ip_output.c revision 1.223.2.2
1 /* $NetBSD: ip_output.c,v 1.223.2.2 2013/08/28 23:59:36 rmind 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 1.223.2.2 2013/08/28 23:59:36 rmind Exp $"); 95 96 #include "opt_inet.h" 97 #include "opt_ipsec.h" 98 #include "opt_mrouting.h" 99 100 #include <sys/param.h> 101 #include <sys/kmem.h> 102 #include <sys/mbuf.h> 103 #include <sys/protosw.h> 104 #include <sys/socket.h> 105 #include <sys/socketvar.h> 106 #include <sys/kauth.h> 107 #ifdef IPSEC 108 #include <sys/domain.h> 109 #endif 110 #include <sys/systm.h> 111 112 #include <net/if.h> 113 #include <net/route.h> 114 #include <net/pfil.h> 115 116 #include <netinet/in.h> 117 #include <netinet/in_systm.h> 118 #include <netinet/ip.h> 119 #include <netinet/in_pcb.h> 120 #include <netinet/in_var.h> 121 #include <netinet/ip_var.h> 122 #include <netinet/ip_private.h> 123 #include <netinet/in_offload.h> 124 #include <netinet/portalgo.h> 125 #include <netinet/udp.h> 126 127 #ifdef MROUTING 128 #include <netinet/ip_mroute.h> 129 #endif 130 131 #include <netipsec/ipsec.h> 132 #include <netipsec/key.h> 133 134 static int ip_pcbopts(inpcb_t *, const struct sockopt *); 135 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 136 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 137 static void ip_mloopback(struct ifnet *, struct mbuf *, 138 const struct sockaddr_in *); 139 static int ip_setmoptions(inpcb_t *, const struct sockopt *); 140 static int ip_getmoptions(inpcb_t *, struct sockopt *); 141 142 extern pfil_head_t *inet_pfil_hook; /* XXX */ 143 144 int ip_do_loopback_cksum = 0; 145 146 /* 147 * IP output. The packet in mbuf chain m contains a skeletal IP 148 * header (with len, off, ttl, proto, tos, src, dst). 149 * The mbuf chain containing the packet will be freed. 150 * The mbuf opt, if present, will not be freed. 151 */ 152 int 153 ip_output(struct mbuf *m0, ...) 154 { 155 struct rtentry *rt; 156 struct ip *ip; 157 struct ifnet *ifp; 158 struct mbuf *m = m0; 159 int hlen = sizeof (struct ip); 160 int len, error = 0; 161 struct route iproute; 162 const struct sockaddr_in *dst; 163 struct in_ifaddr *ia; 164 struct ifaddr *xifa; 165 struct mbuf *opt; 166 struct route *ro; 167 int flags, sw_csum; 168 u_long mtu; 169 struct ip_moptions *imo; 170 struct socket *so; 171 va_list ap; 172 struct secpolicy *sp = NULL; 173 bool natt_frag = false; 174 bool __unused done = false; 175 union { 176 struct sockaddr dst; 177 struct sockaddr_in dst4; 178 } u; 179 struct sockaddr *rdst = &u.dst; /* real IP destination, as opposed 180 * to the nexthop 181 */ 182 183 len = 0; 184 va_start(ap, m0); 185 opt = va_arg(ap, struct mbuf *); 186 ro = va_arg(ap, struct route *); 187 flags = va_arg(ap, int); 188 imo = va_arg(ap, struct ip_moptions *); 189 so = va_arg(ap, struct socket *); 190 va_end(ap); 191 192 MCLAIM(m, &ip_tx_mowner); 193 194 KASSERT((m->m_flags & M_PKTHDR) != 0); 195 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 0); 196 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) != 197 (M_CSUM_TCPv4|M_CSUM_UDPv4)); 198 199 if (opt) { 200 m = ip_insertoptions(m, opt, &len); 201 if (len >= sizeof(struct ip)) 202 hlen = len; 203 } 204 ip = mtod(m, struct ip *); 205 /* 206 * Fill in IP header. 207 */ 208 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 209 ip->ip_v = IPVERSION; 210 ip->ip_off = htons(0); 211 /* ip->ip_id filled in after we find out source ia */ 212 ip->ip_hl = hlen >> 2; 213 IP_STATINC(IP_STAT_LOCALOUT); 214 } else { 215 hlen = ip->ip_hl << 2; 216 } 217 /* 218 * Route packet. 219 */ 220 memset(&iproute, 0, sizeof(iproute)); 221 if (ro == NULL) 222 ro = &iproute; 223 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 224 dst = satocsin(rtcache_getdst(ro)); 225 /* 226 * If there is a cached route, 227 * check that it is to the same destination 228 * and is still up. If not, free it and try again. 229 * The address family should also be checked in case of sharing the 230 * cache with IPv6. 231 */ 232 if (dst == NULL) 233 ; 234 else if (dst->sin_family != AF_INET || 235 !in_hosteq(dst->sin_addr, ip->ip_dst)) 236 rtcache_free(ro); 237 238 if ((rt = rtcache_validate(ro)) == NULL && 239 (rt = rtcache_update(ro, 1)) == NULL) { 240 dst = &u.dst4; 241 rtcache_setdst(ro, &u.dst); 242 } 243 /* 244 * If routing to interface only, 245 * short circuit routing lookup. 