ip_output.c revision 1.210.6.3
1 /* $NetBSD: ip_output.c,v 1.210.6.3 2012/06/02 11:09:38 mrg Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 50 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 51 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 52 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 53 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 54 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 55 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 56 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 57 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 58 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 59 * POSSIBILITY OF SUCH DAMAGE. 60 */ 61 62 /* 63 * Copyright (c) 1982, 1986, 1988, 1990, 1993 64 * The Regents of the University of California. All rights reserved. 65 * 66 * Redistribution and use in source and binary forms, with or without 67 * modification, are permitted provided that the following conditions 68 * are met: 69 * 1. Redistributions of source code must retain the above copyright 70 * notice, this list of conditions and the following disclaimer. 71 * 2. Redistributions in binary form must reproduce the above copyright 72 * notice, this list of conditions and the following disclaimer in the 73 * documentation and/or other materials provided with the distribution. 74 * 3. Neither the name of the University nor the names of its contributors 75 * may be used to endorse or promote products derived from this software 76 * without specific prior written permission. 77 * 78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 88 * SUCH DAMAGE. 89 * 90 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 91 */ 92 93 #include <sys/cdefs.h> 94 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.210.6.3 2012/06/02 11:09:38 mrg Exp $"); 95 96 #include "opt_pfil_hooks.h" 97 #include "opt_inet.h" 98 #include "opt_ipsec.h" 99 #include "opt_mrouting.h" 100 101 #include <sys/param.h> 102 #include <sys/malloc.h> 103 #include <sys/kmem.h> 104 #include <sys/mbuf.h> 105 #include <sys/errno.h> 106 #include <sys/protosw.h> 107 #include <sys/socket.h> 108 #include <sys/socketvar.h> 109 #include <sys/kauth.h> 110 #ifdef FAST_IPSEC 111 #include <sys/domain.h> 112 #endif 113 #include <sys/systm.h> 114 #include <sys/proc.h> 115 116 #include <net/if.h> 117 #include <net/route.h> 118 #include <net/pfil.h> 119 120 #include <netinet/in.h> 121 #include <netinet/in_systm.h> 122 #include <netinet/ip.h> 123 #include <netinet/in_pcb.h> 124 #include <netinet/in_var.h> 125 #include <netinet/ip_var.h> 126 #include <netinet/ip_private.h> 127 #include <netinet/in_offload.h> 128 129 #ifdef MROUTING 130 #include <netinet/ip_mroute.h> 131 #endif 132 133 #ifdef FAST_IPSEC 134 #include <netipsec/ipsec.h> 135 #include <netipsec/key.h> 136 #include <netipsec/xform.h> 137 #endif /* FAST_IPSEC*/ 138 139 #ifdef IPSEC_NAT_T 140 #include <netinet/udp.h> 141 #endif 142 143 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 144 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 145 static void ip_mloopback(struct ifnet *, struct mbuf *, 146 const struct sockaddr_in *); 147 148 #ifdef PFIL_HOOKS 149 extern struct pfil_head inet_pfil_hook; /* XXX */ 150 #endif 151 152 int ip_do_loopback_cksum = 0; 153 154 /* 155 * IP output. The packet in mbuf chain m contains a skeletal IP 156 * header (with len, off, ttl, proto, tos, src, dst). 157 * The mbuf chain containing the packet will be freed. 158 * The mbuf opt, if present, will not be freed. 159 */ 160 int 161 ip_output(struct mbuf *m0, ...) 162 { 163 struct rtentry *rt; 164 struct ip *ip; 165 struct ifnet *ifp; 166 struct mbuf *m = m0; 167 int hlen = sizeof (struct ip); 168 int len, error = 0; 169 struct route iproute; 170 const struct sockaddr_in *dst; 171 struct in_ifaddr *ia; 172 struct ifaddr *xifa; 173 struct mbuf *opt; 174 struct route *ro; 175 int flags, sw_csum; 176 int *mtu_p; 177 u_long mtu; 178 struct ip_moptions *imo; 179 struct socket *so; 180 va_list ap; 181 #ifdef IPSEC_NAT_T 182 int natt_frag = 0; 183 #endif 184 #ifdef FAST_IPSEC 185 struct inpcb *inp; 186 struct secpolicy *sp = NULL; 187 int s; 188 #endif 189 u_int16_t ip_len; 190 union { 191 struct sockaddr dst; 192 struct sockaddr_in dst4; 193 } u; 194 struct sockaddr *rdst = &u.dst; /* real IP destination, as opposed 195 * to the nexthop 196 */ 197 198 len = 0; 199 va_start(ap, m0); 200 opt = va_arg(ap, struct mbuf *); 201 ro = va_arg(ap, struct route *); 202 flags = va_arg(ap, int); 203 imo = va_arg(ap, struct ip_moptions *); 204 so = va_arg(ap, struct socket *); 205 if (flags & IP_RETURNMTU) 206 mtu_p = va_arg(ap, int *); 207 else 208 mtu_p = NULL; 209 va_end(ap); 210 211 MCLAIM(m, &ip_tx_mowner); 212 #ifdef FAST_IPSEC 213 if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET) 214 inp = (struct inpcb *)so->so_pcb; 215 else 216 inp = NULL; 217 #endif /* FAST_IPSEC */ 218 219 #ifdef DIAGNOSTIC 220 if ((m->m_flags & M_PKTHDR) == 0) 221 panic("ip_output: no HDR"); 222 223 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) != 0) { 224 panic("ip_output: IPv6 checksum offload flags: %d", 225 m->m_pkthdr.csum_flags); 226 } 227 228 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) == 229 (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 230 panic("ip_output: conflicting checksum offload flags: %d", 231 m->m_pkthdr.csum_flags); 232 } 233 #endif 234 if (opt) { 235 m = ip_insertoptions(m, opt, &len); 236 if (len >= sizeof(struct ip)) 237 hlen = len; 238 } 239 ip = mtod(m, struct ip *); 240 /* 241 * Fill in IP header. 