ip_output.c revision 1.252
1 /* $NetBSD: ip_output.c,v 1.252 2016/04/26 09:30:01 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 1.252 2016/04/26 09:30:01 ozaki-r Exp $"); 95 96 #ifdef _KERNEL_OPT 97 #include "opt_inet.h" 98 #include "opt_ipsec.h" 99 #include "opt_mrouting.h" 100 #include "opt_net_mpsafe.h" 101 #include "opt_mpls.h" 102 #endif 103 104 #include <sys/param.h> 105 #include <sys/kmem.h> 106 #include <sys/mbuf.h> 107 #include <sys/protosw.h> 108 #include <sys/socket.h> 109 #include <sys/socketvar.h> 110 #include <sys/kauth.h> 111 #ifdef IPSEC 112 #include <sys/domain.h> 113 #endif 114 #include <sys/systm.h> 115 116 #include <net/if.h> 117 #include <net/if_types.h> 118 #include <net/route.h> 119 #include <net/pfil.h> 120 121 #include <netinet/in.h> 122 #include <netinet/in_systm.h> 123 #include <netinet/ip.h> 124 #include <netinet/in_pcb.h> 125 #include <netinet/in_var.h> 126 #include <netinet/ip_var.h> 127 #include <netinet/ip_private.h> 128 #include <netinet/in_offload.h> 129 #include <netinet/portalgo.h> 130 #include <netinet/udp.h> 131 132 #ifdef INET6 133 #include <netinet6/ip6_var.h> 134 #endif 135 136 #ifdef MROUTING 137 #include <netinet/ip_mroute.h> 138 #endif 139 140 #ifdef IPSEC 141 #include <netipsec/ipsec.h> 142 #include <netipsec/key.h> 143 #endif 144 145 #ifdef MPLS 146 #include <netmpls/mpls.h> 147 #include <netmpls/mpls_var.h> 148 #endif 149 150 static int ip_pcbopts(struct inpcb *, const struct sockopt *); 151 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 152 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 153 static void ip_mloopback(struct ifnet *, struct mbuf *, 154 const struct sockaddr_in *); 155 156 extern pfil_head_t *inet_pfil_hook; /* XXX */ 157 158 int ip_do_loopback_cksum = 0; 159 160 static int 161 ip_mark_mpls(struct ifnet * const ifp, struct mbuf * const m, struct rtentry *rt) 162 { 163 int error = 0; 164 #ifdef MPLS 165 union mpls_shim msh; 166 167 if (rt == NULL || rt_gettag(rt) == NULL || 168 rt_gettag(rt)->sa_family != AF_MPLS || 169 (m->m_flags & (M_MCAST | M_BCAST)) != 0 || 170 ifp->if_type != IFT_ETHER) 171 return 0; 172 173 msh.s_addr = MPLS_GETSADDR(rt); 174 if (msh.shim.label != MPLS_LABEL_IMPLNULL) { 175 struct m_tag *mtag; 176 /* 177 * XXX tentative solution to tell ether_output 178 * it's MPLS. Need some more efficient solution. 179 */ 180 mtag = m_tag_get(PACKET_TAG_MPLS, 181 sizeof(int) /* dummy */, 182 M_NOWAIT); 183 if (mtag == NULL) 184 return ENOMEM; 185 m_tag_prepend(m, mtag); 186 } 187 #endif 188 return error; 189 } 190 191 /* 192 * Send an IP packet to a host. 193 */ 194 int 195 ip_if_output(struct ifnet * const ifp, struct mbuf * const m, 196 const struct sockaddr * const dst, struct rtentry *rt) 197 { 198 int error = 0; 199 200 if (rt != NULL) { 201 error = rt_check_reject_route(rt, ifp); 202 if (error != 0) { 203 m_freem(m); 204 return error; 205 } 206 } 207 208 error = ip_mark_mpls(ifp, m, rt); 209 if (error != 0) { 210 m_freem(m); 211 return error; 212 } 213 214 #ifndef NET_MPSAFE 215 KERNEL_LOCK(1, NULL); 216 #endif 217 218 error = (*ifp->if_output)(ifp, m, dst, rt); 219 220 #ifndef NET_MPSAFE 221 KERNEL_UNLOCK_ONE(NULL); 222 #endif 223 return error; 224 } 225 226 /* 227 * IP output. The packet in mbuf chain m contains a skeletal IP 228 * header (with len, off, ttl, proto, tos, src, dst). 229 * The mbuf chain containing the packet will be freed. 