246 */ 247 if (flags & IP_ROUTETOIF) { 248 if ((ia = ifatoia(ifa_ifwithladdr(sintocsa(dst)))) == NULL) { 249 IP_STATINC(IP_STAT_NOROUTE); 250 error = ENETUNREACH; 251 goto bad; 252 } 253 ifp = ia->ia_ifp; 254 mtu = ifp->if_mtu; 255 ip->ip_ttl = 1; 256 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 257 ip->ip_dst.s_addr == INADDR_BROADCAST) && 258 imo != NULL && imo->imo_multicast_ifp != NULL) { 259 ifp = imo->imo_multicast_ifp; 260 mtu = ifp->if_mtu; 261 IFP_TO_IA(ifp, ia); 262 } else { 263 if (rt == NULL) 264 rt = rtcache_init(ro); 265 if (rt == NULL) { 266 IP_STATINC(IP_STAT_NOROUTE); 267 error = EHOSTUNREACH; 268 goto bad; 269 } 270 ia = ifatoia(rt->rt_ifa); 271 ifp = rt->rt_ifp; 272 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 273 mtu = ifp->if_mtu; 274 rt->rt_use++; 275 if (rt->rt_flags & RTF_GATEWAY) 276 dst = satosin(rt->rt_gateway); 277 } 278 if (IN_MULTICAST(ip->ip_dst.s_addr) || 279 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 280 struct in_multi *inm; 281 282 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 283 M_BCAST : M_MCAST; 284 /* 285 * See if the caller provided any multicast options 286 */ 287 if (imo != NULL) 288 ip->ip_ttl = imo->imo_multicast_ttl; 289 else 290 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 291 292 /* 293 * if we don't know the outgoing ifp yet, we can't generate 294 * output 295 */ 296 if (!ifp) { 297 IP_STATINC(IP_STAT_NOROUTE); 298 error = ENETUNREACH; 299 goto bad; 300 } 301 302 /* 303 * If the packet is multicast or broadcast, confirm that 304 * the outgoing interface can transmit it. 305 */ 306 if (((m->m_flags & M_MCAST) && 307 (ifp->if_flags & IFF_MULTICAST) == 0) || 308 ((m->m_flags & M_BCAST) && 309 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 310 IP_STATINC(IP_STAT_NOROUTE); 311 error = ENETUNREACH; 312 goto bad; 313 } 314 /* 315 * If source address not specified yet, use an address 316 * of outgoing interface. 317 */ 318 if (in_nullhost(ip->ip_src)) { 319 struct in_ifaddr *xia; 320 321 IFP_TO_IA(ifp, xia); 322 if (!xia) { 323 error = EADDRNOTAVAIL; 324 goto bad; 325 } 326 xifa = &xia->ia_ifa; 327 if (xifa->ifa_getifa != NULL) { 328 xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 329 } 330 ip->ip_src = xia->ia_addr.sin_addr; 331 } 332 333 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 334 if (inm != NULL && 335 (imo == NULL || imo->imo_multicast_loop)) { 336 /* 337 * If we belong to the destination multicast group 338 * on the outgoing interface, and the caller did not 339 * forbid loopback, loop back a copy. 340 */ 341 ip_mloopback(ifp, m, &u.dst4); 342 } 343 #ifdef MROUTING 344 else { 345 /* 346 * If we are acting as a multicast router, perform 347 * multicast forwarding as if the packet had just 348 * arrived on the interface to which we are about 349 * to send. The multicast forwarding function 350 * recursively calls this function, using the 351 * IP_FORWARDING flag to prevent infinite recursion. 352 * 353 * Multicasts that are looped back by ip_mloopback(), 354 * above, will be forwarded by the ip_input() routine, 355 * if necessary. 356 */ 357 extern struct socket *ip_mrouter; 358 359 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 360 if (ip_mforward(m, ifp) != 0) { 361 m_freem(m); 362 goto done; 363 } 364 } 365 } 366 #endif 367 /* 368 * Multicasts with a time-to-live of zero may be looped- 369 * back, above, but must not be transmitted on a network. 370 * Also, multicasts addressed to the loopback interface 371 * are not sent -- the above call to ip_mloopback() will 372 * loop back a copy if this host actually belongs to the 373 * destination group on the loopback interface. 374 */ 375 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 376 m_freem(m); 377 goto done; 378 } 379 380 goto sendit; 381 } 382 /* 383 * If source address not specified yet, use address 384 * of outgoing interface. 385 */ 386 if (in_nullhost(ip->ip_src)) { 387 xifa = &ia->ia_ifa; 388 if (xifa->ifa_getifa != NULL) 389 ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 390 ip->ip_src = ia->ia_addr.sin_addr; 391 } 392 393 /* 394 * packets with Class-D address as source are not valid per 395 * RFC 1112 396 */ 397 if (IN_MULTICAST(ip->ip_src.s_addr)) { 398 IP_STATINC(IP_STAT_ODROPPED); 399 error = EADDRNOTAVAIL; 400 goto bad; 401 } 402 403 /* 404 * Look for broadcast address and 405 * and verify user is allowed to send 406 * such a packet. 407 */ 408 if (in_broadcast(dst->sin_addr, ifp)) { 409 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 410 error = EADDRNOTAVAIL; 411 goto bad; 412 } 413 if ((flags & IP_ALLOWBROADCAST) == 0) { 414 error = EACCES; 415 goto bad; 416 } 417 /* don't allow broadcast messages to be fragmented */ 418 if (ntohs(ip->ip_len) > ifp->if_mtu) { 419 error = EMSGSIZE; 420 goto bad; 421 } 422 m->m_flags |= M_BCAST; 423 } else 424 m->m_flags &= ~M_BCAST; 425 426 sendit: 427 if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) { 428 if (m->m_pkthdr.len < IP_MINFRAGSIZE) { 429 ip->ip_id = 0; 430 } else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 431 ip->ip_id = ip_newid(ia); 432 } else { 433 434 /* 435 * TSO capable interfaces (typically?) increment 436 * ip_id for each segment. 437 * "allocate" enough ids here to increase the chance 438 * for them to be unique. 439 * 440 * note that the following calculation is not 441 * needed to be precise. wasting some ip_id is fine. 442 */ 443 444 unsigned int segsz = m->m_pkthdr.segsz; 445 unsigned int datasz = ntohs(ip->ip_len) - hlen; 446 unsigned int num = howmany(datasz, segsz); 447 448 ip->ip_id = ip_newid_range(ia, num); 449 } 450 } 451 /* 452 * If we're doing Path MTU Discovery, we need to set DF unless 453 * the route's MTU is locked. 