ 1379 1380 default: 1381 error = ENOPROTOOPT; 1382 break; 1383 } 1384 if (m) 1385 (void)m_free(m); 1386 break; 1387 1388 case PRCO_GETOPT: 1389 switch (optname) { 1390 case IP_OPTIONS: 1391 case IP_RETOPTS: 1392 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1393 MCLAIM(m, so->so_mowner); 1394 if (inp->inp_options) { 1395 m->m_len = inp->inp_options->m_len; 1396 bcopy(mtod(inp->inp_options, caddr_t), 1397 mtod(m, caddr_t), (unsigned)m->m_len); 1398 } else 1399 m->m_len = 0; 1400 break; 1401 1402 case IP_TOS: 1403 case IP_TTL: 1404 case IP_RECVOPTS: 1405 case IP_RECVRETOPTS: 1406 case IP_RECVDSTADDR: 1407 case IP_RECVIF: 1408 case IP_ERRORMTU: 1409 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1410 MCLAIM(m, so->so_mowner); 1411 m->m_len = sizeof(int); 1412 switch (optname) { 1413 1414 case IP_TOS: 1415 optval = inp->inp_ip.ip_tos; 1416 break; 1417 1418 case IP_TTL: 1419 optval = inp->inp_ip.ip_ttl; 1420 break; 1421 1422 case IP_ERRORMTU: 1423 optval = inp->inp_errormtu; 1424 break; 1425 1426 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1427 1428 case IP_RECVOPTS: 1429 optval = OPTBIT(INP_RECVOPTS); 1430 break; 1431 1432 case IP_RECVRETOPTS: 1433 optval = OPTBIT(INP_RECVRETOPTS); 1434 break; 1435 1436 case IP_RECVDSTADDR: 1437 optval = OPTBIT(INP_RECVDSTADDR); 1438 break; 1439 1440 case IP_RECVIF: 1441 optval = OPTBIT(INP_RECVIF); 1442 break; 1443 } 1444 *mtod(m, int *) = optval; 1445 break; 1446 1447 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */ 1448 /* XXX: code broken */ 1449 case IP_IPSEC_POLICY: 1450 { 1451 caddr_t req = NULL; 1452 size_t len = 0; 1453 1454 if (m) { 1455 req = mtod(m, caddr_t); 1456 len = m->m_len; 1457 } 1458 error = ipsec4_get_policy(inp, req, len, mp); 1459 break; 1460 } 1461 #endif /*IPSEC*/ 1462 1463 case IP_MULTICAST_IF: 1464 case IP_MULTICAST_TTL: 1465 case IP_MULTICAST_LOOP: 1466 case IP_ADD_MEMBERSHIP: 1467 case IP_DROP_MEMBERSHIP: 1468 error = ip_getmoptions(optname, inp->inp_moptions, mp); 1469 if (*mp) 1470 MCLAIM(*mp, so->so_mowner); 1471 break; 1472 1473 case IP_PORTRANGE: 1474 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1475 MCLAIM(m, so->so_mowner); 1476 m->m_len = sizeof(int); 1477 1478 if (inp->inp_flags & INP_LOWPORT) 1479 optval = IP_PORTRANGE_LOW; 1480 else 1481 optval = IP_PORTRANGE_DEFAULT; 1482 1483 *mtod(m, int *) = optval; 1484 break; 1485 1486 default: 1487 error = ENOPROTOOPT; 1488 break; 1489 } 1490 break; 1491 } 1492 return (error); 1493 } 1494 1495 /* 1496 * Set up IP options in pcb for insertion in output packets. 1497 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1498 * with destination address if source routed. 1499 */ 1500 int 1501 #ifdef notyet 1502 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m) 1503 #else 1504 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m) 1505 #endif 1506 { 1507 int cnt, optlen; 1508 u_char *cp; 1509 u_char opt; 1510 1511 /* turn off any old options */ 1512 if (*pcbopt) 1513 (void)m_free(*pcbopt); 1514 *pcbopt = 0; 1515 if (m == (struct mbuf *)0 || m->m_len == 0) { 1516 /* 1517 * Only turning off any previous options. 1518 */ 1519 if (m) 1520 (void)m_free(m); 1521 return (0); 1522 } 1523 1524 #ifndef __vax__ 1525 if (m->m_len % sizeof(int32_t)) 1526 goto bad; 1527 #endif 1528 /* 1529 * IP first-hop destination address will be stored before 1530 * actual options; move other options back 1531 * and clear it when none present. 