230 * The mbuf opt, if present, will not be freed. 231 */ 232 int 233 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, int flags, 234 struct ip_moptions *imo, struct socket *so) 235 { 236 struct rtentry *rt; 237 struct ip *ip; 238 struct ifnet *ifp; 239 struct mbuf *m = m0; 240 int hlen = sizeof (struct ip); 241 int len, error = 0; 242 struct route iproute; 243 const struct sockaddr_in *dst; 244 struct in_ifaddr *ia; 245 int isbroadcast; 246 int sw_csum; 247 u_long mtu; 248 #ifdef IPSEC 249 struct secpolicy *sp = NULL; 250 #endif 251 bool natt_frag = false; 252 bool rtmtu_nolock; 253 union { 254 struct sockaddr dst; 255 struct sockaddr_in dst4; 256 } u; 257 struct sockaddr *rdst = &u.dst; /* real IP destination, as opposed 258 * to the nexthop 259 */ 260 261 len = 0; 262 263 MCLAIM(m, &ip_tx_mowner); 264 265 KASSERT((m->m_flags & M_PKTHDR) != 0); 266 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 0); 267 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) != 268 (M_CSUM_TCPv4|M_CSUM_UDPv4)); 269 270 if (opt) { 271 m = ip_insertoptions(m, opt, &len); 272 if (len >= sizeof(struct ip)) 273 hlen = len; 274 } 275 ip = mtod(m, struct ip *); 276 277 /* 278 * Fill in IP header. 279 */ 280 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 281 ip->ip_v = IPVERSION; 282 ip->ip_off = htons(0); 283 /* ip->ip_id filled in after we find out source ia */ 284 ip->ip_hl = hlen >> 2; 285 IP_STATINC(IP_STAT_LOCALOUT); 286 } else { 287 hlen = ip->ip_hl << 2; 288 } 289 290 /* 291 * Route packet. 292 */ 293 if (ro == NULL) { 294 memset(&iproute, 0, sizeof(iproute)); 295 ro = &iproute; 296 } 297 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 298 dst = satocsin(rtcache_getdst(ro)); 299 300 /* 301 * If there is a cached route, check that it is to the same 302 * destination and is still up. If not, free it and try again. 303 * The address family should also be checked in case of sharing 304 * the cache with IPv6. 305 */ 306 if (dst && (dst->sin_family != AF_INET || 307 !in_hosteq(dst->sin_addr, ip->ip_dst))) 308 rtcache_free(ro); 309 310 if ((rt = rtcache_validate(ro)) == NULL && 311 (rt = rtcache_update(ro, 1)) == NULL) { 312 dst = &u.dst4; 313 error = rtcache_setdst(ro, &u.dst); 314 if (error != 0) 315 goto bad; 316 } 317 318 /* 319 * If routing to interface only, short circuit routing lookup. 320 */ 321 if (flags & IP_ROUTETOIF) { 322 if ((ia = ifatoia(ifa_ifwithladdr(sintocsa(dst)))) == NULL) { 323 IP_STATINC(IP_STAT_NOROUTE); 324 error = ENETUNREACH; 325 goto bad; 326 } 327 ifp = ia->ia_ifp; 328 mtu = ifp->if_mtu; 329 ip->ip_ttl = 1; 330 isbroadcast = in_broadcast(dst->sin_addr, ifp); 331 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 332 ip->ip_dst.s_addr == INADDR_BROADCAST) && 333 imo != NULL && imo->imo_multicast_ifp != NULL) { 334 ifp = imo->imo_multicast_ifp; 335 mtu = ifp->if_mtu; 336 IFP_TO_IA(ifp, ia); 337 isbroadcast = 0; 338 } else { 339 if (rt == NULL) 340 rt = rtcache_init(ro); 341 if (rt == NULL) { 342 IP_STATINC(IP_STAT_NOROUTE); 343 error = EHOSTUNREACH; 344 goto bad; 345 } 346 ia = ifatoia(rt->rt_ifa); 347 ifp = rt->rt_ifp; 348 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 349 mtu = ifp->if_mtu; 350 rt->rt_use++; 351 if (rt->rt_flags & RTF_GATEWAY) 352 dst = satosin(rt->rt_gateway); 353 if (rt->rt_flags & RTF_HOST) 354 isbroadcast = rt->rt_flags & RTF_BROADCAST; 355 else 356 