242 */ 243 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 244 ip->ip_v = IPVERSION; 245 ip->ip_off = htons(0); 246 /* ip->ip_id filled in after we find out source ia */ 247 ip->ip_hl = hlen >> 2; 248 IP_STATINC(IP_STAT_LOCALOUT); 249 } else { 250 hlen = ip->ip_hl << 2; 251 } 252 /* 253 * Route packet. 254 */ 255 memset(&iproute, 0, sizeof(iproute)); 256 if (ro == NULL) 257 ro = &iproute; 258 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 259 dst = satocsin(rtcache_getdst(ro)); 260 /* 261 * If there is a cached route, 262 * check that it is to the same destination 263 * and is still up. If not, free it and try again. 264 * The address family should also be checked in case of sharing the 265 * cache with IPv6. 266 */ 267 if (dst == NULL) 268 ; 269 else if (dst->sin_family != AF_INET || 270 !in_hosteq(dst->sin_addr, ip->ip_dst)) 271 rtcache_free(ro); 272 273 if ((rt = rtcache_validate(ro)) == NULL && 274 (rt = rtcache_update(ro, 1)) == NULL) { 275 dst = &u.dst4; 276 rtcache_setdst(ro, &u.dst); 277 } 278 /* 279 * If routing to interface only, 280 * short circuit routing lookup. 281 */ 282 if (flags & IP_ROUTETOIF) { 283 if ((ia = ifatoia(ifa_ifwithladdr(sintocsa(dst)))) == NULL) { 284 IP_STATINC(IP_STAT_NOROUTE); 285 error = ENETUNREACH; 286 goto bad; 287 } 288 ifp = ia->ia_ifp; 289 mtu = ifp->if_mtu; 290 ip->ip_ttl = 1; 291 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 292 ip->ip_dst.s_addr == INADDR_BROADCAST) && 293 imo != NULL && imo->imo_multicast_ifp != NULL) { 294 ifp = imo->imo_multicast_ifp; 295 mtu = ifp->if_mtu; 296 IFP_TO_IA(ifp, ia); 297 } else { 298 if (rt == NULL) 299 rt = rtcache_init(ro); 300 if (rt == NULL) { 301 IP_STATINC(IP_STAT_NOROUTE); 302 error = EHOSTUNREACH; 303 goto bad; 304 } 305 ia = ifatoia(rt->rt_ifa); 306 ifp = rt->rt_ifp; 307 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 308 mtu = ifp->if_mtu; 309 rt->rt_use++; 310 if (rt->rt_flags & RTF_GATEWAY) 311 dst = satosin(rt->rt_gateway); 312 } 313 if (IN_MULTICAST(ip->ip_dst.s_addr) || 314 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 315 struct in_multi *inm; 316 317 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 318 M_BCAST : M_MCAST; 319 /* 320 * See if the caller provided any multicast options 321 */ 322 if (imo != NULL) 323 ip->ip_ttl = imo->imo_multicast_ttl; 324 else 325 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 326 327 /* 328 * if we don't know the outgoing ifp yet, we can't generate 329 * output 330 */ 331 if (!ifp) { 332 IP_STATINC(IP_STAT_NOROUTE); 333 error = ENETUNREACH; 334 goto bad; 335 } 336 337 /* 338 * If the packet is multicast or broadcast, confirm that 339 * the outgoing interface can transmit it. 340 */ 341 if (((m->m_flags & M_MCAST) && 342 (ifp->if_flags & IFF_MULTICAST) == 0) || 343 ((m->m_flags & M_BCAST) && 344 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 345 IP_STATINC(IP_STAT_NOROUTE); 346 error = ENETUNREACH; 347 goto bad; 348 } 349 /* 350 * If source address not specified yet, use an address 351 * of outgoing interface. 352 */ 353 if (in_nullhost(ip->ip_src)) { 354 struct in_ifaddr *xia; 355 356 IFP_TO_IA(ifp, xia); 357 if (!xia) { 358 error = EADDRNOTAVAIL; 359 goto bad; 360 } 361 xifa = &xia->ia_ifa; 362 if (xifa->ifa_getifa != NULL) { 363 xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 364 } 365 ip->ip_src = xia->ia_addr.sin_addr; 366 } 367 368 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 369 if (inm != NULL && 370 (imo == NULL || imo->imo_multicast_loop)) { 371 /* 372 * If we belong to the destination multicast group 373 * on the outgoing interface, and the caller did not 374 * forbid loopback, loop back a copy. 375 */ 376 ip_mloopback(ifp, m, &u.dst4); 377 } 378 #ifdef MROUTING 379 else { 380 /* 381 * If we are acting as a multicast router, perform 382 * multicast forwarding as if the packet had just 383 * arrived on the interface to which we are about 384 * to send. The multicast forwarding function 385 * recursively calls this function, using the 386 * IP_FORWARDING flag to prevent infinite recursion. 387 * 388 * Multicasts that are looped back by ip_mloopback(), 389 * above, will be forwarded by the ip_input() routine, 390 * if necessary. 391 */ 392 extern struct socket *ip_mrouter; 393 394 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 395 if (ip_mforward(m, ifp) != 0) { 396 m_freem(m); 397 goto done; 398 } 399 } 400 } 401 #endif 402 /* 403 * Multicasts with a time-to-live of zero may be looped- 404 * back, above, but must not be transmitted on a network. 405 * Also, multicasts addressed to the loopback interface 406 * are not sent -- the above call to ip_mloopback() will 407 * loop back a copy if this host actually belongs to the 408 * destination group on the loopback interface. 409 */ 410 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 411 m_freem(m); 412 goto done; 413 } 414 415 goto sendit; 416 } 417 /* 418 * If source address not specified yet, use address 419 * of outgoing interface. 420 */ 421 if (in_nullhost(ip->ip_src)) { 422 xifa = &ia->ia_ifa; 423 if (xifa->ifa_getifa != NULL) 424 ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 425 ip->ip_src = ia->ia_addr.sin_addr; 426 } 427 428 /* 429 * packets with Class-D address as source are not valid per 430 * RFC 1112 431 */ 432 if (IN_MULTICAST(ip->ip_src.s_addr)) { 433 IP_STATINC(IP_STAT_ODROPPED); 434 error = EADDRNOTAVAIL; 435 goto bad; 436 } 437 438 /* 439 * Look for broadcast address and 440 * and verify user is allowed to send 441 * such a packet. 