1532 */ 1533 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1534 goto bad; 1535 cnt = m->m_len; 1536 m->m_len += sizeof(struct in_addr); 1537 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1538 memmove(cp, mtod(m, caddr_t), (unsigned)cnt); 1539 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1540 1541 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1542 opt = cp[IPOPT_OPTVAL]; 1543 if (opt == IPOPT_EOL) 1544 break; 1545 if (opt == IPOPT_NOP) 1546 optlen = 1; 1547 else { 1548 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1549 goto bad; 1550 optlen = cp[IPOPT_OLEN]; 1551 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1552 goto bad; 1553 } 1554 switch (opt) { 1555 1556 default: 1557 break; 1558 1559 case IPOPT_LSRR: 1560 case IPOPT_SSRR: 1561 /* 1562 * user process specifies route as: 1563 * ->A->B->C->D 1564 * D must be our final destination (but we can't 1565 * check that since we may not have connected yet). 1566 * A is first hop destination, which doesn't appear in 1567 * actual IP option, but is stored before the options. 1568 */ 1569 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1570 goto bad; 1571 m->m_len -= sizeof(struct in_addr); 1572 cnt -= sizeof(struct in_addr); 1573 optlen -= sizeof(struct in_addr); 1574 cp[IPOPT_OLEN] = optlen; 1575 /* 1576 * Move first hop before start of options. 1577 */ 1578 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1579 sizeof(struct in_addr)); 1580 /* 1581 * Then copy rest of options back 1582 * to close up the deleted entry. 1583 */ 1584 (void)memmove(&cp[IPOPT_OFFSET+1], 1585 &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr), 1586 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1587 break; 1588 } 1589 } 1590 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1591 goto bad; 1592 *pcbopt = m; 1593 return (0); 1594 1595 bad: 1596 (void)m_free(m); 1597 return (EINVAL); 1598 } 1599 1600 /* 1601 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1602 */ 1603 static struct ifnet * 1604 ip_multicast_if(struct in_addr *a, int *ifindexp) 1605 { 1606 int ifindex; 1607 struct ifnet *ifp = NULL; 1608 struct in_ifaddr *ia; 1609 1610 if (ifindexp) 1611 *ifindexp = 0; 1612 if (ntohl(a->s_addr) >> 24 == 0) { 1613 ifindex = ntohl(a->s_addr) & 0xffffff; 1614 if (ifindex < 0 || if_indexlim <= ifindex) 1615 return NULL; 1616 ifp = ifindex2ifnet[ifindex]; 1617 if (!ifp) 1618 return NULL; 1619 if (ifindexp) 1620 *ifindexp = ifindex; 1621 } else { 1622 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1623 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1624 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1625 ifp = ia->ia_ifp; 1626 break; 1627 } 1628 } 1629 } 1630 return ifp; 1631 } 1632 1633 /* 1634 * Set the IP multicast options in response to user setsockopt(). 1635 */ 1636 int 1637 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m) 1638 { 1639 int error = 0; 1640 u_char loop; 1641 int i; 1642 struct in_addr addr; 1643 struct ip_mreq *mreq; 1644 struct ifnet *ifp; 1645 struct ip_moptions *imo = *imop; 1646 struct route ro; 1647 struct sockaddr_in *dst; 1648 int ifindex; 1649 1650 if (imo == NULL) { 1651 /* 1652 * No multicast option buffer attached to the pcb; 1653 * allocate one and initialize to default values. 1654 */ 1655 imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS, 1656 M_WAITOK); 1657 1658 if (imo == NULL) 1659 return (ENOBUFS); 1660 *imop = imo; 1661 imo->imo_multicast_ifp = NULL; 1662 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1663 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1664 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1665 imo->imo_num_memberships = 0; 1666 } 1667 1668 switch (optname) { 1669 1670 case IP_MULTICAST_IF: 1671 /* 1672 * Select the interface for outgoing multicast packets. 1673 */ 1674 if (m == NULL || m->m_len != sizeof(struct in_addr)) { 1675 error = EINVAL; 1676 break; 1677 } 1678 addr = *(mtod(m, struct in_addr *)); 1679 /* 1680 * INADDR_ANY is used to remove a previous selection. 