isbroadcast = in_broadcast(dst->sin_addr, ifp); 357 } 358 rtmtu_nolock = rt && (rt->rt_rmx.rmx_locks & RTV_MTU) == 0; 359 360 if (IN_MULTICAST(ip->ip_dst.s_addr) || 361 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 362 bool inmgroup; 363 364 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 365 M_BCAST : M_MCAST; 366 /* 367 * See if the caller provided any multicast options 368 */ 369 if (imo != NULL) 370 ip->ip_ttl = imo->imo_multicast_ttl; 371 else 372 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 373 374 /* 375 * if we don't know the outgoing ifp yet, we can't generate 376 * output 377 */ 378 if (!ifp) { 379 IP_STATINC(IP_STAT_NOROUTE); 380 error = ENETUNREACH; 381 goto bad; 382 } 383 384 /* 385 * If the packet is multicast or broadcast, confirm that 386 * the outgoing interface can transmit it. 387 */ 388 if (((m->m_flags & M_MCAST) && 389 (ifp->if_flags & IFF_MULTICAST) == 0) || 390 ((m->m_flags & M_BCAST) && 391 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 392 IP_STATINC(IP_STAT_NOROUTE); 393 error = ENETUNREACH; 394 goto bad; 395 } 396 /* 397 * If source address not specified yet, use an address 398 * of outgoing interface. 399 */ 400 if (in_nullhost(ip->ip_src)) { 401 struct in_ifaddr *xia; 402 struct ifaddr *xifa; 403 404 IFP_TO_IA(ifp, xia); 405 if (!xia) { 406 error = EADDRNOTAVAIL; 407 goto bad; 408 } 409 xifa = &xia->ia_ifa; 410 if (xifa->ifa_getifa != NULL) { 411 xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 412 if (xia == NULL) { 413 error = EADDRNOTAVAIL; 414 goto bad; 415 } 416 } 417 ip->ip_src = xia->ia_addr.sin_addr; 418 } 419 420 inmgroup = in_multi_group(ip->ip_dst, ifp, flags); 421 if (inmgroup && (imo == NULL || imo->imo_multicast_loop)) { 422 /* 423 * If we belong to the destination multicast group 424 * on the outgoing interface, and the caller did not 425 * forbid loopback, loop back a copy. 426 */ 427 ip_mloopback(ifp, m, &u.dst4); 428 } 429 #ifdef MROUTING 430 else { 431 /* 432 * If we are acting as a multicast router, perform 433 * multicast forwarding as if the packet had just 434 * arrived on the interface to which we are about 435 * to send. The multicast forwarding function 436 * recursively calls this function, using the 437 * IP_FORWARDING flag to prevent infinite recursion. 438 * 439 * Multicasts that are looped back by ip_mloopback(), 440 * above, will be forwarded by the ip_input() routine, 441 * if necessary. 442 */ 443 extern struct socket *ip_mrouter; 444 445 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 446 if (ip_mforward(m, ifp) != 0) { 447 m_freem(m); 448 goto done; 449 } 450 } 451 } 452 #endif 453 /* 454 * Multicasts with a time-to-live of zero may be looped- 455 * back, above, but must not be transmitted on a network. 456 * Also, multicasts addressed to the loopback interface 457 * are not sent -- the above call to ip_mloopback() will 458 * loop back a copy if this host actually belongs to the 459 * destination group on the loopback interface. 460 */ 461 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 462 m_freem(m); 463 goto done; 464 } 465 goto sendit; 466 } 467 468 /* 469 * If source address not specified yet, use address 470 * of outgoing interface. 