454 */ 455 if ((flags & IP_MTUDISC) != 0 && rt != NULL && 456 (rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 457 ip->ip_off |= htons(IP_DF); 458 459 #ifdef IPSEC 460 /* Perform IPsec processing, if any. */ 461 error = ipsec4_output(m, so, flags, &sp, &mtu, &natt_frag, &done); 462 if (error || done) { 463 goto done; 464 } 465 #endif 466 467 /* 468 * Run through list of hooks for output packets. 469 */ 470 if ((error = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 471 goto done; 472 if (m == NULL) 473 goto done; 474 475 ip = mtod(m, struct ip *); 476 hlen = ip->ip_hl << 2; 477 478 m->m_pkthdr.csum_data |= hlen << 16; 479 480 #if IFA_STATS 481 /* 482 * search for the source address structure to 483 * maintain output statistics. 484 */ 485 INADDR_TO_IA(ip->ip_src, ia); 486 #endif 487 488 /* Maybe skip checksums on loopback interfaces. */ 489 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 490 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 491 } 492 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 493 /* 494 * If small enough for mtu of path, or if using TCP segmentation 495 * offload, can just send directly. 496 */ 497 if (ntohs(ip->ip_len) <= mtu || 498 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 499 #if IFA_STATS 500 if (ia) 501 ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len); 502 #endif 503 /* 504 * Always initialize the sum to 0! Some HW assisted 505 * checksumming requires this. 506 */ 507 ip->ip_sum = 0; 508 509 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 510 /* 511 * Perform any checksums that the hardware can't do 512 * for us. 513 * 514 * XXX Does any hardware require the {th,uh}_sum 515 * XXX fields to be 0? 516 */ 517 if (sw_csum & M_CSUM_IPv4) { 518 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 519 ip->ip_sum = in_cksum(m, hlen); 520 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 521 } 522 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 523 if (IN_NEED_CHECKSUM(ifp, 524 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 525 in_delayed_cksum(m); 526 } 527 m->m_pkthdr.csum_flags &= 528 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 529 } 530 } 531 532 if (__predict_true( 533 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 || 534 (ifp->if_capenable & IFCAP_TSOv4) != 0)) { 535 KERNEL_LOCK(1, NULL); 536 error = 537 (*ifp->if_output)(ifp, m, 538 (m->m_flags & M_MCAST) ? 539 sintocsa(rdst) : sintocsa(dst), 540 rt); 541 KERNEL_UNLOCK_ONE(NULL); 542 } else { 543 error = 544 ip_tso_output(ifp, m, 545 (m->m_flags & M_MCAST) ? 546 sintocsa(rdst) : sintocsa(dst), 547 rt); 548 } 549 goto done; 550 } 551 552 /* 553 * We can't use HW checksumming if we're about to 554 * to fragment the packet. 555 * 556 * XXX Some hardware can do this. 557 */ 558 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 559 if (IN_NEED_CHECKSUM(ifp, 560 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 561 in_delayed_cksum(m); 562 } 563 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 564 } 565 566 /* 567 * Too large for interface; fragment if possible. 568 * Must be able to put at least 8 bytes per fragment. 569 */ 570 if (ntohs(ip->ip_off) & IP_DF) { 571 if (flags & IP_RETURNMTU) { 572 inpcb_t *inp = sotoinpcb(so); 573 inpcb_set_errormtu(inp, mtu); 574 } 575 error = EMSGSIZE; 576 IP_STATINC(IP_STAT_CANTFRAG); 577 goto bad; 578 } 579 580 error = ip_fragment(m, ifp, mtu); 581 if (error) { 582 m = NULL; 583 goto bad; 584 } 585 586 for (; m; m = m0) { 587 m0 = m->m_nextpkt; 588 m->m_nextpkt = 0; 589 if (error == 0) { 590 #if IFA_STATS 591 if (ia) 592 ia->ia_ifa.ifa_data.ifad_outbytes += 593 ntohs(ip->ip_len); 594 #endif 595 /* 596 * If we get there, the packet has not been handled by 597 * IPsec whereas it should have. Now that it has been 598 * fragmented, re-inject it in ip_output so that IPsec 599 * processing can occur. 600 */ 601 if (natt_frag) { 602 error = ip_output(m, opt, ro, 603 flags | IP_RAWOUTPUT | IP_NOIPNEWID, 604 imo, so); 605 } else { 606 KASSERT((m->m_pkthdr.csum_flags & 607 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 608 KERNEL_LOCK(1, NULL); 609 error = (*ifp->if_output)(ifp, m, 610 (m->m_flags & M_MCAST) ? 611 sintocsa(rdst) : sintocsa(dst), rt); 612 KERNEL_UNLOCK_ONE(NULL); 613 } 614 } else 615 m_freem(m); 616 } 617 618 if (error == 0) 619 IP_STATINC(IP_STAT_FRAGMENTED); 620 done: 621 rtcache_free(&iproute); 622 if (sp) { 623 #ifdef IPSEC 624 KEY_FREESP(&sp); 625 #endif 626 } 627 return error; 628 bad: 629 m_freem(m); 630 goto done; 631 } 632 633 int 634 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 635 { 636 struct ip *ip, *mhip; 637 struct mbuf *m0; 638 int len, hlen, off; 639 int mhlen, firstlen; 640 struct mbuf **mnext; 641 int sw_csum = m->m_pkthdr.csum_flags; 642 int fragments = 0; 643 int s; 644 int error = 0; 645 646 ip = mtod(m, struct ip *); 647 hlen = ip->ip_hl << 2; 648 if (ifp != NULL) 649 sw_csum &= ~ifp->if_csum_flags_tx; 650 651 len = (mtu - hlen) &~ 7; 652 if (len < 8) { 653 m_freem(m); 654 return (EMSGSIZE); 655 } 656 657 firstlen = len; 658 mnext = &m->m_nextpkt; 659 660 /* 661 * Loop through length of segment after first fragment, 662 * make new header and copy data of each part and link onto chain. 