ip_output.c revision 1.238
1 /* $NetBSD: ip_output.c,v 1.238 2015/04/27 10:14:44 ozaki-r 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 1.238 2015/04/27 10:14:44 ozaki-r Exp $"); 95 96 #include "opt_inet.h" 97 #include "opt_ipsec.h" 98 #include "opt_mrouting.h" 99 #include "opt_net_mpsafe.h" 100 101 #include <sys/param.h> 102 #include <sys/kmem.h> 103 #include <sys/mbuf.h> 104 #include <sys/protosw.h> 105 #include <sys/socket.h> 106 #include <sys/socketvar.h> 107 #include <sys/kauth.h> 108 #ifdef IPSEC 109 #include <sys/domain.h> 110 #endif 111 #include <sys/systm.h> 112 113 #include <net/if.h> 114 #include <net/route.h> 115 #include <net/pfil.h> 116 117 #include <netinet/in.h> 118 #include <netinet/in_systm.h> 119 #include <netinet/ip.h> 120 #include <netinet/in_pcb.h> 121 #include <netinet/in_var.h> 122 #include <netinet/ip_var.h> 123 #include <netinet/ip_private.h> 124 #include <netinet/in_offload.h> 125 #include <netinet/portalgo.h> 126 #include <netinet/udp.h> 127 128 #ifdef INET6 129 #include <netinet6/ip6_var.h> 130 #endif 131 132 #ifdef MROUTING 133 #include <netinet/ip_mroute.h> 134 #endif 135 136 #ifdef IPSEC 137 #include <netipsec/ipsec.h> 138 #include <netipsec/key.h> 139 #endif 140 141 static int ip_pcbopts(struct inpcb *, const struct sockopt *); 142 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 143 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 144 static void ip_mloopback(struct ifnet *, struct mbuf *, 145 const struct sockaddr_in *); 146 147 extern pfil_head_t *inet_pfil_hook; /* XXX */ 148 149 int ip_do_loopback_cksum = 0; 150 151 /* 152 * IP output. The packet in mbuf chain m contains a skeletal IP 153 * header (with len, off, ttl, proto, tos, src, dst). 154 * The mbuf chain containing the packet will be freed. 155 * The mbuf opt, if present, will not be freed. 156 */ 157 int 158 ip_output(struct mbuf *m0, ...) 159 { 160 struct rtentry *rt; 161 struct ip *ip; 162 struct ifnet *ifp; 163 struct mbuf *m = m0; 164 int hlen = sizeof (struct ip); 165 int len, error = 0; 166 struct route iproute; 167 const struct sockaddr_in *dst; 168 struct in_ifaddr *ia; 169 int isbroadcast; 170 struct mbuf *opt; 171 struct route *ro; 172 int flags, sw_csum; 173 u_long mtu; 174 struct ip_moptions *imo; 175 struct socket *so; 176 va_list ap; 177 #ifdef IPSEC 178 struct secpolicy *sp = NULL; 179 #endif 180 bool natt_frag = false; 181 bool rtmtu_nolock; 182 union { 183 struct sockaddr dst; 184 struct sockaddr_in dst4; 185 } u; 186 struct sockaddr *rdst = &u.dst; /* real IP destination, as opposed 187 * to the nexthop 188 */ 189 190 len = 0; 191 va_start(ap, m0); 192 opt = va_arg(ap, struct mbuf *); 193 ro = va_arg(ap, struct route *); 194 flags = va_arg(ap, int); 195 imo = va_arg(ap, struct ip_moptions *); 196 so = va_arg(ap, struct socket *); 197 va_end(ap); 198 199 MCLAIM(m, &ip_tx_mowner); 200 201 KASSERT((m->m_flags & M_PKTHDR) != 0); 202 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 0); 203 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) != 204 (M_CSUM_TCPv4|M_CSUM_UDPv4)); 205 206 if (opt) { 207 m = ip_insertoptions(m, opt, &len); 208 if (len >= sizeof(struct ip)) 209 hlen = len; 210 } 211 ip = mtod(m, struct ip *); 212 213 /* 214 * Fill in IP header. 215 */ 216 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 217 ip->ip_v = IPVERSION; 218 ip->ip_off = htons(0); 219 /* ip->ip_id filled in after we find out source ia */ 220 ip->ip_hl = hlen >> 2; 221 IP_STATINC(IP_STAT_LOCALOUT); 222 } else { 223 hlen = ip->ip_hl << 2; 224 } 225 226 /* 227 * Route packet. 228 */ 229 if (ro == NULL) { 230 memset(&iproute, 0, sizeof(iproute)); 231 ro = &iproute; 232 } 233 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 234 dst = satocsin(rtcache_getdst(ro)); 235 236 /* 237 * If there is a cached route, check that it is to the same 238 * destination and is still up. If not, free it and try again. 239 * The address family should also be checked in case of sharing 240 * the cache with IPv6. 241 */ 242 if (dst && (dst->sin_family != AF_INET || 243 !in_hosteq(dst->sin_addr, ip->ip_dst))) 244 rtcache_free(ro); 245 246 if ((rt = rtcache_validate(ro)) == NULL && 247 (rt = rtcache_update(ro, 1)) == NULL) { 248 dst = &u.dst4; 249 error = rtcache_setdst(ro, &u.dst); 250 if (error != 0) 251 goto bad; 252 } 253 254 /* 255 * If routing to interface only, short circuit routing lookup. 256 */ 257 if (flags & IP_ROUTETOIF) { 258 if ((ia = ifatoia(ifa_ifwithladdr(sintocsa(dst)))) == NULL) { 259 IP_STATINC(IP_STAT_NOROUTE); 260 error = ENETUNREACH; 261 goto bad; 262 } 263 ifp = ia->ia_ifp; 264 mtu = ifp->if_mtu; 265 ip->ip_ttl = 1; 266 isbroadcast = in_broadcast(dst->sin_addr, ifp); 267 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 268 ip->ip_dst.s_addr == INADDR_BROADCAST) && 269 imo != NULL && imo->imo_multicast_ifp != NULL) { 270 ifp = imo->imo_multicast_ifp; 271 mtu = ifp->if_mtu; 272 IFP_TO_IA(ifp, ia); 273 isbroadcast = 0; 274 } else { 275 if (rt == NULL) 276 rt = rtcache_init(ro); 277 if (rt == NULL) { 278 IP_STATINC(IP_STAT_NOROUTE); 279 error = EHOSTUNREACH; 280 goto bad; 281 } 282 ia = ifatoia(rt->rt_ifa); 283 ifp = rt->rt_ifp; 284 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 285 mtu = ifp->if_mtu; 286 rt->rt_use++; 287 if (rt->rt_flags & RTF_GATEWAY) 288 dst = satosin(rt->rt_gateway); 289 if (rt->rt_flags & RTF_HOST) 290 isbroadcast = rt->rt_flags & RTF_BROADCAST; 291 else 292 isbroadcast = in_broadcast(dst->sin_addr, ifp); 293 } 294 rtmtu_nolock = rt && (rt->rt_rmx.rmx_locks & RTV_MTU) == 0; 295 296 if (IN_MULTICAST(ip->ip_dst.s_addr) || 297 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 298 bool inmgroup; 299 300 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 301 M_BCAST : M_MCAST; 302 /* 303 * See if the caller provided any multicast options 304 */ 305 if (imo != NULL) 306 ip->ip_ttl = imo->imo_multicast_ttl; 307 else 308 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 309 310 /* 311 * if we don't know the outgoing ifp yet, we can't generate 312 * output 313 */ 314 if (!ifp) { 315 IP_STATINC(IP_STAT_NOROUTE); 316 error = ENETUNREACH; 317 goto bad; 318 } 319 320 /* 321 * If the packet is multicast or broadcast, confirm that 322 * the outgoing interface can transmit it. 323 */ 324 if (((m->m_flags & M_MCAST) && 325 (ifp->if_flags & IFF_MULTICAST) == 0) || 326 ((m->m_flags & M_BCAST) && 327 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 328 IP_STATINC(IP_STAT_NOROUTE); 329 error = ENETUNREACH; 330 goto bad; 331 } 332 /* 333 * If source address not specified yet, use an address 334 * of outgoing interface. 335 */ 336 if (in_nullhost(ip->ip_src)) { 337 struct in_ifaddr *xia; 338 struct ifaddr *xifa; 339 340 IFP_TO_IA(ifp, xia); 341 if (!xia) { 342 error = EADDRNOTAVAIL; 343 goto bad; 344 } 345 xifa = &xia->ia_ifa; 346 if (xifa->ifa_getifa != NULL) { 347 xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 348 } 349 ip->ip_src = xia->ia_addr.sin_addr; 350 } 351 352 inmgroup = in_multi_group(ip->ip_dst, ifp, flags); 353 if (inmgroup && (imo == NULL || imo->imo_multicast_loop)) { 354 /* 355 * If we belong to the destination multicast group 356 * on the outgoing interface, and the caller did not 357 * forbid loopback, loop back a copy. 358 */ 359 ip_mloopback(ifp, m, &u.dst4); 360 } 361 #ifdef MROUTING 362 else { 363 /* 364 * If we are acting as a multicast router, perform 365 * multicast forwarding as if the packet had just 366 * arrived on the interface to which we are about 367 * to send. The multicast forwarding function 368 * recursively calls this function, using the 369 * IP_FORWARDING flag to prevent infinite recursion. 370 * 371 * Multicasts that are looped back by ip_mloopback(), 372 * above, will be forwarded by the ip_input() routine, 373 * if necessary. 374 */ 375 extern struct socket *ip_mrouter; 376 377 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 378 if (ip_mforward(m, ifp) != 0) { 379 m_freem(m); 380 goto done; 381 } 382 } 383 } 384 #endif 385 /* 386 * Multicasts with a time-to-live of zero may be looped- 387 * back, above, but must not be transmitted on a network. 388 * Also, multicasts addressed to the loopback interface 389 * are not sent -- the above call to ip_mloopback() will 390 * loop back a copy if this host actually belongs to the 391 * destination group on the loopback interface. 392 */ 393 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 394 m_freem(m); 395 goto done; 396 } 397 goto sendit; 398 } 399 400 /* 401 * If source address not specified yet, use address 402 * of outgoing interface. 403 */ 404 if (in_nullhost(ip->ip_src)) { 405 struct ifaddr *xifa; 406 407 xifa = &ia->ia_ifa; 408 if (xifa->ifa_getifa != NULL) 409 ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 410 ip->ip_src = ia->ia_addr.sin_addr; 411 } 412 413 /* 414 * packets with Class-D address as source are not valid per 415 * RFC 1112 416 */ 417 if (IN_MULTICAST(ip->ip_src.s_addr)) { 418 IP_STATINC(IP_STAT_ODROPPED); 419 error = EADDRNOTAVAIL; 420 goto bad; 421 } 422 423 /* 424 * Look for broadcast address and and verify user is allowed to 425 * send such a packet. 426 */ 427 if (isbroadcast) { 428 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 429 error = EADDRNOTAVAIL; 430 goto bad; 431 } 432 if ((flags & IP_ALLOWBROADCAST) == 0) { 433 error = EACCES; 434 goto bad; 435 } 436 /* don't allow broadcast messages to be fragmented */ 437 if (ntohs(ip->ip_len) > ifp->if_mtu) { 438 error = EMSGSIZE; 439 goto bad; 440 } 441 m->m_flags |= M_BCAST; 442 } else 443 m->m_flags &= ~M_BCAST; 444 445 sendit: 446 if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) { 447 if (m->m_pkthdr.len < IP_MINFRAGSIZE) { 448 ip->ip_id = 0; 449 } else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 450 ip->ip_id = ip_newid(ia); 451 } else { 452 453 /* 454 * TSO capable interfaces (typically?) increment 455 * ip_id for each segment. 456 * "allocate" enough ids here to increase the chance 457 * for them to be unique. 458 * 459 * note that the following calculation is not 460 * needed to be precise. wasting some ip_id is fine. 461 */ 462 463 unsigned int segsz = m->m_pkthdr.segsz; 464 unsigned int datasz = ntohs(ip->ip_len) - hlen; 465 unsigned int num = howmany(datasz, segsz); 466 467 ip->ip_id = ip_newid_range(ia, num); 468 } 469 } 470 471 /* 472 * If we're doing Path MTU Discovery, we need to set DF unless 473 * the route's MTU is locked. 