442 */ 443 if (in_broadcast(dst->sin_addr, ifp)) { 444 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 445 error = EADDRNOTAVAIL; 446 goto bad; 447 } 448 if ((flags & IP_ALLOWBROADCAST) == 0) { 449 error = EACCES; 450 goto bad; 451 } 452 /* don't allow broadcast messages to be fragmented */ 453 if (ntohs(ip->ip_len) > ifp->if_mtu) { 454 error = EMSGSIZE; 455 goto bad; 456 } 457 m->m_flags |= M_BCAST; 458 } else 459 m->m_flags &= ~M_BCAST; 460 461 sendit: 462 if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) { 463 if (m->m_pkthdr.len < IP_MINFRAGSIZE) { 464 ip->ip_id = 0; 465 } else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 466 ip->ip_id = ip_newid(ia); 467 } else { 468 469 /* 470 * TSO capable interfaces (typically?) increment 471 * ip_id for each segment. 472 * "allocate" enough ids here to increase the chance 473 * for them to be unique. 474 * 475 * note that the following calculation is not 476 * needed to be precise. wasting some ip_id is fine. 477 */ 478 479 unsigned int segsz = m->m_pkthdr.segsz; 480 unsigned int datasz = ntohs(ip->ip_len) - hlen; 481 unsigned int num = howmany(datasz, segsz); 482 483 ip->ip_id = ip_newid_range(ia, num); 484 } 485 } 486 /* 487 * If we're doing Path MTU Discovery, we need to set DF unless 488 * the route's MTU is locked. 489 */ 490 if ((flags & IP_MTUDISC) != 0 && rt != NULL && 491 (rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 492 ip->ip_off |= htons(IP_DF); 493 494 /* Remember the current ip_len */ 495 ip_len = ntohs(ip->ip_len); 496 497 #ifdef FAST_IPSEC 498 /* 499 * Check the security policy (SP) for the packet and, if 500 * required, do IPsec-related processing. There are two 501 * cases here; the first time a packet is sent through 502 * it will be untagged and handled by ipsec4_checkpolicy. 503 * If the packet is resubmitted to ip_output (e.g. after 504 * AH, ESP, etc. processing), there will be a tag to bypass 505 * the lookup and related policy checking. 506 */ 507 if (!ipsec_outdone(m)) { 508 s = splsoftnet(); 509 if (inp != NULL && 510 IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND)) { 511 splx(s); 512 goto spd_done; 513 } 514 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 515 &error, inp); 516 /* 517 * There are four return cases: 518 * sp != NULL apply IPsec policy 519 * sp == NULL, error == 0 no IPsec handling needed 520 * sp == NULL, error == -EINVAL discard packet w/o error 521 * sp == NULL, error != 0 discard packet, report error 522 */ 523 if (sp != NULL) { 524 #ifdef IPSEC_NAT_T 525 /* 526 * NAT-T ESP fragmentation: don't do IPSec processing now, 527 * we'll do it on each fragmented packet. 528 */ 529 if (sp->req->sav && 530 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 531 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 532 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 533 natt_frag = 1; 534 mtu = sp->req->sav->esp_frag; 535 splx(s); 536 goto spd_done; 537 } 538 } 539 #endif /* IPSEC_NAT_T */ 540 541 /* 542 * Do delayed checksums now because we send before 543 * this is done in the normal processing path. 544 */ 545 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 546 in_delayed_cksum(m); 547 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 548 } 549 550 #ifdef __FreeBSD__ 551 ip->ip_len = htons(ip->ip_len); 552 ip->ip_off = htons(ip->ip_off); 553 #endif 554 555 /* NB: callee frees mbuf */ 556 error = ipsec4_process_packet(m, sp->req, flags, 0); 557 /* 558 * Preserve KAME behaviour: ENOENT can be returned 559 * when an SA acquire is in progress. Don't propagate 560 * this to user-level; it confuses applications. 561 * 562 * XXX this will go away when the SADB is redone. 563 */ 564 if (error == ENOENT) 565 error = 0; 566 splx(s); 567 goto done; 568 } else { 569 splx(s); 570 571 if (error != 0) { 572 /* 573 * Hack: -EINVAL is used to signal that a packet 574 * should be silently discarded. This is typically 575 * because we asked key management for an SA and 576 * it was delayed (e.g. kicked up to IKE). 577 */ 578 if (error == -EINVAL) 579 error = 0; 580 goto bad; 581 } else { 582 /* No IPsec processing for this packet. */ 583 } 584 } 585 } 586 spd_done: 587 #endif /* FAST_IPSEC */ 588 589 #ifdef PFIL_HOOKS 590 /* 591 * Run through list of hooks for output packets. 592 */ 593 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 594 goto done; 595 if (m == NULL) 596 goto done; 597 598 ip = mtod(m, struct ip *); 599 hlen = ip->ip_hl << 2; 600 ip_len = ntohs(ip->ip_len); 601 #endif /* PFIL_HOOKS */ 602 603 m->m_pkthdr.csum_data |= hlen << 16; 604 605 #if IFA_STATS 606 /* 607 * search for the source address structure to 608 * maintain output statistics. 609 */ 610 INADDR_TO_IA(ip->ip_src, ia); 611 #endif 612 613 /* Maybe skip checksums on loopback interfaces. */ 614 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 615 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 616 } 617 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 618 /* 619 * If small enough for mtu of path, or if using TCP segmentation 620 * offload, can just send directly. 621 */ 622 if (ip_len <= mtu || 623 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 624 #if IFA_STATS 625 if (ia) 626 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 627 #endif 628 /* 629 * Always initialize the sum to 0! Some HW assisted 630 * checksumming requires this. 631 */ 632 ip->ip_sum = 0; 633 634 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 635 /* 636 * Perform any checksums that the hardware can't do 637 * for us. 638 * 639 * XXX Does any hardware require the {th,uh}_sum 640 * XXX fields to be 0? 