663 */ 664 m0 = m; 665 mhlen = sizeof (struct ip); 666 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 667 MGETHDR(m, M_DONTWAIT, MT_HEADER); 668 if (m == 0) { 669 error = ENOBUFS; 670 IP_STATINC(IP_STAT_ODROPPED); 671 goto sendorfree; 672 } 673 MCLAIM(m, m0->m_owner); 674 *mnext = m; 675 mnext = &m->m_nextpkt; 676 m->m_data += max_linkhdr; 677 mhip = mtod(m, struct ip *); 678 *mhip = *ip; 679 /* we must inherit MCAST and BCAST flags */ 680 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 681 if (hlen > sizeof (struct ip)) { 682 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 683 mhip->ip_hl = mhlen >> 2; 684 } 685 m->m_len = mhlen; 686 mhip->ip_off = ((off - hlen) >> 3) + 687 (ntohs(ip->ip_off) & ~IP_MF); 688 if (ip->ip_off & htons(IP_MF)) 689 mhip->ip_off |= IP_MF; 690 if (off + len >= ntohs(ip->ip_len)) 691 len = ntohs(ip->ip_len) - off; 692 else 693 mhip->ip_off |= IP_MF; 694 HTONS(mhip->ip_off); 695 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 696 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 697 if (m->m_next == 0) { 698 error = ENOBUFS; /* ??? */ 699 IP_STATINC(IP_STAT_ODROPPED); 700 goto sendorfree; 701 } 702 m->m_pkthdr.len = mhlen + len; 703 m->m_pkthdr.rcvif = NULL; 704 mhip->ip_sum = 0; 705 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 706 if (sw_csum & M_CSUM_IPv4) { 707 mhip->ip_sum = in_cksum(m, mhlen); 708 } else { 709 /* 710 * checksum is hw-offloaded or not necessary. 711 */ 712 m->m_pkthdr.csum_flags |= 713 m0->m_pkthdr.csum_flags & M_CSUM_IPv4; 714 m->m_pkthdr.csum_data |= mhlen << 16; 715 KASSERT(!(ifp != NULL && 716 IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) 717 || (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 718 } 719 IP_STATINC(IP_STAT_OFRAGMENTS); 720 fragments++; 721 } 722 /* 723 * Update first fragment by trimming what's been copied out 724 * and updating header, then send each fragment (in order). 725 */ 726 m = m0; 727 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 728 m->m_pkthdr.len = hlen + firstlen; 729 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 730 ip->ip_off |= htons(IP_MF); 731 ip->ip_sum = 0; 732 if (sw_csum & M_CSUM_IPv4) { 733 ip->ip_sum = in_cksum(m, hlen); 734 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 735 } else { 736 /* 737 * checksum is hw-offloaded or not necessary. 738 */ 739 KASSERT(!(ifp != NULL && IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) 740 || (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 741 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 742 sizeof(struct ip)); 743 } 744 sendorfree: 745 /* 746 * If there is no room for all the fragments, don't queue 747 * any of them. 748 */ 749 if (ifp != NULL) { 750 s = splnet(); 751 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 752 error == 0) { 753 error = ENOBUFS; 754 IP_STATINC(IP_STAT_ODROPPED); 755 IFQ_INC_DROPS(&ifp->if_snd); 756 } 757 splx(s); 758 } 759 if (error) { 760 for (m = m0; m; m = m0) { 761 m0 = m->m_nextpkt; 762 m->m_nextpkt = NULL; 763 m_freem(m); 764 } 765 } 766 return (error); 767 } 768 769 /* 770 * Process a delayed payload checksum calculation. 771 */ 772 void 773 in_delayed_cksum(struct mbuf *m) 774 { 775 struct ip *ip; 776 u_int16_t csum, offset; 777 778 ip = mtod(m, struct ip *); 779 offset = ip->ip_hl << 2; 780 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 781 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 782 csum = 0xffff; 783 784 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 785 786 if ((offset + sizeof(u_int16_t)) > m->m_len) { 787 /* This happen when ip options were inserted 788 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 789 m->m_len, offset, ip->ip_p); 790 */ 791 m_copyback(m, offset, sizeof(csum), (void *) &csum); 792 } else 793 *(u_int16_t *)(mtod(m, char *) + offset) = csum; 794 } 795 796 /* 797 * Determine the maximum length of the options to be inserted; 798 * we would far rather allocate too much space rather than too little. 799 */ 800 801 u_int 802 ip_optlen(inpcb_t *inp) 803 { 804 struct mbuf *m = inpcb_get_options(inp); 805 806 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) { 807 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 808 } 809 return 0; 810 } 811 812 /* 813 * Insert IP options into preformed packet. 814 * Adjust IP destination as required for IP source routing, 815 * as indicated by a non-zero in_addr at the start of the options. 