1681 * When no interface is selected, a default one is 1682 * chosen every time a multicast packet is sent. 1683 */ 1684 if (in_nullhost(addr)) { 1685 imo->imo_multicast_ifp = NULL; 1686 break; 1687 } 1688 /* 1689 * The selected interface is identified by its local 1690 * IP address. Find the interface and confirm that 1691 * it supports multicasting. 1692 */ 1693 ifp = ip_multicast_if(&addr, &ifindex); 1694 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1695 error = EADDRNOTAVAIL; 1696 break; 1697 } 1698 imo->imo_multicast_ifp = ifp; 1699 if (ifindex) 1700 imo->imo_multicast_addr = addr; 1701 else 1702 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1703 break; 1704 1705 case IP_MULTICAST_TTL: 1706 /* 1707 * Set the IP time-to-live for outgoing multicast packets. 1708 */ 1709 if (m == NULL || m->m_len != 1) { 1710 error = EINVAL; 1711 break; 1712 } 1713 imo->imo_multicast_ttl = *(mtod(m, u_char *)); 1714 break; 1715 1716 case IP_MULTICAST_LOOP: 1717 /* 1718 * Set the loopback flag for outgoing multicast packets. 1719 * Must be zero or one. 1720 */ 1721 if (m == NULL || m->m_len != 1 || 1722 (loop = *(mtod(m, u_char *))) > 1) { 1723 error = EINVAL; 1724 break; 1725 } 1726 imo->imo_multicast_loop = loop; 1727 break; 1728 1729 case IP_ADD_MEMBERSHIP: 1730 /* 1731 * Add a multicast group membership. 1732 * Group must be a valid IP multicast address. 1733 */ 1734 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1735 error = EINVAL; 1736 break; 1737 } 1738 mreq = mtod(m, struct ip_mreq *); 1739 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1740 error = EINVAL; 1741 break; 1742 } 1743 /* 1744 * If no interface address was provided, use the interface of 1745 * the route to the given multicast address. 1746 */ 1747 if (in_nullhost(mreq->imr_interface)) { 1748 bzero((caddr_t)&ro, sizeof(ro)); 1749 ro.ro_rt = NULL; 1750 dst = satosin(&ro.ro_dst); 1751 dst->sin_len = sizeof(*dst); 1752 dst->sin_family = AF_INET; 1753 dst->sin_addr = mreq->imr_multiaddr; 1754 rtalloc(&ro); 1755 if (ro.ro_rt == NULL) { 1756 error = EADDRNOTAVAIL; 1757 break; 1758 } 1759 ifp = ro.ro_rt->rt_ifp; 1760 rtfree(ro.ro_rt); 1761 } else { 1762 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1763 } 1764 /* 1765 * See if we found an interface, and confirm that it 1766 * supports multicast. 1767 */ 1768 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1769 error = EADDRNOTAVAIL; 1770 break; 1771 } 1772 /* 1773 * See if the membership already exists or if all the 1774 * membership slots are full. 1775 */ 1776 for (i = 0; i < imo->imo_num_memberships; ++i) { 1777 if (imo->imo_membership[i]->inm_ifp == ifp && 1778 in_hosteq(imo->imo_membership[i]->inm_addr, 1779 mreq->imr_multiaddr)) 1780 break; 1781 } 1782 if (i < imo->imo_num_memberships) { 1783 error = EADDRINUSE; 1784 break; 1785 } 1786 if (i == IP_MAX_MEMBERSHIPS) { 1787 error = ETOOMANYREFS; 1788 break; 1789 } 1790 /* 1791 * Everything looks good; add a new record to the multicast 1792 * address list for the given interface. 1793 */ 1794 if ((imo->imo_membership[i] = 1795 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1796 error = ENOBUFS; 1797 break; 1798 } 1799 ++imo->imo_num_memberships; 1800 break; 1801 1802 case IP_DROP_MEMBERSHIP: 1803 /* 1804 * Drop a multicast group membership. 1805 * Group must be a valid IP multicast address. 1806 */ 1807 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1808 error = EINVAL; 1809 break; 1810 } 1811 mreq = mtod(m, struct ip_mreq *); 1812 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1813 error = EINVAL; 1814 break; 1815 } 1816 /* 1817 * If an interface address was specified, get a pointer 1818 * to its ifnet structure. 1819 */ 1820 if (in_nullhost(mreq->imr_interface)) 1821 ifp = NULL; 1822 else { 1823 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1824 if (ifp == NULL) { 1825 error = EADDRNOTAVAIL; 1826 break; 1827 } 1828 } 1829 /* 1830 * Find the membership in the membership array. 