474 */ 475 if ((flags & IP_MTUDISC) != 0 && rtmtu_nolock) { 476 ip->ip_off |= htons(IP_DF); 477 } 478 479 #ifdef IPSEC 480 if (ipsec_used) { 481 bool ipsec_done = false; 482 483 /* Perform IPsec processing, if any. */ 484 error = ipsec4_output(m, so, flags, &sp, &mtu, &natt_frag, 485 &ipsec_done); 486 if (error || ipsec_done) 487 goto done; 488 } 489 #endif 490 491 /* 492 * Run through list of hooks for output packets. 493 */ 494 error = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_OUT); 495 if (error) 496 goto done; 497 if (m == NULL) 498 goto done; 499 500 ip = mtod(m, struct ip *); 501 hlen = ip->ip_hl << 2; 502 503 m->m_pkthdr.csum_data |= hlen << 16; 504 505 #if IFA_STATS 506 /* 507 * search for the source address structure to 508 * maintain output statistics. 509 */ 510 INADDR_TO_IA(ip->ip_src, ia); 511 #endif 512 513 /* Maybe skip checksums on loopback interfaces. */ 514 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 515 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 516 } 517 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 518 /* 519 * If small enough for mtu of path, or if using TCP segmentation 520 * offload, can just send directly. 521 */ 522 if (ntohs(ip->ip_len) <= mtu || 523 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 524 const struct sockaddr *sa; 525 526 #if IFA_STATS 527 if (ia) 528 ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len); 529 #endif 530 /* 531 * Always initialize the sum to 0! Some HW assisted 532 * checksumming requires this. 533 */ 534 ip->ip_sum = 0; 535 536 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 537 /* 538 * Perform any checksums that the hardware can't do 539 * for us. 540 * 541 * XXX Does any hardware require the {th,uh}_sum 542 * XXX fields to be 0? 543 */ 544 if (sw_csum & M_CSUM_IPv4) { 545 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 546 ip->ip_sum = in_cksum(m, hlen); 547 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 548 } 549 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 550 if (IN_NEED_CHECKSUM(ifp, 551 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 552 in_delayed_cksum(m); 553 } 554 m->m_pkthdr.csum_flags &= 555 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 556 } 557 } 558 559 sa = (m->m_flags & M_MCAST) ? sintocsa(rdst) : sintocsa(dst); 560 if (__predict_true( 561 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 || 562 (ifp->if_capenable & IFCAP_TSOv4) != 0)) { 563 #ifndef NET_MPSAFE 564 KERNEL_LOCK(1, NULL); 565 #endif 566 error = (*ifp->if_output)(ifp, m, sa, rt); 567 #ifndef NET_MPSAFE 568 KERNEL_UNLOCK_ONE(NULL); 569 #endif 570 } else { 571 error = ip_tso_output(ifp, m, sa, rt); 572 } 573 goto done; 574 } 575 576 /* 577 * We can't use HW checksumming if we're about to 578 * to fragment the packet. 579 * 580 * XXX Some hardware can do this. 581 */ 582 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 583 if (IN_NEED_CHECKSUM(ifp, 584 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 585 in_delayed_cksum(m); 586 } 587 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 588 } 589 590 /* 591 * Too large for interface; fragment if possible. 592 * Must be able to put at least 8 bytes per fragment. 593 */ 594 if (ntohs(ip->ip_off) & IP_DF) { 595 if (flags & IP_RETURNMTU) { 596 struct inpcb *inp; 597 598 KASSERT(so && solocked(so)); 599 inp = sotoinpcb(so); 600 inp->inp_errormtu = mtu; 601 } 602 error = EMSGSIZE; 603 IP_STATINC(IP_STAT_CANTFRAG); 604 goto bad; 605 } 606 607 error = ip_fragment(m, ifp, mtu); 608 if (error) { 609 m = NULL; 610 goto bad; 611 } 612 613 for (; m; m = m0) { 614 m0 = m->m_nextpkt; 615 m->m_nextpkt = 0; 616 if (error) { 617 m_freem(m); 618 continue; 619 } 620 #if IFA_STATS 621 if (ia) 622 ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len); 623 #endif 624 /* 625 * If we get there, the packet has not been handled by 626 * IPsec whereas it should have. Now that it has been 627 * fragmented, re-inject it in ip_output so that IPsec 628 * processing can occur. 629 */ 630 if (natt_frag) { 631 error = ip_output(m, opt, ro, 632 flags | IP_RAWOUTPUT | IP_NOIPNEWID, 633 imo, so); 634 } else { 635 KASSERT((m->m_pkthdr.csum_flags & 636 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 637 #ifndef NET_MPSAFE 638 KERNEL_LOCK(1, NULL); 639 #endif 640 error = (*ifp->if_output)(ifp, m, 641 (m->m_flags & M_MCAST) ? 642 sintocsa(rdst) : sintocsa(dst), rt); 643 #ifndef NET_MPSAFE 644 KERNEL_UNLOCK_ONE(NULL); 645 #endif 646 } 647 } 648 if (error == 0) { 649 IP_STATINC(IP_STAT_FRAGMENTED); 650 } 651 done: 652 if (ro == &iproute) { 653 rtcache_free(&iproute); 654 } 655 #ifdef IPSEC 656 if (sp) { 657 KEY_FREESP(&sp); 658 } 659 #endif 660 return error; 661 bad: 662 m_freem(m); 663 goto done; 664 } 665 666 int 667 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 668 { 669 struct ip *ip, *mhip; 670 struct mbuf *m0; 671 int len, hlen, off; 672 int mhlen, firstlen; 673 struct mbuf **mnext; 674 int sw_csum = m->m_pkthdr.csum_flags; 675 int fragments = 0; 676 int s; 677 int error = 0; 678 679 ip = mtod(m, struct ip *); 680 hlen = ip->ip_hl << 2; 681 if (ifp != NULL) 682 sw_csum &= ~ifp->if_csum_flags_tx; 683 684 len = (mtu - hlen) &~ 7; 685 if (len < 8) { 686 m_freem(m); 687 return (EMSGSIZE); 688 } 689 690 firstlen = len; 691 mnext = &m->m_nextpkt; 692 693 /* 694 * Loop through length of segment after first fragment, 695 * make new header and copy data of each part and link onto chain. 