641 */ 642 if (sw_csum & M_CSUM_IPv4) { 643 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 644 ip->ip_sum = in_cksum(m, hlen); 645 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 646 } 647 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 648 if (IN_NEED_CHECKSUM(ifp, 649 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 650 in_delayed_cksum(m); 651 } 652 m->m_pkthdr.csum_flags &= 653 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 654 } 655 } 656 657 if (__predict_true( 658 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 || 659 (ifp->if_capenable & IFCAP_TSOv4) != 0)) { 660 KERNEL_LOCK(1, NULL); 661 error = 662 (*ifp->if_output)(ifp, m, 663 (m->m_flags & M_MCAST) ? 664 sintocsa(rdst) : sintocsa(dst), 665 rt); 666 KERNEL_UNLOCK_ONE(NULL); 667 } else { 668 error = 669 ip_tso_output(ifp, m, 670 (m->m_flags & M_MCAST) ? 671 sintocsa(rdst) : sintocsa(dst), 672 rt); 673 } 674 goto done; 675 } 676 677 /* 678 * We can't use HW checksumming if we're about to 679 * to fragment the packet. 680 * 681 * XXX Some hardware can do this. 682 */ 683 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 684 if (IN_NEED_CHECKSUM(ifp, 685 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 686 in_delayed_cksum(m); 687 } 688 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 689 } 690 691 /* 692 * Too large for interface; fragment if possible. 693 * Must be able to put at least 8 bytes per fragment. 694 */ 695 if (ntohs(ip->ip_off) & IP_DF) { 696 if (flags & IP_RETURNMTU) 697 *mtu_p = mtu; 698 error = EMSGSIZE; 699 IP_STATINC(IP_STAT_CANTFRAG); 700 goto bad; 701 } 702 703 error = ip_fragment(m, ifp, mtu); 704 if (error) { 705 m = NULL; 706 goto bad; 707 } 708 709 for (; m; m = m0) { 710 m0 = m->m_nextpkt; 711 m->m_nextpkt = 0; 712 if (error == 0) { 713 #if IFA_STATS 714 if (ia) 715 ia->ia_ifa.ifa_data.ifad_outbytes += 716 ntohs(ip->ip_len); 717 #endif 718 #ifdef IPSEC_NAT_T 719 /* 720 * If we get there, the packet has not been handeld by 721 * IPSec whereas it should have. Now that it has been 722 * fragmented, re-inject it in ip_output so that IPsec 723 * processing can occur. 724 */ 725 if (natt_frag) { 726 error = ip_output(m, opt, 727 ro, flags | IP_RAWOUTPUT | IP_NOIPNEWID, imo, so, mtu_p); 728 } else 729 #endif /* IPSEC_NAT_T */ 730 { 731 KASSERT((m->m_pkthdr.csum_flags & 732 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 733 KERNEL_LOCK(1, NULL); 734 error = (*ifp->if_output)(ifp, m, 735 (m->m_flags & M_MCAST) ? 736 sintocsa(rdst) : sintocsa(dst), 737 rt); 738 KERNEL_UNLOCK_ONE(NULL); 739 } 740 } else 741 m_freem(m); 742 } 743 744 if (error == 0) 745 IP_STATINC(IP_STAT_FRAGMENTED); 746 done: 747 rtcache_free(&iproute); 748 749 #ifdef FAST_IPSEC 750 if (sp != NULL) 751 KEY_FREESP(&sp); 752 #endif /* FAST_IPSEC */ 753 754 return (error); 755 bad: 756 m_freem(m); 757 goto done; 758 } 759 760 int 761 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 762 { 763 struct ip *ip, *mhip; 764 struct mbuf *m0; 765 int len, hlen, off; 766 int mhlen, firstlen; 767 struct mbuf **mnext; 768 int sw_csum = m->m_pkthdr.csum_flags; 769 int fragments = 0; 770 int s; 771 int error = 0; 772 773 ip = mtod(m, struct ip *); 774 hlen = ip->ip_hl << 2; 775 if (ifp != NULL) 776 sw_csum &= ~ifp->if_csum_flags_tx; 777 778 len = (mtu - hlen) &~ 7; 779 if (len < 8) { 780 m_freem(m); 781 return (EMSGSIZE); 782 } 783 784 firstlen = len; 785 mnext = &m->m_nextpkt; 786 787 /* 788 * Loop through length of segment after first fragment, 789 * make new header and copy data of each part and link onto chain. 790 */ 791 m0 = m; 792 mhlen = sizeof (struct ip); 793 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 794 MGETHDR(m, M_DONTWAIT, MT_HEADER); 795 if (m == 0) { 796 error = ENOBUFS; 797 IP_STATINC(IP_STAT_ODROPPED); 798 goto sendorfree; 799 } 800 MCLAIM(m, m0->m_owner); 801 *mnext = m; 802 mnext = &m->m_nextpkt; 803 m->m_data += max_linkhdr; 804 mhip = mtod(m, struct ip *); 805 *mhip = *ip; 806 /* we must inherit MCAST and BCAST flags */ 807 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 808 if (hlen > sizeof (struct ip)) { 809 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 810 mhip->ip_hl = mhlen >> 2; 811 } 812 m->m_len = mhlen; 813 mhip->ip_off = ((off - hlen) >> 3) + 814 (ntohs(ip->ip_off) & ~IP_MF); 815 if (ip->ip_off & htons(IP_MF)) 816 mhip->ip_off |= IP_MF; 817 if (off + len >= ntohs(ip->ip_len)) 818 len = ntohs(ip->ip_len) - off; 819 else 820 mhip->ip_off |= IP_MF; 821 HTONS(mhip->ip_off); 822 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 823 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 824 if (m->m_next == 0) { 825 error = ENOBUFS; /* ??? */ 826 IP_STATINC(IP_STAT_ODROPPED); 827 goto sendorfree; 828 } 829 m->m_pkthdr.len = mhlen + len; 830 m->m_pkthdr.rcvif = NULL; 831 mhip->ip_sum = 0; 832 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 833 if (sw_csum & M_CSUM_IPv4) { 834 mhip->ip_sum = in_cksum(m, mhlen); 835 } else { 836 /* 837 * checksum is hw-offloaded or not necessary. 838 */ 839 m->m_pkthdr.csum_flags |= 840 m0->m_pkthdr.csum_flags & M_CSUM_IPv4; 841 m->m_pkthdr.csum_data |= mhlen << 16; 842 KASSERT(!(ifp != NULL && 843 IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) 844 || (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 845 } 846 IP_STATINC(IP_STAT_OFRAGMENTS); 847 fragments++; 848 } 849 /* 850 * Update first fragment by trimming what's been copied out 851 * and updating header, then send each fragment (in order). 852 */ 853 m = m0; 854 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 855 m->m_pkthdr.len = hlen + firstlen; 856 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 857 ip->ip_off |= htons(IP_MF); 858 ip->ip_sum = 0; 859 if (sw_csum & M_CSUM_IPv4) { 860 ip->ip_sum = in_cksum(m, hlen); 861 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 862 } else { 863 /* 864 * checksum is hw-offloaded or not necessary. 865 */ 866 KASSERT(!