816 */ 817 static struct mbuf * 818 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 819 { 820 struct ipoption *p = mtod(opt, struct ipoption *); 821 struct mbuf *n; 822 struct ip *ip = mtod(m, struct ip *); 823 unsigned optlen; 824 825 optlen = opt->m_len - sizeof(p->ipopt_dst); 826 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 827 return (m); /* XXX should fail */ 828 if (!in_nullhost(p->ipopt_dst)) 829 ip->ip_dst = p->ipopt_dst; 830 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 831 MGETHDR(n, M_DONTWAIT, MT_HEADER); 832 if (n == 0) 833 return (m); 834 MCLAIM(n, m->m_owner); 835 M_MOVE_PKTHDR(n, m); 836 m->m_len -= sizeof(struct ip); 837 m->m_data += sizeof(struct ip); 838 n->m_next = m; 839 m = n; 840 m->m_len = optlen + sizeof(struct ip); 841 m->m_data += max_linkhdr; 842 bcopy((void *)ip, mtod(m, void *), sizeof(struct ip)); 843 } else { 844 m->m_data -= optlen; 845 m->m_len += optlen; 846 memmove(mtod(m, void *), ip, sizeof(struct ip)); 847 } 848 m->m_pkthdr.len += optlen; 849 ip = mtod(m, struct ip *); 850 bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen); 851 *phlen = sizeof(struct ip) + optlen; 852 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 853 return (m); 854 } 855 856 /* 857 * Copy options from ip to jp, 858 * omitting those not copied during fragmentation. 859 */ 860 int 861 ip_optcopy(struct ip *ip, struct ip *jp) 862 { 863 u_char *cp, *dp; 864 int opt, optlen, cnt; 865 866 cp = (u_char *)(ip + 1); 867 dp = (u_char *)(jp + 1); 868 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 869 for (; cnt > 0; cnt -= optlen, cp += optlen) { 870 opt = cp[0]; 871 if (opt == IPOPT_EOL) 872 break; 873 if (opt == IPOPT_NOP) { 874 /* Preserve for IP mcast tunnel's LSRR alignment. */ 875 *dp++ = IPOPT_NOP; 876 optlen = 1; 877 continue; 878 } 879 880 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp)); 881 optlen = cp[IPOPT_OLEN]; 882 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen < cnt); 883 884 /* Invalid lengths should have been caught by ip_dooptions. */ 885 if (optlen > cnt) 886 optlen = cnt; 887 if (IPOPT_COPIED(opt)) { 888 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 889 dp += optlen; 890 } 891 } 892 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 893 *dp++ = IPOPT_EOL; 894 return (optlen); 895 } 896 897 /* 898 * IP socket option processing. 899 */ 900 int 901 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt) 902 { 903 inpcb_t *inp = sotoinpcb(so); 904 struct ip *ip = in_getiphdr(inp); 905 int inpflags = inpcb_get_flags(inp); 906 int optval = 0, error = 0; 907 908 if (sopt->sopt_level != IPPROTO_IP) { 909 if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER) 910 return 0; 911 return ENOPROTOOPT; 912 } 913 914 switch (op) { 915 case PRCO_SETOPT: 916 switch (sopt->sopt_name) { 917 case IP_OPTIONS: 918 #ifdef notyet 919 case IP_RETOPTS: 920 #endif 921 error = ip_pcbopts(inp, sopt); 922 break; 923 924 case IP_TOS: 925 case IP_TTL: 926 case IP_MINTTL: 927 case IP_PKTINFO: 928 case IP_RECVOPTS: 929 case IP_RECVRETOPTS: 930 case IP_RECVDSTADDR: 931 case IP_RECVIF: 932 case IP_RECVPKTINFO: 933 case IP_RECVTTL: 934 error = sockopt_getint(sopt, &optval); 935 if (error) 936 break; 937 938 switch (sopt->sopt_name) { 939 case IP_TOS: 940 ip->ip_tos = optval; 941 break; 942 943 case IP_TTL: 944 ip->ip_ttl = optval; 945 break; 946 947 case IP_MINTTL: 948 if (optval > 0 && optval <= MAXTTL) 949 inpcb_set_minttl(inp, optval); 950 else 951 error = EINVAL; 952 break; 953 #define OPTSET(bit) \ 954 if (optval) \ 955 inpflags |= bit; \ 956 else \ 957 inpflags &= ~bit; 958 959 case IP_PKTINFO: 960 OPTSET(INP_PKTINFO); 961 break; 962 963 case IP_RECVOPTS: 964 OPTSET(INP_RECVOPTS); 965 break; 966 967 case IP_RECVPKTINFO: 968 OPTSET(INP_RECVPKTINFO); 969 break; 970 971 case IP_RECVRETOPTS: 972 OPTSET(INP_RECVRETOPTS); 973 break; 974 975 case IP_RECVDSTADDR: 976 OPTSET(INP_RECVDSTADDR); 977 break; 978 979 case IP_RECVIF: 980 OPTSET(INP_RECVIF); 981 break; 982 983 case IP_RECVTTL: 984 OPTSET(INP_RECVTTL); 985 break; 986 } 987 break; 988 #undef OPTSET 989 990 case IP_MULTICAST_IF: 991 case IP_MULTICAST_TTL: 992 case IP_MULTICAST_LOOP: 993 case IP_ADD_MEMBERSHIP: 994 case IP_DROP_MEMBERSHIP: 995 error = ip_setmoptions(inp, sopt); 996 break; 997 998 case IP_PORTRANGE: 999 error = sockopt_getint(sopt, &optval); 1000 if (error) 1001 break; 1002 1003 switch (optval) { 1004 case IP_PORTRANGE_DEFAULT: 1005 case IP_PORTRANGE_HIGH: 1006 inpflags &= ~(INP_LOWPORT); 1007 break; 1008 1009 case IP_PORTRANGE_LOW: 1010 inpflags |= INP_LOWPORT; 1011 break; 1012 1013 default: 1014 error = EINVAL; 1015 break; 1016 } 1017 break; 1018 1019 case IP_PORTALGO: 1020 error = sockopt_getint(sopt, &optval); 1021 if (error) 1022 break; 1023 1024 error = portalgo_algo_index_select( 1025 (struct inpcb_hdr *)inp, optval); 1026 break; 1027 1028 #if defined(IPSEC) 1029 case IP_IPSEC_POLICY: 1030 error = ipsec4_set_policy(inp, sopt->sopt_name, 1031 sopt->sopt_data, sopt->sopt_size, curlwp->l_cred); 1032 break; 1033 #endif /*IPSEC*/ 1034 1035 default: 1036 error = ENOPROTOOPT; 1037 break; 1038 } 1039 break; 1040 1041 case PRCO_GETOPT: 1042 switch (sopt->sopt_name) { 1043 case IP_OPTIONS: 1044 case IP_RETOPTS: { 1045 struct mbuf *mopts = inpcb_get_options(inp); 1046 1047 if (mopts) { 1048 struct mbuf *m; 1049 1050 m = m_copym(mopts, 0, M_COPYALL, M_DONTWAIT); 1051 if (m == NULL) { 1052 error = ENOBUFS; 1053 break; 1054 } 1055 error = sockopt_setmbuf(sopt, m); 1056 } 1057 break; 1058 } 1059 case IP_PKTINFO: 1060 case IP_TOS: 1061 case IP_TTL: 1062 case IP_MINTTL: 1063 case IP_RECVOPTS: 1064 case IP_RECVRETOPTS: 1065 case IP_RECVDSTADDR: 1066 case IP_RECVIF: 1067 case IP_RECVPKTINFO: 1068 case IP_RECVTTL: 1069 case IP_ERRORMTU: 1070 switch (sopt->sopt_name) { 1071 case IP_TOS: 1072 optval = ip->ip_tos; 1073 break; 1074 1075 case IP_TTL: 1076 optval = ip->ip_ttl; 1077 break; 1078 1079 case IP_MINTTL: 1080 optval = inpcb_get_minttl(inp); 1081 break; 1082 1083 case IP_ERRORMTU: 1084 optval = inpcb_get_errormtu(inp); 1085 break; 1086 1087 #define OPTBIT(bit) (inpflags & bit ? 