696 */ 697 m0 = m; 698 mhlen = sizeof (struct ip); 699 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 700 MGETHDR(m, M_DONTWAIT, MT_HEADER); 701 if (m == 0) { 702 error = ENOBUFS; 703 IP_STATINC(IP_STAT_ODROPPED); 704 goto sendorfree; 705 } 706 MCLAIM(m, m0->m_owner); 707 *mnext = m; 708 mnext = &m->m_nextpkt; 709 m->m_data += max_linkhdr; 710 mhip = mtod(m, struct ip *); 711 *mhip = *ip; 712 /* we must inherit MCAST and BCAST flags */ 713 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 714 if (hlen > sizeof (struct ip)) { 715 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 716 mhip->ip_hl = mhlen >> 2; 717 } 718 m->m_len = mhlen; 719 mhip->ip_off = ((off - hlen) >> 3) + 720 (ntohs(ip->ip_off) & ~IP_MF); 721 if (ip->ip_off & htons(IP_MF)) 722 mhip->ip_off |= IP_MF; 723 if (off + len >= ntohs(ip->ip_len)) 724 len = ntohs(ip->ip_len) - off; 725 else 726 mhip->ip_off |= IP_MF; 727 HTONS(mhip->ip_off); 728 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 729 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 730 if (m->m_next == 0) { 731 error = ENOBUFS; /* ??? */ 732 IP_STATINC(IP_STAT_ODROPPED); 733 goto sendorfree; 734 } 735 m->m_pkthdr.len = mhlen + len; 736 m->m_pkthdr.rcvif = NULL; 737 mhip->ip_sum = 0; 738 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 739 if (sw_csum & M_CSUM_IPv4) { 740 mhip->ip_sum = in_cksum(m, mhlen); 741 } else { 742 /* 743 * checksum is hw-offloaded or not necessary. 744 */ 745 m->m_pkthdr.csum_flags |= 746 m0->m_pkthdr.csum_flags & M_CSUM_IPv4; 747 m->m_pkthdr.csum_data |= mhlen << 16; 748 KASSERT(!(ifp != NULL && 749 IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) || 750 (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 751 } 752 IP_STATINC(IP_STAT_OFRAGMENTS); 753 fragments++; 754 } 755 /* 756 * Update first fragment by trimming what's been copied out 757 * and updating header, then send each fragment (in order). 758 */ 759 m = m0; 760 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 761 m->m_pkthdr.len = hlen + firstlen; 762 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 763 ip->ip_off |= htons(IP_MF); 764 ip->ip_sum = 0; 765 if (sw_csum & M_CSUM_IPv4) { 766 ip->ip_sum = in_cksum(m, hlen); 767 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 768 } else { 769 /* 770 * checksum is hw-offloaded or not necessary. 771 */ 772 KASSERT(!(ifp != NULL && IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) || 773 (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 774 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 775 sizeof(struct ip)); 776 } 777 sendorfree: 778 /* 779 * If there is no room for all the fragments, don't queue 780 * any of them. 781 */ 782 if (ifp != NULL) { 783 s = splnet(); 784 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 785 error == 0) { 786 error = ENOBUFS; 787 IP_STATINC(IP_STAT_ODROPPED); 788 IFQ_INC_DROPS(&ifp->if_snd); 789 } 790 splx(s); 791 } 792 if (error) { 793 for (m = m0; m; m = m0) { 794 m0 = m->m_nextpkt; 795 m->m_nextpkt = NULL; 796 m_freem(m); 797 } 798 } 799 return (error); 800 } 801 802 /* 803 * Process a delayed payload checksum calculation. 804 */ 805 void 806 in_delayed_cksum(struct mbuf *m) 807 { 808 struct ip *ip; 809 u_int16_t csum, offset; 810 811 ip = mtod(m, struct ip *); 812 offset = ip->ip_hl << 2; 813 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 814 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 815 csum = 0xffff; 816 817 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 818 819 if ((offset + sizeof(u_int16_t)) > m->m_len) { 820 /* This happen when ip options were inserted 821 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 822 m->m_len, offset, ip->ip_p); 823 */ 824 m_copyback(m, offset, sizeof(csum), (void *) &csum); 825 } else 826 *(u_int16_t *)(mtod(m, char *) + offset) = csum; 827 } 828 829 /* 830 * Determine the maximum length of the options to be inserted; 831 * we would far rather allocate too much space rather than too little. 832 */ 833 834 u_int 835 ip_optlen(struct inpcb *inp) 836 { 837 struct mbuf *m = inp->inp_options; 838 839 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) { 840 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 841 } 842 return 0; 843 } 844 845 /* 846 * Insert IP options into preformed packet. 847 * Adjust IP destination as required for IP source routing, 848 * as indicated by a non-zero in_addr at the start of the options. 849 */ 850 static struct mbuf * 851 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 852 { 853 struct ipoption *p = mtod(opt, struct ipoption *); 854 struct mbuf *n; 855 struct ip *ip = mtod(m, struct ip *); 856 unsigned optlen; 857 858 optlen = opt->m_len - sizeof(p->ipopt_dst); 859 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 860 return (m); /* XXX should fail */ 861 if (!in_nullhost(p->ipopt_dst)) 862 ip->ip_dst = p->ipopt_dst; 863 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 864 MGETHDR(n, M_DONTWAIT, MT_HEADER); 865 if (n == 0) 866 return (m); 867 MCLAIM(n, m->m_owner); 868 M_MOVE_PKTHDR(n, m); 869 m->m_len -= sizeof(struct ip); 870 m->m_data += sizeof(struct ip); 871 n->m_next = m; 872 m = n; 873 m->m_len = optlen + sizeof(struct ip); 874 m->m_data += max_linkhdr; 875 bcopy((void *)ip, mtod(m, void *), sizeof(struct ip)); 876 } else { 877 m->m_data -= optlen; 878 m->m_len += optlen; 879 memmove(mtod(m, void *), ip, sizeof(struct ip)); 880 } 881 m->m_pkthdr.len += optlen; 882 ip = mtod(m, struct ip *); 883 bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen); 884 *phlen = sizeof(struct ip) + optlen; 885 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 886 return (m); 887 } 888 889 /* 890 * Copy options from ip to jp, 891 * omitting those not copied during fragmentation. 