(ifp != NULL && IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) 867 || (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 868 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 869 sizeof(struct ip)); 870 } 871 sendorfree: 872 /* 873 * If there is no room for all the fragments, don't queue 874 * any of them. 875 */ 876 if (ifp != NULL) { 877 s = splnet(); 878 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 879 error == 0) { 880 error = ENOBUFS; 881 IP_STATINC(IP_STAT_ODROPPED); 882 IFQ_INC_DROPS(&ifp->if_snd); 883 } 884 splx(s); 885 } 886 if (error) { 887 for (m = m0; m; m = m0) { 888 m0 = m->m_nextpkt; 889 m->m_nextpkt = NULL; 890 m_freem(m); 891 } 892 } 893 return (error); 894 } 895 896 /* 897 * Process a delayed payload checksum calculation. 898 */ 899 void 900 in_delayed_cksum(struct mbuf *m) 901 { 902 struct ip *ip; 903 u_int16_t csum, offset; 904 905 ip = mtod(m, struct ip *); 906 offset = ip->ip_hl << 2; 907 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 908 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 909 csum = 0xffff; 910 911 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 912 913 if ((offset + sizeof(u_int16_t)) > m->m_len) { 914 /* This happen when ip options were inserted 915 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 916 m->m_len, offset, ip->ip_p); 917 */ 918 m_copyback(m, offset, sizeof(csum), (void *) &csum); 919 } else 920 *(u_int16_t *)(mtod(m, char *) + offset) = csum; 921 } 922 923 /* 924 * Determine the maximum length of the options to be inserted; 925 * we would far rather allocate too much space rather than too little. 926 */ 927 928 u_int 929 ip_optlen(struct inpcb *inp) 930 { 931 struct mbuf *m = inp->inp_options; 932 933 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 934 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 935 else 936 return 0; 937 } 938 939 940 /* 941 * Insert IP options into preformed packet. 942 * Adjust IP destination as required for IP source routing, 943 * as indicated by a non-zero in_addr at the start of the options. 944 */ 945 static struct mbuf * 946 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 947 { 948 struct ipoption *p = mtod(opt, struct ipoption *); 949 struct mbuf *n; 950 struct ip *ip = mtod(m, struct ip *); 951 unsigned optlen; 952 953 optlen = opt->m_len - sizeof(p->ipopt_dst); 954 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 955 return (m); /* XXX should fail */ 956 if (!in_nullhost(p->ipopt_dst)) 957 ip->ip_dst = p->ipopt_dst; 958 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 959 MGETHDR(n, M_DONTWAIT, MT_HEADER); 960 if (n == 0) 961 return (m); 962 MCLAIM(n, m->m_owner); 963 M_MOVE_PKTHDR(n, m); 964 m->m_len -= sizeof(struct ip); 965 m->m_data += sizeof(struct ip); 966 n->m_next = m; 967 m = n; 968 m->m_len = optlen + sizeof(struct ip); 969 m->m_data += max_linkhdr; 970 bcopy((void *)ip, mtod(m, void *), sizeof(struct ip)); 971 } else { 972 m->m_data -= optlen; 973 m->m_len += optlen; 974 memmove(mtod(m, void *), ip, sizeof(struct ip)); 975 } 976 m->m_pkthdr.len += optlen; 977 ip = mtod(m, struct ip *); 978 bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen); 979 *phlen = sizeof(struct ip) + optlen; 980 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 981 return (m); 982 } 983 984 /* 985 * Copy options from ip to jp, 986 * omitting those not copied during fragmentation. 987 */ 988 int 989 ip_optcopy(struct ip *ip, struct ip *jp) 990 { 991 u_char *cp, *dp; 992 int opt, optlen, cnt; 993 994 cp = (u_char *)(ip + 1); 995 dp = (u_char *)(jp + 1); 996 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 997 for (; cnt > 0; cnt -= optlen, cp += optlen) { 998 opt = cp[0]; 999 if (opt == IPOPT_EOL) 1000 break; 1001 if (opt == IPOPT_NOP) { 1002 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1003 *dp++ = IPOPT_NOP; 1004 optlen = 1; 1005 continue; 1006 } 1007 #ifdef DIAGNOSTIC 1008 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1009 panic("malformed IPv4 option passed to ip_optcopy"); 1010 #endif 1011 optlen = cp[IPOPT_OLEN]; 1012 #ifdef DIAGNOSTIC 1013 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1014 panic("malformed IPv4 option passed to ip_optcopy"); 1015 #endif 1016 /* bogus lengths should have been caught by ip_dooptions */ 1017 if (optlen > cnt) 1018 optlen = cnt; 1019 if (IPOPT_COPIED(opt)) { 1020 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 1021 dp += optlen; 1022 } 1023 } 1024 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1025 *dp++ = IPOPT_EOL; 1026 return (optlen); 1027 } 1028 1029 /* 1030 * IP socket option processing. 1031 */ 1032 int 1033 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1034 { 1035 struct inpcb *inp = sotoinpcb(so); 1036 int optval = 0; 1037 int error = 0; 1038 #if defined(FAST_IPSEC) 1039 struct lwp *l = curlwp; /*XXX*/ 1040 #endif 1041 1042 if (sopt->sopt_level != IPPROTO_IP) { 1043 if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER) 1044 return 0; 1045 return ENOPROTOOPT; 1046 } 1047 1048 switch (op) { 1049 case PRCO_SETOPT: 1050 switch (sopt->sopt_name) { 1051 case IP_OPTIONS: 1052 #ifdef notyet 1053 case IP_RETOPTS: 1054 #endif 1055 error = ip_pcbopts(&inp->inp_options, sopt); 1056 break; 1057 1058 case IP_TOS: 1059 case IP_TTL: 1060 case IP_MINTTL: 1061 case IP_RECVOPTS: 1062 case IP_RECVRETOPTS: 1063 case IP_RECVDSTADDR: 1064 case IP_RECVIF: 1065 case IP_RECVTTL: 1066 error = sockopt_getint(sopt, &optval); 1067 if (error) 1068 break; 1069 1070 switch (sopt->sopt_name) { 1071 case IP_TOS: 1072 inp->inp_ip.ip_tos = optval; 1073 break; 1074 1075 case IP_TTL: 1076 inp->inp_ip.ip_ttl = optval; 1077 break; 