1831 */ 1832 for (i = 0; i < imo->imo_num_memberships; ++i) { 1833 if ((ifp == NULL || 1834 imo->imo_membership[i]->inm_ifp == ifp) && 1835 in_hosteq(imo->imo_membership[i]->inm_addr, 1836 mreq->imr_multiaddr)) 1837 break; 1838 } 1839 if (i == imo->imo_num_memberships) { 1840 error = EADDRNOTAVAIL; 1841 break; 1842 } 1843 /* 1844 * Give up the multicast address record to which the 1845 * membership points. 1846 */ 1847 in_delmulti(imo->imo_membership[i]); 1848 /* 1849 * Remove the gap in the membership array. 1850 */ 1851 for (++i; i < imo->imo_num_memberships; ++i) 1852 imo->imo_membership[i-1] = imo->imo_membership[i]; 1853 --imo->imo_num_memberships; 1854 break; 1855 1856 default: 1857 error = EOPNOTSUPP; 1858 break; 1859 } 1860 1861 /* 1862 * If all options have default values, no need to keep the mbuf. 1863 */ 1864 if (imo->imo_multicast_ifp == NULL && 1865 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1866 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1867 imo->imo_num_memberships == 0) { 1868 free(*imop, M_IPMOPTS); 1869 *imop = NULL; 1870 } 1871 1872 return (error); 1873 } 1874 1875 /* 1876 * Return the IP multicast options in response to user getsockopt(). 1877 */ 1878 int 1879 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp) 1880 { 1881 u_char *ttl; 1882 u_char *loop; 1883 struct in_addr *addr; 1884 struct in_ifaddr *ia; 1885 1886 *mp = m_get(M_WAIT, MT_SOOPTS); 1887 1888 switch (optname) { 1889 1890 case IP_MULTICAST_IF: 1891 addr = mtod(*mp, struct in_addr *); 1892 (*mp)->m_len = sizeof(struct in_addr); 1893 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1894 *addr = zeroin_addr; 1895 else if (imo->imo_multicast_addr.s_addr) { 1896 /* return the value user has set */ 1897 *addr = imo->imo_multicast_addr; 1898 } else { 1899 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1900 *addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1901 } 1902 return (0); 1903 1904 case IP_MULTICAST_TTL: 1905 ttl = mtod(*mp, u_char *); 1906 (*mp)->m_len = 1; 1907 *ttl = imo ? imo->imo_multicast_ttl 1908 : IP_DEFAULT_MULTICAST_TTL; 1909 return (0); 1910 1911 case IP_MULTICAST_LOOP: 1912 loop = mtod(*mp, u_char *); 1913 (*mp)->m_len = 1; 1914 *loop = imo ? imo->imo_multicast_loop 1915 : IP_DEFAULT_MULTICAST_LOOP; 1916 return (0); 1917 1918 default: 1919 return (EOPNOTSUPP); 1920 } 1921 } 1922 1923 /* 1924 * Discard the IP multicast options. 1925 */ 1926 void 1927 ip_freemoptions(struct ip_moptions *imo) 1928 { 1929 int i; 1930 1931 if (imo != NULL) { 1932 for (i = 0; i < imo->imo_num_memberships; ++i) 1933 in_delmulti(imo->imo_membership[i]); 1934 free(imo, M_IPMOPTS); 1935 } 1936 } 1937 1938 /* 1939 * Routine called from ip_output() to loop back a copy of an IP multicast 1940 * packet to the input queue of a specified interface. Note that this 1941 * calls the output routine of the loopback "driver", but with an interface 1942 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1943 */ 1944 static void 1945 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst) 1946 { 1947 struct ip *ip; 1948 struct mbuf *copym; 1949 1950 copym = m_copy(m, 0, M_COPYALL); 1951 if (copym != NULL 1952 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1953 copym = m_pullup(copym, sizeof(struct ip)); 1954 if (copym != NULL) { 1955 /* 1956 * We don't bother to fragment if the IP length is greater 1957 * than the interface's MTU. Can this possibly matter? 1958 */ 1959 ip = mtod(copym, struct ip *); 1960 1961 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1962 in_delayed_cksum(copym); 1963 copym->m_pkthdr.csum_flags &= 1964 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1965 } 1966 1967 ip->ip_sum = 0; 1968 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1969 (void) looutput(ifp, copym, sintosa(dst), NULL); 1970 } 1971 } 1972