892 */ 893 int 894 ip_optcopy(struct ip *ip, struct ip *jp) 895 { 896 u_char *cp, *dp; 897 int opt, optlen, cnt; 898 899 cp = (u_char *)(ip + 1); 900 dp = (u_char *)(jp + 1); 901 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 902 for (; cnt > 0; cnt -= optlen, cp += optlen) { 903 opt = cp[0]; 904 if (opt == IPOPT_EOL) 905 break; 906 if (opt == IPOPT_NOP) { 907 /* Preserve for IP mcast tunnel's LSRR alignment. */ 908 *dp++ = IPOPT_NOP; 909 optlen = 1; 910 continue; 911 } 912 913 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp)); 914 optlen = cp[IPOPT_OLEN]; 915 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen < cnt); 916 917 /* Invalid lengths should have been caught by ip_dooptions. */ 918 if (optlen > cnt) 919 optlen = cnt; 920 if (IPOPT_COPIED(opt)) { 921 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 922 dp += optlen; 923 } 924 } 925 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 926 *dp++ = IPOPT_EOL; 927 return (optlen); 928 } 929 930 /* 931 * IP socket option processing. 932 */ 933 int 934 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt) 935 { 936 struct inpcb *inp = sotoinpcb(so); 937 struct ip *ip = &inp->inp_ip; 938 int inpflags = inp->inp_flags; 939 int optval = 0, error = 0; 940 941 if (sopt->sopt_level != IPPROTO_IP) { 942 if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER) 943 return 0; 944 return ENOPROTOOPT; 945 } 946 947 switch (op) { 948 case PRCO_SETOPT: 949 switch (sopt->sopt_name) { 950 case IP_OPTIONS: 951 #ifdef notyet 952 case IP_RETOPTS: 953 #endif 954 error = ip_pcbopts(inp, sopt); 955 break; 956 957 case IP_TOS: 958 case IP_TTL: 959 case IP_MINTTL: 960 case IP_PKTINFO: 961 case IP_RECVOPTS: 962 case IP_RECVRETOPTS: 963 case IP_RECVDSTADDR: 964 case IP_RECVIF: 965 case IP_RECVPKTINFO: 966 case IP_RECVTTL: 967 error = sockopt_getint(sopt, &optval); 968 if (error) 969 break; 970 971 switch (sopt->sopt_name) { 972 case IP_TOS: 973 ip->ip_tos = optval; 974 break; 975 976 case IP_TTL: 977 ip->ip_ttl = optval; 978 break; 979 980 case IP_MINTTL: 981 if (optval > 0 && optval <= MAXTTL) 982 inp->inp_ip_minttl = optval; 983 else 984 error = EINVAL; 985 break; 986 #define OPTSET(bit) \ 987 if (optval) \ 988 inpflags |= bit; \ 989 else \ 990 inpflags &= ~bit; 991 992 case IP_PKTINFO: 993 OPTSET(INP_PKTINFO); 994 break; 995 996 case IP_RECVOPTS: 997 OPTSET(INP_RECVOPTS); 998 break; 999 1000 case IP_RECVPKTINFO: 1001 OPTSET(INP_RECVPKTINFO); 1002 break; 1003 1004 case IP_RECVRETOPTS: 1005 OPTSET(INP_RECVRETOPTS); 1006 break; 1007 1008 case IP_RECVDSTADDR: 1009 OPTSET(INP_RECVDSTADDR); 1010 break; 1011 1012 case IP_RECVIF: 1013 OPTSET(INP_RECVIF); 1014 break; 1015 1016 case IP_RECVTTL: 1017 OPTSET(INP_RECVTTL); 1018 break; 1019 } 1020 break; 1021 #undef OPTSET 1022 1023 case IP_MULTICAST_IF: 1024 case IP_MULTICAST_TTL: 1025 case IP_MULTICAST_LOOP: 1026 case IP_ADD_MEMBERSHIP: 1027 case IP_DROP_MEMBERSHIP: 1028 error = ip_setmoptions(&inp->inp_moptions, sopt); 1029 break; 1030 1031 case IP_PORTRANGE: 1032 error = sockopt_getint(sopt, &optval); 1033 if (error) 1034 break; 1035 1036 switch (optval) { 1037 case IP_PORTRANGE_DEFAULT: 1038 case IP_PORTRANGE_HIGH: 1039 inpflags &= ~(INP_LOWPORT); 1040 break; 1041 1042 case IP_PORTRANGE_LOW: 1043 inpflags |= INP_LOWPORT; 1044 break; 1045 1046 default: 1047 error = EINVAL; 1048 break; 1049 } 1050 break; 1051 1052 case IP_PORTALGO: 1053 error = sockopt_getint(sopt, &optval); 1054 if (error) 1055 break; 1056 1057 error = portalgo_algo_index_select( 1058 (struct inpcb_hdr *)inp, optval); 1059 break; 1060 1061 #if defined(IPSEC) 1062 case IP_IPSEC_POLICY: 1063 if (ipsec_enabled) { 1064 error = ipsec4_set_policy(inp, sopt->sopt_name, 1065 sopt->sopt_data, sopt->sopt_size, 1066 curlwp->l_cred); 1067 break; 1068 } 1069 /*FALLTHROUGH*/ 1070 #endif /* IPSEC */ 1071 1072 default: 1073 error = ENOPROTOOPT; 1074 break; 1075 } 1076 break; 1077 1078 case PRCO_GETOPT: 1079 switch (sopt->sopt_name) { 1080 case IP_OPTIONS: 1081 case IP_RETOPTS: { 1082 struct mbuf *mopts = inp->inp_options; 1083 1084 if (mopts) { 1085 struct mbuf *m; 1086 1087 m = m_copym(mopts, 0, M_COPYALL, M_DONTWAIT); 1088 if (m == NULL) { 1089 error = ENOBUFS; 1090 break; 1091 } 1092 error = sockopt_setmbuf(sopt, m); 1093 } 1094 break; 1095 } 1096 case IP_PKTINFO: 1097 case IP_TOS: 1098 case IP_TTL: 1099 case IP_MINTTL: 1100 case IP_RECVOPTS: 1101 case IP_RECVRETOPTS: 1102 case IP_RECVDSTADDR: 1103 case IP_RECVIF: 1104 case IP_RECVPKTINFO: 1105 case IP_RECVTTL: 1106 case IP_ERRORMTU: 1107 switch (sopt->sopt_name) { 1108 case IP_TOS: 1109 optval = ip->ip_tos; 1110 break; 1111 1112 case IP_TTL: 1113 optval = ip->ip_ttl; 1114 break; 1115 1116 case IP_MINTTL: 1117 optval = inp->inp_ip_minttl; 1118 break; 1119 1120 case IP_ERRORMTU: 1121 optval = inp->inp_errormtu; 1122 break; 1123 1124 #define OPTBIT(bit) (inpflags & bit ? 