1078 1079 case IP_MINTTL: 1080 if (optval > 0 && optval <= MAXTTL) 1081 inp->inp_ip_minttl = optval; 1082 else 1083 error = EINVAL; 1084 break; 1085 #define OPTSET(bit) \ 1086 if (optval) \ 1087 inp->inp_flags |= bit; \ 1088 else \ 1089 inp->inp_flags &= ~bit; 1090 1091 case IP_RECVOPTS: 1092 OPTSET(INP_RECVOPTS); 1093 break; 1094 1095 case IP_RECVRETOPTS: 1096 OPTSET(INP_RECVRETOPTS); 1097 break; 1098 1099 case IP_RECVDSTADDR: 1100 OPTSET(INP_RECVDSTADDR); 1101 break; 1102 1103 case IP_RECVIF: 1104 OPTSET(INP_RECVIF); 1105 break; 1106 1107 case IP_RECVTTL: 1108 OPTSET(INP_RECVTTL); 1109 break; 1110 } 1111 break; 1112 #undef OPTSET 1113 1114 case IP_MULTICAST_IF: 1115 case IP_MULTICAST_TTL: 1116 case IP_MULTICAST_LOOP: 1117 case IP_ADD_MEMBERSHIP: 1118 case IP_DROP_MEMBERSHIP: 1119 error = ip_setmoptions(&inp->inp_moptions, sopt); 1120 break; 1121 1122 case IP_PORTRANGE: 1123 error = sockopt_getint(sopt, &optval); 1124 if (error) 1125 break; 1126 1127 /* INP_LOCK(inp); */ 1128 switch (optval) { 1129 case IP_PORTRANGE_DEFAULT: 1130 case IP_PORTRANGE_HIGH: 1131 inp->inp_flags &= ~(INP_LOWPORT); 1132 break; 1133 1134 case IP_PORTRANGE_LOW: 1135 inp->inp_flags |= INP_LOWPORT; 1136 break; 1137 1138 default: 1139 error = EINVAL; 1140 break; 1141 } 1142 /* INP_UNLOCK(inp); */ 1143 break; 1144 1145 #if defined(FAST_IPSEC) 1146 case IP_IPSEC_POLICY: 1147 { 1148 error = ipsec4_set_policy(inp, sopt->sopt_name, 1149 sopt->sopt_data, sopt->sopt_size, l->l_cred); 1150 break; 1151 } 1152 #endif /*IPSEC*/ 1153 1154 default: 1155 error = ENOPROTOOPT; 1156 break; 1157 } 1158 break; 1159 1160 case PRCO_GETOPT: 1161 switch (sopt->sopt_name) { 1162 case IP_OPTIONS: 1163 case IP_RETOPTS: 1164 if (inp->inp_options) { 1165 struct mbuf *m; 1166 1167 m = m_copym(inp->inp_options, 0, M_COPYALL, 1168 M_DONTWAIT); 1169 if (m == NULL) { 1170 error = ENOBUFS; 1171 break; 1172 } 1173 1174 error = sockopt_setmbuf(sopt, m); 1175 } 1176 break; 1177 1178 case IP_TOS: 1179 case IP_TTL: 1180 case IP_MINTTL: 1181 case IP_RECVOPTS: 1182 case IP_RECVRETOPTS: 1183 case IP_RECVDSTADDR: 1184 case IP_RECVIF: 1185 case IP_RECVTTL: 1186 case IP_ERRORMTU: 1187 switch (sopt->sopt_name) { 1188 case IP_TOS: 1189 optval = inp->inp_ip.ip_tos; 1190 break; 1191 1192 case IP_TTL: 1193 optval = inp->inp_ip.ip_ttl; 1194 break; 1195 1196 case IP_MINTTL: 1197 optval = inp->inp_ip_minttl; 1198 break; 1199 1200 case IP_ERRORMTU: 1201 optval = inp->inp_errormtu; 1202 break; 1203 1204 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1205 1206 case IP_RECVOPTS: 1207 optval = OPTBIT(INP_RECVOPTS); 1208 break; 1209 1210 case IP_RECVRETOPTS: 1211 optval = OPTBIT(INP_RECVRETOPTS); 1212 break; 1213 1214 case IP_RECVDSTADDR: 1215 optval = OPTBIT(INP_RECVDSTADDR); 1216 break; 1217 1218 case IP_RECVIF: 1219 optval = OPTBIT(INP_RECVIF); 1220 break; 1221 1222 case IP_RECVTTL: 1223 optval = OPTBIT(INP_RECVTTL); 1224 break; 1225 } 1226 error = sockopt_setint(sopt, optval); 1227 break; 1228 1229 #if 0 /* defined(FAST_IPSEC) */ 1230 case IP_IPSEC_POLICY: 1231 { 1232 struct mbuf *m = NULL; 1233 1234 /* XXX this will return EINVAL as sopt is empty */ 1235 error = ipsec4_get_policy(inp, sopt->sopt_data, 1236 sopt->sopt_size, &m); 1237 if (error == 0) 1238 error = sockopt_setmbuf(sopt, m); 1239 break; 1240 } 1241 #endif /*IPSEC*/ 1242 1243 case IP_MULTICAST_IF: 1244 case IP_MULTICAST_TTL: 1245 case IP_MULTICAST_LOOP: 1246 case IP_ADD_MEMBERSHIP: 1247 case IP_DROP_MEMBERSHIP: 1248 error = ip_getmoptions(inp->inp_moptions, sopt); 1249 break; 1250 1251 case IP_PORTRANGE: 1252 if (inp->inp_flags & INP_LOWPORT) 1253 optval = IP_PORTRANGE_LOW; 1254 else 1255 optval = IP_PORTRANGE_DEFAULT; 1256 1257 error = sockopt_setint(sopt, optval); 1258 1259 break; 1260 1261 default: 1262 error = ENOPROTOOPT; 1263 break; 1264 } 1265 break; 1266 } 1267 return (error); 1268 } 1269 1270 /* 1271 * Set up IP options in pcb for insertion in output packets. 1272 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1273 * with destination address if source routed. 1274 */ 1275 int 1276 ip_pcbopts(struct mbuf **pcbopt, const struct sockopt *sopt) 1277 { 1278 struct mbuf *m; 1279 const u_char *cp; 1280 u_char *dp; 1281 int cnt; 1282 uint8_t optval, olen, offset; 1283 1284 /* turn off any old options */ 1285 if (*pcbopt) 1286 (void)m_free(*pcbopt); 1287 *pcbopt = NULL; 1288 1289 cp = sopt->sopt_data; 1290 cnt = sopt->sopt_size; 1291 1292 if (cnt == 0) 1293 return (0); /* Only turning off any previous options */ 1294 1295 #ifndef __vax__ 1296 if (cnt % sizeof(int32_t)) 1297 return (EINVAL); 1298 #endif 1299 1300 m = m_get(M_DONTWAIT, MT_SOOPTS); 1301 if (m == NULL) 1302 return (ENOBUFS); 1303 1304 dp = mtod(m, u_char *); 1305 memset(dp, 0, sizeof(struct in_addr)); 1306 dp += sizeof(struct in_addr); 1307 m->m_len = sizeof(struct in_addr); 1308 1309 /* 1310 * IP option list according to RFC791. Each option is of the form 1311 * 1312 * [optval] [olen] [(olen - 2) data bytes] 1313 * 1314 * we validate the list and copy options to an mbuf for prepending 1315 * to data packets. The IP first-hop destination address will be 1316 * stored before actual options and is zero if unset. 1317 */ 1318 while (cnt > 0) { 1319 optval = cp[IPOPT_OPTVAL]; 1320 1321 if (optval == IPOPT_EOL || optval == IPOPT_NOP) { 1322 olen = 1; 1323 } else { 1324 if (cnt < IPOPT_OLEN + 1) 1325 goto bad; 1326 1327 olen = cp[IPOPT_OLEN]; 1328 if (olen < IPOPT_OLEN + 1 || olen > cnt) 1329 goto bad; 1330 } 1331 1332 if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) { 1333 /* 1334 * user process specifies route as: 1335 * ->A->B->C->D 1336 * D must be our final destination (but we can't 1337 * check that since we may not have connected yet). 