471 */ 472 if (in_nullhost(ip->ip_src)) { 473 struct ifaddr *xifa; 474 475 xifa = &ia->ia_ifa; 476 if (xifa->ifa_getifa != NULL) { 477 ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 478 if (ia == NULL) { 479 error = EADDRNOTAVAIL; 480 goto bad; 481 } 482 } 483 ip->ip_src = ia->ia_addr.sin_addr; 484 } 485 486 /* 487 * packets with Class-D address as source are not valid per 488 * RFC 1112 489 */ 490 if (IN_MULTICAST(ip->ip_src.s_addr)) { 491 IP_STATINC(IP_STAT_ODROPPED); 492 error = EADDRNOTAVAIL; 493 goto bad; 494 } 495 496 /* 497 * Look for broadcast address and and verify user is allowed to 498 * send such a packet. 499 */ 500 if (isbroadcast) { 501 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 502 error = EADDRNOTAVAIL; 503 goto bad; 504 } 505 if ((flags & IP_ALLOWBROADCAST) == 0) { 506 error = EACCES; 507 goto bad; 508 } 509 /* don't allow broadcast messages to be fragmented */ 510 if (ntohs(ip->ip_len) > ifp->if_mtu) { 511 error = EMSGSIZE; 512 goto bad; 513 } 514 m->m_flags |= M_BCAST; 515 } else 516 m->m_flags &= ~M_BCAST; 517 518 sendit: 519 if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) { 520 if (m->m_pkthdr.len < IP_MINFRAGSIZE) { 521 ip->ip_id = 0; 522 } else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 523 ip->ip_id = ip_newid(ia); 524 } else { 525 526 /* 527 * TSO capable interfaces (typically?) increment 528 * ip_id for each segment. 529 * "allocate" enough ids here to increase the chance 530 * for them to be unique. 531 * 532 * note that the following calculation is not 533 * needed to be precise. wasting some ip_id is fine. 534 */ 535 536 unsigned int segsz = m->m_pkthdr.segsz; 537 unsigned int datasz = ntohs(ip->ip_len) - hlen; 538 unsigned int num = howmany(datasz, segsz); 539 540 ip->ip_id = ip_newid_range(ia, num); 541 } 542 } 543 544 /* 545 * If we're doing Path MTU Discovery, we need to set DF unless 546 * the route's MTU is locked. 547 */ 548 if ((flags & IP_MTUDISC) != 0 && rtmtu_nolock) { 549 ip->ip_off |= htons(IP_DF); 550 } 551 552 #ifdef IPSEC 553 if (ipsec_used) { 554 bool ipsec_done = false; 555 556 /* Perform IPsec processing, if any. */ 557 error = ipsec4_output(m, so, flags, &sp, &mtu, &natt_frag, 558 &ipsec_done); 559 if (error || ipsec_done) 560 goto done; 561 } 562 #endif 563 564 /* 565 * Run through list of hooks for output packets. 566 */ 567 error = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_OUT); 568 if (error) 569 goto done; 570 if (m == NULL) 571 goto done; 572 573 ip = mtod(m, struct ip *); 574 hlen = ip->ip_hl << 2; 575 576 m->m_pkthdr.csum_data |= hlen << 16; 577 578 #if IFA_STATS 579 /* 580 * search for the source address structure to 581 * maintain output statistics. 582 */ 583 INADDR_TO_IA(ip->ip_src, ia); 584 #endif 585 586 /* Maybe skip checksums on loopback interfaces. */ 587 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 588 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 589 } 590 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 591 /* 592 * If small enough for mtu of path, or if using TCP segmentation 593 * offload, can just send directly. 594 */ 595 if (ntohs(ip->ip_len) <= mtu || 596 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 597 const struct sockaddr *sa; 598 599 #if IFA_STATS 600 if (ia) 601 ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len); 602 #endif 603 /* 604 * Always initialize the sum to 0! Some HW assisted 605 * checksumming requires this. 606 */ 607 ip->ip_sum = 0; 608 609 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 610 /* 611 * Perform any checksums that the hardware can't do 612 * for us. 613 * 614 * XXX Does any hardware require the {th,uh}_sum 615 * XXX fields to be 0? 