1 : 0) 1125 1126 case IP_PKTINFO: 1127 optval = OPTBIT(INP_PKTINFO); 1128 break; 1129 1130 case IP_RECVOPTS: 1131 optval = OPTBIT(INP_RECVOPTS); 1132 break; 1133 1134 case IP_RECVPKTINFO: 1135 optval = OPTBIT(INP_RECVPKTINFO); 1136 break; 1137 1138 case IP_RECVRETOPTS: 1139 optval = OPTBIT(INP_RECVRETOPTS); 1140 break; 1141 1142 case IP_RECVDSTADDR: 1143 optval = OPTBIT(INP_RECVDSTADDR); 1144 break; 1145 1146 case IP_RECVIF: 1147 optval = OPTBIT(INP_RECVIF); 1148 break; 1149 1150 case IP_RECVTTL: 1151 optval = OPTBIT(INP_RECVTTL); 1152 break; 1153 } 1154 error = sockopt_setint(sopt, optval); 1155 break; 1156 1157 #if 0 /* defined(IPSEC) */ 1158 case IP_IPSEC_POLICY: 1159 { 1160 struct mbuf *m = NULL; 1161 1162 /* XXX this will return EINVAL as sopt is empty */ 1163 error = ipsec4_get_policy(inp, sopt->sopt_data, 1164 sopt->sopt_size, &m); 1165 if (error == 0) 1166 error = sockopt_setmbuf(sopt, m); 1167 break; 1168 } 1169 #endif /*IPSEC*/ 1170 1171 case IP_MULTICAST_IF: 1172 case IP_MULTICAST_TTL: 1173 case IP_MULTICAST_LOOP: 1174 case IP_ADD_MEMBERSHIP: 1175 case IP_DROP_MEMBERSHIP: 1176 error = ip_getmoptions(inp->inp_moptions, sopt); 1177 break; 1178 1179 case IP_PORTRANGE: 1180 if (inpflags & INP_LOWPORT) 1181 optval = IP_PORTRANGE_LOW; 1182 else 1183 optval = IP_PORTRANGE_DEFAULT; 1184 error = sockopt_setint(sopt, optval); 1185 break; 1186 1187 case IP_PORTALGO: 1188 optval = inp->inp_portalgo; 1189 error = sockopt_setint(sopt, optval); 1190 break; 1191 1192 default: 1193 error = ENOPROTOOPT; 1194 break; 1195 } 1196 break; 1197 } 1198 1199 if (!error) { 1200 inp->inp_flags = inpflags; 1201 } 1202 return error; 1203 } 1204 1205 /* 1206 * Set up IP options in pcb for insertion in output packets. 1207 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1208 * with destination address if source routed. 1209 */ 1210 static int 1211 ip_pcbopts(struct inpcb *inp, const struct sockopt *sopt) 1212 { 1213 struct mbuf *m; 1214 const u_char *cp; 1215 u_char *dp; 1216 int cnt; 1217 1218 /* Turn off any old options. */ 1219 if (inp->inp_options) { 1220 m_free(inp->inp_options); 1221 } 1222 inp->inp_options = NULL; 1223 if ((cnt = sopt->sopt_size) == 0) { 1224 /* Only turning off any previous options. */ 1225 return 0; 1226 } 1227 cp = sopt->sopt_data; 1228 1229 #ifndef __vax__ 1230 if (cnt % sizeof(int32_t)) 1231 return (EINVAL); 1232 #endif 1233 1234 m = m_get(M_DONTWAIT, MT_SOOPTS); 1235 if (m == NULL) 1236 return (ENOBUFS); 1237 1238 dp = mtod(m, u_char *); 1239 memset(dp, 0, sizeof(struct in_addr)); 1240 dp += sizeof(struct in_addr); 1241 m->m_len = sizeof(struct in_addr); 1242 1243 /* 1244 * IP option list according to RFC791. Each option is of the form 1245 * 1246 * [optval] [olen] [(olen - 2) data bytes] 1247 * 1248 * We validate the list and copy options to an mbuf for prepending 1249 * to data packets. The IP first-hop destination address will be 1250 * stored before actual options and is zero if unset. 1251 */ 1252 while (cnt > 0) { 1253 uint8_t optval, olen, offset; 1254 1255 optval = cp[IPOPT_OPTVAL]; 1256 1257 if (optval == IPOPT_EOL || optval == IPOPT_NOP) { 1258 olen = 1; 1259 } else { 1260 if (cnt < IPOPT_OLEN + 1) 1261 goto bad; 1262 1263 olen = cp[IPOPT_OLEN]; 1264 if (olen < IPOPT_OLEN + 1 || olen > cnt) 1265 goto bad; 1266 } 1267 1268 if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) { 1269 /* 1270 * user process specifies route as: 1271 * ->A->B->C->D 1272 * D must be our final destination (but we can't 1273 * check that since we may not have connected yet). 1274 * A is first hop destination, which doesn't appear in 1275 * actual IP option, but is stored before the options. 1276 */ 1277 if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr)) 1278 goto bad; 1279 1280 offset = cp[IPOPT_OFFSET]; 1281 memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1, 1282 sizeof(struct in_addr)); 1283 1284 cp += sizeof(struct in_addr); 1285 cnt -= sizeof(struct in_addr); 1286 olen -= sizeof(struct in_addr); 1287 1288 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1289 goto bad; 1290 1291 memcpy(dp, cp, olen); 1292 dp[IPOPT_OPTVAL] = optval; 1293 dp[IPOPT_OLEN] = olen; 1294 dp[IPOPT_OFFSET] = offset; 1295 break; 1296 } else { 1297 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1298 goto bad; 1299 1300 memcpy(dp, cp, olen); 1301 break; 1302 } 1303 1304 dp += olen; 1305 m->m_len += olen; 1306 1307 if (optval == IPOPT_EOL) 1308 break; 1309 1310 cp += olen; 1311 cnt -= olen; 1312 } 1313 1314 inp->inp_options = m; 1315 return 0; 1316 bad: 1317 (void)m_free(m); 1318 return EINVAL; 1319 } 1320 1321 /* 1322 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1323 */ 1324 static struct ifnet * 1325 ip_multicast_if(struct in_addr *a, int *ifindexp) 1326 { 1327 int ifindex; 1328 struct ifnet *ifp = NULL; 1329 struct in_ifaddr *ia; 1330 1331 if (ifindexp) 1332 *ifindexp = 0; 1333 if (ntohl(a->s_addr) >> 24 == 0) { 1334 ifindex = ntohl(a->s_addr) & 0xffffff; 1335 ifp = if_byindex(ifindex); 1336 if (!ifp) 1337 return NULL; 1338 if (ifindexp) 1339 *ifindexp = ifindex; 1340 } else { 1341 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1342 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1343 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1344 ifp = ia->ia_ifp; 1345 break; 1346 } 1347 } 1348 } 1349 return ifp; 1350 } 1351 1352 static int 1353 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval) 1354 { 1355 u_int tval; 1356 u_char cval; 1357 int error; 1358 1359 if (sopt == NULL) 1360 return EINVAL; 1361 1362 switch (sopt->sopt_size) { 1363 case sizeof(u_char): 1364 error = sockopt_get(sopt, &cval, sizeof(u_char)); 1365 tval = cval; 1366 break; 1367 1368 case sizeof(u_int): 1369 error = sockopt_get(sopt, &tval, sizeof(u_int)); 1370 break; 1371 1372 default: 1373 error = EINVAL; 1374 } 1375 1376 if (error) 1377 return error; 1378 1379 if (tval > maxval) 1380 return EINVAL; 1381 1382 *val = tval; 1383 return 0; 1384 } 1385 1386 static int 1387 ip_get_membership(const struct sockopt *sopt, struct ifnet **ifp, 1388 struct in_addr *ia, bool add) 1389 { 1390 int error; 1391 struct ip_mreq mreq; 1392 1393 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 1394 if (error) 1395 return error; 1396 1397 if (!IN_MULTICAST(mreq.imr_multiaddr.s_addr)) 1398 return EINVAL; 1399 1400 memcpy(ia, &mreq.imr_multiaddr, sizeof(*ia)); 1401 1402 if (in_nullhost(mreq.imr_interface)) { 1403 union { 1404 struct sockaddr dst; 1405 struct sockaddr_in dst4; 1406 } u; 1407 struct route ro; 1408 1409 if (!add) { 1410 *ifp = NULL; 1411 return 0; 1412 } 1413 /* 1414 * If no interface address was provided, use the interface of 1415 * the route to the given multicast address. 1416 */ 1417 struct rtentry *rt; 1418 memset(&ro, 0, sizeof(ro)); 1419 1420 sockaddr_in_init(&u.dst4, ia, 0); 1421 error = rtcache_setdst(&ro, &u.dst); 1422 if (error != 0) 1423 return error; 1424 *ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL; 1425 rtcache_free(&ro); 1426 } else { 1427 *ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1428 if (!add && *ifp == NULL) 1429 return EADDRNOTAVAIL; 1430 } 1431 return 0; 1432 } 1433 1434 /* 1435 * Add a multicast group membership. 1436 * Group must be a valid IP multicast address. 1437 */ 1438 static int 1439 ip_add_membership(struct ip_moptions *imo, const struct sockopt *sopt) 1440 { 1441 struct ifnet *ifp; 1442 struct in_addr ia; 1443 int i, error; 1444 1445 if (sopt->sopt_size == sizeof(struct ip_mreq)) 1446 error = ip_get_membership(sopt, &ifp, &ia, true); 1447 else 1448 #ifdef INET6 1449 error = ip6_get_membership(sopt, &ifp, &ia, sizeof(ia)); 1450 #else 1451 return EINVAL; 1452 #endif 1453 1454 if (error) 1455 return error; 1456 1457 /* 1458 * See if we found an interface, and confirm that it 1459 * supports multicast. 1460 */ 1461 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) 1462 return EADDRNOTAVAIL; 1463 1464 /* 1465 * See if the membership already exists or if all the 1466 * membership slots are full. 1467 */ 1468 for (i = 0; i < imo->imo_num_memberships; ++i) { 1469 if (imo->imo_membership[i]->inm_ifp == ifp && 1470 in_hosteq(imo->imo_membership[i]->inm_addr, ia)) 1471 break; 1472 } 1473 if (i < imo->imo_num_memberships) 1474 return EADDRINUSE; 1475 1476 if (i == IP_MAX_MEMBERSHIPS) 1477 return ETOOMANYREFS; 1478 1479 /* 1480 * Everything looks good; add a new record to the multicast 1481 * address list for the given interface. 1482 */ 1483 if ((imo->imo_membership[i] = in_addmulti(&ia, ifp)) == NULL) 1484 return ENOBUFS; 1485 1486 ++imo->imo_num_memberships; 1487 return 0; 1488 } 1489 1490 /* 1491 * Drop a multicast group membership. 1492 * Group must be a valid IP multicast address. 1493 */ 1494 static int 1495 ip_drop_membership(struct ip_moptions *imo, const struct sockopt *sopt) 1496 { 1497 struct in_addr ia; 1498 struct ifnet *ifp; 1499 int i, error; 1500 1501 if (sopt->sopt_size == sizeof(struct ip_mreq)) 1502 error = ip_get_membership(sopt, &ifp, &ia, false); 1503 else 1504 #ifdef INET6 1505 error = ip6_get_membership(sopt, &ifp, &ia, sizeof(ia)); 1506 #else 1507 return EINVAL; 1508 #endif 1509 1510 if (error) 1511 return error; 1512 1513 /* 1514 * Find the membership in the membership array. 1515 */ 1516 for (i = 0; i < imo->imo_num_memberships; ++i) { 1517 if ((ifp == NULL || 1518 imo->imo_membership[i]->inm_ifp == ifp) && 1519 in_hosteq(imo->imo_membership[i]->inm_addr, ia)) 1520 break; 1521 } 1522 if (i == imo->imo_num_memberships) 1523 return EADDRNOTAVAIL; 1524 1525 /* 1526 * Give up the multicast address record to which the 1527 * membership points. 1528 */ 1529 in_delmulti(imo->imo_membership[i]); 1530 1531 /* 1532 * Remove the gap in the membership array. 1533 */ 1534 for (++i; i < imo->imo_num_memberships; ++i) 1535 imo->imo_membership[i-1] = imo->imo_membership[i]; 1536 --imo->imo_num_memberships; 1537 return 0; 1538 } 1539 1540 /* 1541 * Set the IP multicast options in response to user setsockopt(). 1542 */ 1543 int 1544 ip_setmoptions(struct ip_moptions **pimo, const struct sockopt *sopt) 1545 { 1546 struct ip_moptions *imo = *pimo; 1547 struct in_addr addr; 1548 struct ifnet *ifp; 1549 int ifindex, error = 0; 1550 1551 if (!imo) { 1552 /* 1553 * No multicast option buffer attached to the pcb; 1554 * allocate one and initialize to default values. 