1 : 0) 1088 1089 case IP_PKTINFO: 1090 optval = OPTBIT(INP_PKTINFO); 1091 break; 1092 1093 case IP_RECVOPTS: 1094 optval = OPTBIT(INP_RECVOPTS); 1095 break; 1096 1097 case IP_RECVPKTINFO: 1098 optval = OPTBIT(INP_RECVPKTINFO); 1099 break; 1100 1101 case IP_RECVRETOPTS: 1102 optval = OPTBIT(INP_RECVRETOPTS); 1103 break; 1104 1105 case IP_RECVDSTADDR: 1106 optval = OPTBIT(INP_RECVDSTADDR); 1107 break; 1108 1109 case IP_RECVIF: 1110 optval = OPTBIT(INP_RECVIF); 1111 break; 1112 1113 case IP_RECVTTL: 1114 optval = OPTBIT(INP_RECVTTL); 1115 break; 1116 } 1117 error = sockopt_setint(sopt, optval); 1118 break; 1119 1120 #if 0 /* defined(IPSEC) */ 1121 case IP_IPSEC_POLICY: 1122 { 1123 struct mbuf *m = NULL; 1124 1125 /* XXX this will return EINVAL as sopt is empty */ 1126 error = ipsec4_get_policy(inp, sopt->sopt_data, 1127 sopt->sopt_size, &m); 1128 if (error == 0) 1129 error = sockopt_setmbuf(sopt, m); 1130 break; 1131 } 1132 #endif /*IPSEC*/ 1133 1134 case IP_MULTICAST_IF: 1135 case IP_MULTICAST_TTL: 1136 case IP_MULTICAST_LOOP: 1137 case IP_ADD_MEMBERSHIP: 1138 case IP_DROP_MEMBERSHIP: 1139 error = ip_getmoptions(inp, sopt); 1140 break; 1141 1142 case IP_PORTRANGE: 1143 if (inpflags & INP_LOWPORT) 1144 optval = IP_PORTRANGE_LOW; 1145 else 1146 optval = IP_PORTRANGE_DEFAULT; 1147 error = sockopt_setint(sopt, optval); 1148 break; 1149 1150 case IP_PORTALGO: 1151 optval = inpcb_get_portalgo(inp); 1152 error = sockopt_setint(sopt, optval); 1153 break; 1154 1155 default: 1156 error = ENOPROTOOPT; 1157 break; 1158 } 1159 break; 1160 } 1161 1162 if (!error) { 1163 inpcb_set_flags(inp, inpflags); 1164 } 1165 return error; 1166 } 1167 1168 /* 1169 * Set up IP options in pcb for insertion in output packets. 1170 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1171 * with destination address if source routed. 1172 */ 1173 static int 1174 ip_pcbopts(inpcb_t *inp, const struct sockopt *sopt) 1175 { 1176 struct mbuf *m; 1177 const u_char *cp; 1178 u_char *dp; 1179 int cnt; 1180 1181 /* Turn off any old options. */ 1182 inpcb_set_options(inp, NULL); 1183 if ((cnt = sopt->sopt_size) == 0) { 1184 /* Only turning off any previous options. */ 1185 return 0; 1186 } 1187 cp = sopt->sopt_data; 1188 1189 #ifndef __vax__ 1190 if (cnt % sizeof(int32_t)) 1191 return (EINVAL); 1192 #endif 1193 1194 m = m_get(M_DONTWAIT, MT_SOOPTS); 1195 if (m == NULL) 1196 return (ENOBUFS); 1197 1198 dp = mtod(m, u_char *); 1199 memset(dp, 0, sizeof(struct in_addr)); 1200 dp += sizeof(struct in_addr); 1201 m->m_len = sizeof(struct in_addr); 1202 1203 /* 1204 * IP option list according to RFC791. Each option is of the form 1205 * 1206 * [optval] [olen] [(olen - 2) data bytes] 1207 * 1208 * We validate the list and copy options to an mbuf for prepending 1209 * to data packets. The IP first-hop destination address will be 1210 * stored before actual options and is zero if unset. 1211 */ 1212 while (cnt > 0) { 1213 uint8_t optval, olen, offset; 1214 1215 optval = cp[IPOPT_OPTVAL]; 1216 1217 if (optval == IPOPT_EOL || optval == IPOPT_NOP) { 1218 olen = 1; 1219 } else { 1220 if (cnt < IPOPT_OLEN + 1) 1221 goto bad; 1222 1223 olen = cp[IPOPT_OLEN]; 1224 if (olen < IPOPT_OLEN + 1 || olen > cnt) 1225 goto bad; 1226 } 1227 1228 if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) { 1229 /* 1230 * user process specifies route as: 1231 * ->A->B->C->D 1232 * D must be our final destination (but we can't 1233 * check that since we may not have connected yet). 1234 * A is first hop destination, which doesn't appear in 1235 * actual IP option, but is stored before the options. 1236 */ 1237 if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr)) 1238 goto bad; 1239 1240 offset = cp[IPOPT_OFFSET]; 1241 memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1, 1242 sizeof(struct in_addr)); 1243 1244 cp += sizeof(struct in_addr); 1245 cnt -= sizeof(struct in_addr); 1246 olen -= sizeof(struct in_addr); 1247 1248 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1249 goto bad; 1250 1251 memcpy(dp, cp, olen); 1252 dp[IPOPT_OPTVAL] = optval; 1253 dp[IPOPT_OLEN] = olen; 1254 dp[IPOPT_OFFSET] = offset; 1255 break; 1256 } else { 1257 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1258 goto bad; 1259 1260 memcpy(dp, cp, olen); 1261 break; 1262 } 1263 1264 dp += olen; 1265 m->m_len += olen; 1266 1267 if (optval == IPOPT_EOL) 1268 break; 1269 1270 cp += olen; 1271 cnt -= olen; 1272 } 1273 inpcb_set_options(inp, m); 1274 return 0; 1275 bad: 1276 (void)m_free(m); 1277 return (EINVAL); 1278 } 1279 1280 /* 1281 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1282 */ 