1338 * A is first hop destination, which doesn't appear in 1339 * actual IP option, but is stored before the options. 1340 */ 1341 if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr)) 1342 goto bad; 1343 1344 offset = cp[IPOPT_OFFSET]; 1345 memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1, 1346 sizeof(struct in_addr)); 1347 1348 cp += sizeof(struct in_addr); 1349 cnt -= sizeof(struct in_addr); 1350 olen -= sizeof(struct in_addr); 1351 1352 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1353 goto bad; 1354 1355 memcpy(dp, cp, olen); 1356 dp[IPOPT_OPTVAL] = optval; 1357 dp[IPOPT_OLEN] = olen; 1358 dp[IPOPT_OFFSET] = offset; 1359 break; 1360 } else { 1361 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1362 goto bad; 1363 1364 memcpy(dp, cp, olen); 1365 break; 1366 } 1367 1368 dp += olen; 1369 m->m_len += olen; 1370 1371 if (optval == IPOPT_EOL) 1372 break; 1373 1374 cp += olen; 1375 cnt -= olen; 1376 } 1377 1378 *pcbopt = m; 1379 return (0); 1380 1381 bad: 1382 (void)m_free(m); 1383 return (EINVAL); 1384 } 1385 1386 /* 1387 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1388 */ 1389 static struct ifnet * 1390 ip_multicast_if(struct in_addr *a, int *ifindexp) 1391 { 1392 int ifindex; 1393 struct ifnet *ifp = NULL; 1394 struct in_ifaddr *ia; 1395 1396 if (ifindexp) 1397 *ifindexp = 0; 1398 if (ntohl(a->s_addr) >> 24 == 0) { 1399 ifindex = ntohl(a->s_addr) & 0xffffff; 1400 if (ifindex < 0 || if_indexlim <= ifindex) 1401 return NULL; 1402 ifp = ifindex2ifnet[ifindex]; 1403 if (!ifp) 1404 return NULL; 1405 if (ifindexp) 1406 *ifindexp = ifindex; 1407 } else { 1408 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1409 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1410 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1411 ifp = ia->ia_ifp; 1412 break; 1413 } 1414 } 1415 } 1416 return ifp; 1417 } 1418 1419 static int 1420 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval) 1421 { 1422 u_int tval; 1423 u_char cval; 1424 int error; 1425 1426 if (sopt == NULL) 1427 return EINVAL; 1428 1429 switch (sopt->sopt_size) { 1430 case sizeof(u_char): 1431 error = sockopt_get(sopt, &cval, sizeof(u_char)); 1432 tval = cval; 1433 break; 1434 1435 case sizeof(u_int): 1436 error = sockopt_get(sopt, &tval, sizeof(u_int)); 1437 break; 1438 1439 default: 1440 error = EINVAL; 1441 } 1442 1443 if (error) 1444 return error; 1445 1446 if (tval > maxval) 1447 return EINVAL; 1448 1449 *val = tval; 1450 return 0; 1451 } 1452 1453 /* 1454 * Set the IP multicast options in response to user setsockopt(). 1455 */ 1456 int 1457 ip_setmoptions(struct ip_moptions **imop, const struct sockopt *sopt) 1458 { 1459 struct in_addr addr; 1460 struct ip_mreq lmreq, *mreq; 1461 struct ifnet *ifp; 1462 struct ip_moptions *imo = *imop; 1463 int i, ifindex, error = 0; 1464 1465 if (imo == NULL) { 1466 /* 1467 * No multicast option buffer attached to the pcb; 1468 * allocate one and initialize to default values. 1469 */ 1470 imo = kmem_intr_alloc(sizeof(*imo), KM_NOSLEEP); 1471 if (imo == NULL) 1472 return ENOBUFS; 1473 1474 imo->imo_multicast_ifp = NULL; 1475 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1476 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1477 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1478 imo->imo_num_memberships = 0; 1479 *imop = imo; 1480 } 1481 1482 switch (sopt->sopt_name) { 1483 case IP_MULTICAST_IF: 1484 /* 1485 * Select the interface for outgoing multicast packets. 1486 */ 1487 error = sockopt_get(sopt, &addr, sizeof(addr)); 1488 if (error) 1489 break; 1490 1491 /* 1492 * INADDR_ANY is used to remove a previous selection. 1493 * When no interface is selected, a default one is 1494 * chosen every time a multicast packet is sent. 1495 */ 1496 if (in_nullhost(addr)) { 1497 imo->imo_multicast_ifp = NULL; 1498 break; 1499 } 1500 /* 1501 * The selected interface is identified by its local 1502 * IP address. Find the interface and confirm that 1503 * it supports multicasting. 1504 */ 1505 ifp = ip_multicast_if(&addr, &ifindex); 1506 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1507 error = EADDRNOTAVAIL; 1508 break; 1509 } 1510 imo->imo_multicast_ifp = ifp; 1511 if (ifindex) 1512 imo->imo_multicast_addr = addr; 1513 else 1514 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1515 break; 1516 1517 case IP_MULTICAST_TTL: 1518 /* 1519 * Set the IP time-to-live for outgoing multicast packets. 1520 */ 1521 error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL); 1522 break; 1523 1524 case IP_MULTICAST_LOOP: 1525 /* 1526 * Set the loopback flag for outgoing multicast packets. 1527 * Must be zero or one. 1528 */ 1529 error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1); 1530 break; 1531 1532 case IP_ADD_MEMBERSHIP: 1533 /* 1534 * Add a multicast group membership. 1535 * Group must be a valid IP multicast address. 1536 */ 1537 error = sockopt_get(sopt, &lmreq, sizeof(lmreq)); 1538 if (error) 1539 break; 1540 1541 mreq = &lmreq; 1542 1543 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1544 error = EINVAL; 1545 break; 1546 } 1547 /* 1548 * If no interface address was provided, use the interface of 1549 * the route to the given multicast address. 1550 */ 1551 if (in_nullhost(mreq->imr_interface)) { 1552 struct rtentry *rt; 1553 union { 1554 struct sockaddr dst; 1555 struct sockaddr_in dst4; 1556 } u; 1557 struct route ro; 1558 1559 memset(&ro, 0, sizeof(ro)); 1560 1561 sockaddr_in_init(&u.dst4, &mreq->imr_multiaddr, 0); 1562 rtcache_setdst(&ro, &u.dst); 1563 ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp 1564 : NULL; 1565 rtcache_free(&ro); 1566 } else { 1567 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1568 } 1569 /* 1570 * See if we found an interface, and confirm that it 1571 * supports multicast. 