1555 */ 1556 imo = kmem_intr_alloc(sizeof(*imo), KM_NOSLEEP); 1557 if (imo == NULL) 1558 return ENOBUFS; 1559 1560 imo->imo_multicast_ifp = NULL; 1561 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1562 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1563 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1564 imo->imo_num_memberships = 0; 1565 *pimo = imo; 1566 } 1567 1568 switch (sopt->sopt_name) { 1569 case IP_MULTICAST_IF: 1570 /* 1571 * Select the interface for outgoing multicast packets. 1572 */ 1573 error = sockopt_get(sopt, &addr, sizeof(addr)); 1574 if (error) 1575 break; 1576 1577 /* 1578 * INADDR_ANY is used to remove a previous selection. 1579 * When no interface is selected, a default one is 1580 * chosen every time a multicast packet is sent. 1581 */ 1582 if (in_nullhost(addr)) { 1583 imo->imo_multicast_ifp = NULL; 1584 break; 1585 } 1586 /* 1587 * The selected interface is identified by its local 1588 * IP address. Find the interface and confirm that 1589 * it supports multicasting. 1590 */ 1591 ifp = ip_multicast_if(&addr, &ifindex); 1592 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1593 error = EADDRNOTAVAIL; 1594 break; 1595 } 1596 imo->imo_multicast_ifp = ifp; 1597 if (ifindex) 1598 imo->imo_multicast_addr = addr; 1599 else 1600 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1601 break; 1602 1603 case IP_MULTICAST_TTL: 1604 /* 1605 * Set the IP time-to-live for outgoing multicast packets. 1606 */ 1607 error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL); 1608 break; 1609 1610 case IP_MULTICAST_LOOP: 1611 /* 1612 * Set the loopback flag for outgoing multicast packets. 1613 * Must be zero or one. 1614 */ 1615 error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1); 1616 break; 1617 1618 case IP_ADD_MEMBERSHIP: /* IPV6_JOIN_GROUP */ 1619 error = ip_add_membership(imo, sopt); 1620 break; 1621 1622 case IP_DROP_MEMBERSHIP: /* IPV6_LEAVE_GROUP */ 1623 error = ip_drop_membership(imo, sopt); 1624 break; 1625 1626 default: 1627 error = EOPNOTSUPP; 1628 break; 1629 } 1630 1631 /* 1632 * If all options have default values, no need to keep the mbuf. 1633 */ 1634 if (imo->imo_multicast_ifp == NULL && 1635 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1636 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1637 imo->imo_num_memberships == 0) { 1638 kmem_free(imo, sizeof(*imo)); 1639 *pimo = NULL; 1640 } 1641 1642 return error; 1643 } 1644 1645 /* 1646 * Return the IP multicast options in response to user getsockopt(). 1647 */ 1648 int 1649 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt) 1650 { 1651 struct in_addr addr; 1652 struct in_ifaddr *ia; 1653 uint8_t optval; 1654 int error = 0; 1655 1656 switch (sopt->sopt_name) { 1657 case IP_MULTICAST_IF: 1658 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1659 addr = zeroin_addr; 1660 else if (imo->imo_multicast_addr.s_addr) { 1661 /* return the value user has set */ 1662 addr = imo->imo_multicast_addr; 1663 } else { 1664 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1665 addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1666 } 1667 error = sockopt_set(sopt, &addr, sizeof(addr)); 1668 break; 1669 1670 case IP_MULTICAST_TTL: 1671 optval = imo ? imo->imo_multicast_ttl 1672 : IP_DEFAULT_MULTICAST_TTL; 1673 1674 error = sockopt_set(sopt, &optval, sizeof(optval)); 1675 break; 1676 1677 case IP_MULTICAST_LOOP: 1678 optval = imo ? imo->imo_multicast_loop 1679 : IP_DEFAULT_MULTICAST_LOOP; 1680 1681 error = sockopt_set(sopt, &optval, sizeof(optval)); 1682 break; 1683 1684 default: 1685 error = EOPNOTSUPP; 1686 } 1687 1688 return error; 1689 } 1690 1691 /* 1692 * Discard the IP multicast options. 1693 */ 1694 void 1695 ip_freemoptions(struct ip_moptions *imo) 1696 { 1697 int i; 1698 1699 if (imo != NULL) { 1700 for (i = 0; i < imo->imo_num_memberships; ++i) 1701 in_delmulti(imo->imo_membership[i]); 1702 kmem_free(imo, sizeof(*imo)); 1703 } 1704 } 1705 1706 /* 1707 * Routine called from ip_output() to loop back a copy of an IP multicast 1708 * packet to the input queue of a specified interface. Note that this 1709 * calls the output routine of the loopback "driver", but with an interface 1710 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1711 */ 1712 static void 1713 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 1714 { 1715 struct ip *ip; 1716 struct mbuf *copym; 1717 1718 copym = m_copypacket(m, M_DONTWAIT); 1719 if (copym != NULL && 1720 (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1721 copym = m_pullup(copym, sizeof(struct ip)); 1722 if (copym == NULL) 1723 return; 1724 /* 1725 * We don't bother to fragment if the IP length is greater 1726 * than the interface's MTU. Can this possibly matter? 1727 */ 1728 ip = mtod(copym, struct ip *); 1729 1730 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1731 in_delayed_cksum(copym); 1732 copym->m_pkthdr.csum_flags &= 1733 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1734 } 1735 1736 ip->ip_sum = 0; 1737 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1738 #ifndef NET_MPSAFE 1739 KERNEL_LOCK(1, NULL); 1740 #endif 1741 (void)looutput(ifp, copym, sintocsa(dst), NULL); 1742 #ifndef NET_MPSAFE 1743 KERNEL_UNLOCK_ONE(NULL); 1744 #endif 1745 } 1746
Indexes created Mon Sep 22 21:09:42 GMT 2025