1283 static struct ifnet * 1284 ip_multicast_if(struct in_addr *a, int *ifindexp) 1285 { 1286 int ifindex; 1287 struct ifnet *ifp = NULL; 1288 struct in_ifaddr *ia; 1289 1290 if (ifindexp) 1291 *ifindexp = 0; 1292 if (ntohl(a->s_addr) >> 24 == 0) { 1293 ifindex = ntohl(a->s_addr) & 0xffffff; 1294 ifp = if_byindex(ifindex); 1295 if (!ifp) 1296 return NULL; 1297 if (ifindexp) 1298 *ifindexp = ifindex; 1299 } else { 1300 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1301 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1302 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1303 ifp = ia->ia_ifp; 1304 break; 1305 } 1306 } 1307 } 1308 return ifp; 1309 } 1310 1311 static int 1312 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval) 1313 { 1314 u_int tval; 1315 u_char cval; 1316 int error; 1317 1318 if (sopt == NULL) 1319 return EINVAL; 1320 1321 switch (sopt->sopt_size) { 1322 case sizeof(u_char): 1323 error = sockopt_get(sopt, &cval, sizeof(u_char)); 1324 tval = cval; 1325 break; 1326 1327 case sizeof(u_int): 1328 error = sockopt_get(sopt, &tval, sizeof(u_int)); 1329 break; 1330 1331 default: 1332 error = EINVAL; 1333 } 1334 1335 if (error) 1336 return error; 1337 1338 if (tval > maxval) 1339 return EINVAL; 1340 1341 *val = tval; 1342 return 0; 1343 } 1344 1345 /* 1346 * Set the IP multicast options in response to user setsockopt(). 1347 */ 1348 static int 1349 ip_setmoptions(inpcb_t *inp, const struct sockopt *sopt) 1350 { 1351 struct ip_moptions *imo = inpcb_get_moptions(inp); 1352 struct in_addr addr; 1353 struct ip_mreq lmreq, *mreq; 1354 struct ifnet *ifp; 1355 int i, ifindex, error = 0; 1356 1357 if (!imo) { 1358 /* 1359 * No multicast option buffer attached to the pcb; 1360 * allocate one and initialize to default values. 1361 */ 1362 imo = kmem_intr_alloc(sizeof(*imo), KM_NOSLEEP); 1363 if (imo == NULL) 1364 return ENOBUFS; 1365 1366 imo->imo_multicast_ifp = NULL; 1367 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1368 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1369 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1370 imo->imo_num_memberships = 0; 1371 inpcb_set_moptions(inp, imo); 1372 } 1373 1374 switch (sopt->sopt_name) { 1375 case IP_MULTICAST_IF: 1376 /* 1377 * Select the interface for outgoing multicast packets. 1378 */ 1379 error = sockopt_get(sopt, &addr, sizeof(addr)); 1380 if (error) 1381 break; 1382 1383 /* 1384 * INADDR_ANY is used to remove a previous selection. 1385 * When no interface is selected, a default one is 1386 * chosen every time a multicast packet is sent. 1387 */ 1388 if (in_nullhost(addr)) { 1389 imo->imo_multicast_ifp = NULL; 1390 break; 1391 } 1392 /* 1393 * The selected interface is identified by its local 1394 * IP address. Find the interface and confirm that 1395 * it supports multicasting. 1396 */ 1397 ifp = ip_multicast_if(&addr, &ifindex); 1398 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1399 error = EADDRNOTAVAIL; 1400 break; 1401 } 1402 imo->imo_multicast_ifp = ifp; 1403 if (ifindex) 1404 imo->imo_multicast_addr = addr; 1405 else 1406 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1407 break; 1408 1409 case IP_MULTICAST_TTL: 1410 /* 1411 * Set the IP time-to-live for outgoing multicast packets. 1412 */ 1413 error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL); 1414 break; 1415 1416 case IP_MULTICAST_LOOP: 1417 /* 1418 * Set the loopback flag for outgoing multicast packets. 1419 * Must be zero or one. 1420 */ 1421 error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1); 1422 break; 1423 1424 case IP_ADD_MEMBERSHIP: 1425 /* 1426 * Add a multicast group membership. 1427 * Group must be a valid IP multicast address. 1428 */ 1429 error = sockopt_get(sopt, &lmreq, sizeof(lmreq)); 1430 if (error) 1431 break; 1432 1433 mreq = &lmreq; 1434 1435 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1436 error = EINVAL; 1437 break; 1438 } 1439 /* 1440 * If no interface address was provided, use the interface of 1441 * the route to the given multicast address. 1442 */ 1443 if (in_nullhost(mreq->imr_interface)) { 1444 struct rtentry *rt; 1445 union { 1446 struct sockaddr dst; 1447 struct sockaddr_in dst4; 1448 } u; 1449 struct route ro; 1450 1451 memset(&ro, 0, sizeof(ro)); 1452 1453 sockaddr_in_init(&u.dst4, &mreq->imr_multiaddr, 0); 1454 rtcache_setdst(&ro, &u.dst); 1455 ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp 1456 : NULL; 1457 rtcache_free(&ro); 1458 } else { 1459 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1460 } 1461 /* 1462 * See if we found an interface, and confirm that it 1463 * supports multicast. 1464 */ 1465 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1466 error = EADDRNOTAVAIL; 1467 break; 1468 } 1469 /* 1470 * See if the membership already exists or if all the 1471 * membership slots are full. 1472 */ 1473 for (i = 0; i < imo->imo_num_memberships; ++i) { 1474 if (imo->imo_membership[i]->inm_ifp == ifp && 1475 in_hosteq(imo->imo_membership[i]->inm_addr, 1476 mreq->imr_multiaddr)) 1477 break; 1478 } 1479 if (i < imo->imo_num_memberships) { 1480 error = EADDRINUSE; 1481 break; 1482 } 1483 if (i == IP_MAX_MEMBERSHIPS) { 1484 error = ETOOMANYREFS; 1485 break; 1486 } 1487 /* 1488 * Everything looks good; add a new record to the multicast 1489 * address list for the given interface. 