1572 */ 1573 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1574 error = EADDRNOTAVAIL; 1575 break; 1576 } 1577 /* 1578 * See if the membership already exists or if all the 1579 * membership slots are full. 1580 */ 1581 for (i = 0; i < imo->imo_num_memberships; ++i) { 1582 if (imo->imo_membership[i]->inm_ifp == ifp && 1583 in_hosteq(imo->imo_membership[i]->inm_addr, 1584 mreq->imr_multiaddr)) 1585 break; 1586 } 1587 if (i < imo->imo_num_memberships) { 1588 error = EADDRINUSE; 1589 break; 1590 } 1591 if (i == IP_MAX_MEMBERSHIPS) { 1592 error = ETOOMANYREFS; 1593 break; 1594 } 1595 /* 1596 * Everything looks good; add a new record to the multicast 1597 * address list for the given interface. 1598 */ 1599 if ((imo->imo_membership[i] = 1600 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1601 error = ENOBUFS; 1602 break; 1603 } 1604 ++imo->imo_num_memberships; 1605 break; 1606 1607 case IP_DROP_MEMBERSHIP: 1608 /* 1609 * Drop a multicast group membership. 1610 * Group must be a valid IP multicast address. 1611 */ 1612 error = sockopt_get(sopt, &lmreq, sizeof(lmreq)); 1613 if (error) 1614 break; 1615 1616 mreq = &lmreq; 1617 1618 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1619 error = EINVAL; 1620 break; 1621 } 1622 /* 1623 * If an interface address was specified, get a pointer 1624 * to its ifnet structure. 1625 */ 1626 if (in_nullhost(mreq->imr_interface)) 1627 ifp = NULL; 1628 else { 1629 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1630 if (ifp == NULL) { 1631 error = EADDRNOTAVAIL; 1632 break; 1633 } 1634 } 1635 /* 1636 * Find the membership in the membership array. 1637 */ 1638 for (i = 0; i < imo->imo_num_memberships; ++i) { 1639 if ((ifp == NULL || 1640 imo->imo_membership[i]->inm_ifp == ifp) && 1641 in_hosteq(imo->imo_membership[i]->inm_addr, 1642 mreq->imr_multiaddr)) 1643 break; 1644 } 1645 if (i == imo->imo_num_memberships) { 1646 error = EADDRNOTAVAIL; 1647 break; 1648 } 1649 /* 1650 * Give up the multicast address record to which the 1651 * membership points. 1652 */ 1653 in_delmulti(imo->imo_membership[i]); 1654 /* 1655 * Remove the gap in the membership array. 1656 */ 1657 for (++i; i < imo->imo_num_memberships; ++i) 1658 imo->imo_membership[i-1] = imo->imo_membership[i]; 1659 --imo->imo_num_memberships; 1660 break; 1661 1662 default: 1663 error = EOPNOTSUPP; 1664 break; 1665 } 1666 1667 /* 1668 * If all options have default values, no need to keep the mbuf. 1669 */ 1670 if (imo->imo_multicast_ifp == NULL && 1671 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1672 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1673 imo->imo_num_memberships == 0) { 1674 kmem_free(imo, sizeof(*imo)); 1675 *imop = NULL; 1676 } 1677 1678 return error; 1679 } 1680 1681 /* 1682 * Return the IP multicast options in response to user getsockopt(). 1683 */ 1684 int 1685 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt) 1686 { 1687 struct in_addr addr; 1688 struct in_ifaddr *ia; 1689 int error; 1690 uint8_t optval; 1691 1692 error = 0; 1693 1694 switch (sopt->sopt_name) { 1695 case IP_MULTICAST_IF: 1696 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1697 addr = zeroin_addr; 1698 else if (imo->imo_multicast_addr.s_addr) { 1699 /* return the value user has set */ 1700 addr = imo->imo_multicast_addr; 1701 } else { 1702 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1703 addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1704 } 1705 error = sockopt_set(sopt, &addr, sizeof(addr)); 1706 break; 1707 1708 case IP_MULTICAST_TTL: 1709 optval = imo ? imo->imo_multicast_ttl 1710 : IP_DEFAULT_MULTICAST_TTL; 1711 1712 error = sockopt_set(sopt, &optval, sizeof(optval)); 1713 break; 1714 1715 case IP_MULTICAST_LOOP: 1716 optval = imo ? imo->imo_multicast_loop 1717 : IP_DEFAULT_MULTICAST_LOOP; 1718 1719 error = sockopt_set(sopt, &optval, sizeof(optval)); 1720 break; 1721 1722 default: 1723 error = EOPNOTSUPP; 1724 } 1725 1726 return (error); 1727 } 1728 1729 /* 1730 * Discard the IP multicast options. 1731 */ 1732 void 1733 ip_freemoptions(struct ip_moptions *imo) 1734 { 1735 int i; 1736 1737 if (imo != NULL) { 1738 for (i = 0; i < imo->imo_num_memberships; ++i) 1739 in_delmulti(imo->imo_membership[i]); 1740 kmem_free(imo, sizeof(*imo)); 1741 } 1742 } 1743 1744 /* 1745 * Routine called from ip_output() to loop back a copy of an IP multicast 1746 * packet to the input queue of a specified interface. Note that this 1747 * calls the output routine of the loopback "driver", but with an interface 1748 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1749 */ 1750 static void 1751 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 1752 { 1753 struct ip *ip; 1754 struct mbuf *copym; 1755 1756 copym = m_copypacket(m, M_DONTWAIT); 1757 if (copym != NULL 1758 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1759 copym = m_pullup(copym, sizeof(struct ip)); 1760 if (copym == NULL) 1761 return; 1762 /* 1763 * We don't bother to fragment if the IP length is greater 1764 * than the interface's MTU. Can this possibly matter? 1765 */ 1766 ip = mtod(copym, struct ip *); 1767 1768 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1769 in_delayed_cksum(copym); 1770 copym->m_pkthdr.csum_flags &= 1771 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1772 } 1773 1774 ip->ip_sum = 0; 1775 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1776 (void)looutput(ifp, copym, sintocsa(dst), NULL); 1777 } 1778
Indexes created Wed Oct 01 12:09:54 GMT 2025