616 */ 617 if (sw_csum & M_CSUM_IPv4) { 618 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 619 ip->ip_sum = in_cksum(m, hlen); 620 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 621 } 622 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 623 if (IN_NEED_CHECKSUM(ifp, 624 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 625 in_delayed_cksum(m); 626 } 627 m->m_pkthdr.csum_flags &= 628 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 629 } 630 } 631 632 sa = (m->m_flags & M_MCAST) ? sintocsa(rdst) : sintocsa(dst); 633 if (__predict_true( 634 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 || 635 (ifp->if_capenable & IFCAP_TSOv4) != 0)) { 636 error = ip_if_output(ifp, m, sa, rt); 637 } else { 638 error = ip_tso_output(ifp, m, sa, rt); 639 } 640 goto done; 641 } 642 643 /* 644 * We can't use HW checksumming if we're about to 645 * to fragment the packet. 646 * 647 * XXX Some hardware can do this. 648 */ 649 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 650 if (IN_NEED_CHECKSUM(ifp, 651 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 652 in_delayed_cksum(m); 653 } 654 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 655 } 656 657 /* 658 * Too large for interface; fragment if possible. 659 * Must be able to put at least 8 bytes per fragment. 660 */ 661 if (ntohs(ip->ip_off) & IP_DF) { 662 if (flags & IP_RETURNMTU) { 663 struct inpcb *inp; 664 665 KASSERT(so && solocked(so)); 666 inp = sotoinpcb(so); 667 inp->inp_errormtu = mtu; 668 } 669 error = EMSGSIZE; 670 IP_STATINC(IP_STAT_CANTFRAG); 671 goto bad; 672 } 673 674 error = ip_fragment(m, ifp, mtu); 675 if (error) { 676 m = NULL; 677 goto bad; 678 } 679 680 for (; m; m = m0) { 681 m0 = m->m_nextpkt; 682 m->m_nextpkt = 0; 683 if (error) { 684 m_freem(m); 685 continue; 686 } 687 #if IFA_STATS 688 if (ia) 689 ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len); 690 #endif 691 /* 692 * If we get there, the packet has not been handled by 693 * IPsec whereas it should have. Now that it has been 694 * fragmented, re-inject it in ip_output so that IPsec 695 * processing can occur. 696 */ 697 if (natt_frag) { 698 error = ip_output(m, opt, ro, 699 flags | IP_RAWOUTPUT | IP_NOIPNEWID, 700 imo, so); 701 } else { 702 KASSERT((m->m_pkthdr.csum_flags & 703 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 704 error = ip_if_output(ifp, m, 705 (m->m_flags & M_MCAST) ? 706 sintocsa(rdst) : sintocsa(dst), rt); 707 } 708 } 709 if (error == 0) { 710 IP_STATINC(IP_STAT_FRAGMENTED); 711 } 712 done: 713 if (ro == &iproute) { 714 rtcache_free(&iproute); 715 } 716 #ifdef IPSEC 717 if (sp) { 718 KEY_FREESP(&sp); 719 } 720 #endif 721 return error; 722 bad: 723 m_freem(m); 724 goto done; 725 } 726 727 int 728 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 729 { 730 struct ip *ip, *mhip; 731 struct mbuf *m0; 732 int len, hlen, off; 733 int mhlen, firstlen; 734 struct mbuf **mnext; 735 int sw_csum = m->m_pkthdr.csum_flags; 736 int fragments = 0; 737 int s; 738 int error = 0; 739 740 ip = mtod(m, struct ip *); 741 hlen = ip->ip_hl << 2; 742 if (ifp != NULL) 743 sw_csum &= ~ifp->if_csum_flags_tx; 744 745 len = (mtu - hlen) &~ 7; 746 if (len < 8) { 747 m_freem(m); 748 return (EMSGSIZE); 749 } 750 751 firstlen = len; 752 mnext = &m->m_nextpkt; 753 754 /* 755 * Loop through length of segment after first fragment, 756 * make new header and copy data of each part and link onto chain. 