1490 */ 1491 if ((imo->imo_membership[i] = 1492 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1493 error = ENOBUFS; 1494 break; 1495 } 1496 ++imo->imo_num_memberships; 1497 break; 1498 1499 case IP_DROP_MEMBERSHIP: 1500 /* 1501 * Drop a multicast group membership. 1502 * Group must be a valid IP multicast address. 1503 */ 1504 error = sockopt_get(sopt, &lmreq, sizeof(lmreq)); 1505 if (error) 1506 break; 1507 1508 mreq = &lmreq; 1509 1510 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1511 error = EINVAL; 1512 break; 1513 } 1514 /* 1515 * If an interface address was specified, get a pointer 1516 * to its ifnet structure. 1517 */ 1518 if (in_nullhost(mreq->imr_interface)) 1519 ifp = NULL; 1520 else { 1521 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1522 if (ifp == NULL) { 1523 error = EADDRNOTAVAIL; 1524 break; 1525 } 1526 } 1527 /* 1528 * Find the membership in the membership array. 1529 */ 1530 for (i = 0; i < imo->imo_num_memberships; ++i) { 1531 if ((ifp == NULL || 1532 imo->imo_membership[i]->inm_ifp == ifp) && 1533 in_hosteq(imo->imo_membership[i]->inm_addr, 1534 mreq->imr_multiaddr)) 1535 break; 1536 } 1537 if (i == imo->imo_num_memberships) { 1538 error = EADDRNOTAVAIL; 1539 break; 1540 } 1541 /* 1542 * Give up the multicast address record to which the 1543 * membership points. 1544 */ 1545 in_delmulti(imo->imo_membership[i]); 1546 /* 1547 * Remove the gap in the membership array. 1548 */ 1549 for (++i; i < imo->imo_num_memberships; ++i) 1550 imo->imo_membership[i-1] = imo->imo_membership[i]; 1551 --imo->imo_num_memberships; 1552 break; 1553 1554 default: 1555 error = EOPNOTSUPP; 1556 break; 1557 } 1558 1559 /* 1560 * If all options have default values, no need to keep the mbuf. 1561 */ 1562 if (imo->imo_multicast_ifp == NULL && 1563 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1564 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1565 imo->imo_num_memberships == 0) { 1566 kmem_free(imo, sizeof(*imo)); 1567 inpcb_set_moptions(inp, NULL); 1568 } 1569 1570 return error; 1571 } 1572 1573 /* 1574 * Return the IP multicast options in response to user getsockopt(). 1575 */ 1576 static int 1577 ip_getmoptions(inpcb_t *inp, struct sockopt *sopt) 1578 { 1579 struct ip_moptions *imo = inpcb_get_moptions(inp); 1580 struct in_addr addr; 1581 struct in_ifaddr *ia; 1582 uint8_t optval; 1583 int error = 0; 1584 1585 switch (sopt->sopt_name) { 1586 case IP_MULTICAST_IF: 1587 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1588 addr = zeroin_addr; 1589 else if (imo->imo_multicast_addr.s_addr) { 1590 /* return the value user has set */ 1591 addr = imo->imo_multicast_addr; 1592 } else { 1593 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1594 addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1595 } 1596 error = sockopt_set(sopt, &addr, sizeof(addr)); 1597 break; 1598 1599 case IP_MULTICAST_TTL: 1600 optval = imo ? imo->imo_multicast_ttl 1601 : IP_DEFAULT_MULTICAST_TTL; 1602 1603 error = sockopt_set(sopt, &optval, sizeof(optval)); 1604 break; 1605 1606 case IP_MULTICAST_LOOP: 1607 optval = imo ? imo->imo_multicast_loop 1608 : IP_DEFAULT_MULTICAST_LOOP; 1609 1610 error = sockopt_set(sopt, &optval, sizeof(optval)); 1611 break; 1612 1613 default: 1614 error = EOPNOTSUPP; 1615 } 1616 1617 return error; 1618 } 1619 1620 /* 1621 * Discard the IP multicast options. 1622 */ 1623 void 1624 ip_freemoptions(struct ip_moptions *imo) 1625 { 1626 int i; 1627 1628 if (imo != NULL) { 1629 for (i = 0; i < imo->imo_num_memberships; ++i) 1630 in_delmulti(imo->imo_membership[i]); 1631 kmem_free(imo, sizeof(*imo)); 1632 } 1633 } 1634 1635 /* 1636 * Routine called from ip_output() to loop back a copy of an IP multicast 1637 * packet to the input queue of a specified interface. Note that this 1638 * calls the output routine of the loopback "driver", but with an interface 1639 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1640 */ 1641 static void 1642 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 1643 { 1644 struct ip *ip; 1645 struct mbuf *copym; 1646 1647 copym = m_copypacket(m, M_DONTWAIT); 1648 if (copym != NULL 1649 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1650 copym = m_pullup(copym, sizeof(struct ip)); 1651 if (copym == NULL) 1652 return; 1653 /* 1654 * We don't bother to fragment if the IP length is greater 1655 * than the interface's MTU. Can this possibly matter? 1656 */ 1657 ip = mtod(copym, struct ip *); 1658 1659 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1660 in_delayed_cksum(copym); 1661 copym->m_pkthdr.csum_flags &= 1662 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1663 } 1664 1665 ip->ip_sum = 0; 1666 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1667 (void)looutput(ifp, copym, sintocsa(dst), NULL); 1668 } 1669
Indexes created Wed Oct 01 12:09:54 GMT 2025