757 */ 758 m0 = m; 759 mhlen = sizeof (struct ip); 760 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 761 MGETHDR(m, M_DONTWAIT, MT_HEADER); 762 if (m == 0) { 763 error = ENOBUFS; 764 IP_STATINC(IP_STAT_ODROPPED); 765 goto sendorfree; 766 } 767 MCLAIM(m, m0->m_owner); 768 *mnext = m; 769 mnext = &m->m_nextpkt; 770 m->m_data += max_linkhdr; 771 mhip = mtod(m, struct ip *); 772 *mhip = *ip; 773 /* we must inherit MCAST and BCAST flags */ 774 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 775 if (hlen > sizeof (struct ip)) { 776 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 777 mhip->ip_hl = mhlen >> 2; 778 } 779 m->m_len = mhlen; 780 mhip->ip_off = ((off - hlen) >> 3) + 781 (ntohs(ip->ip_off) & ~IP_MF); 782 if (ip->ip_off & htons(IP_MF)) 783 mhip->ip_off |= IP_MF; 784 if (off + len >= ntohs(ip->ip_len)) 785 len = ntohs(ip->ip_len) - off; 786 else 787 mhip->ip_off |= IP_MF; 788 HTONS(mhip->ip_off); 789 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 790 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 791 if (m->m_next == 0) { 792 error = ENOBUFS; /* ??? */ 793 IP_STATINC(IP_STAT_ODROPPED); 794 goto sendorfree; 795 } 796 m->m_pkthdr.len = mhlen + len; 797 m->m_pkthdr.rcvif = NULL; 798 mhip->ip_sum = 0; 799 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 800 if (sw_csum & M_CSUM_IPv4) { 801 mhip->ip_sum = in_cksum(m, mhlen); 802 } else { 803 /* 804 * checksum is hw-offloaded or not necessary. 805 */ 806 m->m_pkthdr.csum_flags |= 807 m0->m_pkthdr.csum_flags & M_CSUM_IPv4; 808 m->m_pkthdr.csum_data |= mhlen << 16; 809 KASSERT(!(ifp != NULL && 810 IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) || 811 (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 812 } 813 IP_STATINC(IP_STAT_OFRAGMENTS); 814 fragments++; 815 } 816 /* 817 * Update first fragment by trimming what's been copied out 818 * and updating header, then send each fragment (in order). 819 */ 820 m = m0; 821 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 822 m->m_pkthdr.len = hlen + firstlen; 823 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 824 ip->ip_off |= htons(IP_MF); 825 ip->ip_sum = 0; 826 if (sw_csum & M_CSUM_IPv4) { 827 ip->ip_sum = in_cksum(m, hlen); 828 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 829 } else { 830 /* 831 * checksum is hw-offloaded or not necessary. 832 */ 833 KASSERT(!(ifp != NULL && IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) || 834 (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 835 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 836 sizeof(struct ip)); 837 } 838 sendorfree: 839 /* 840 * If there is no room for all the fragments, don't queue 841 * any of them. 842 */ 843 if (ifp != NULL) { 844 s = splnet(); 845 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 846 error == 0) { 847 error = ENOBUFS; 848 IP_STATINC(IP_STAT_ODROPPED); 849 IFQ_INC_DROPS(&ifp->if_snd); 850 } 851 splx(s); 852 } 853 if (error) { 854 for (m = m0; m; m = m0) { 855 m0 = m->m_nextpkt; 856 m->m_nextpkt = NULL; 857 m_freem(m); 858 } 859 } 860 return (error); 861 } 862 863 /* 864 * Process a delayed payload checksum calculation. 865 */ 866 void 867 in_delayed_cksum(struct mbuf *m) 868 { 869 struct ip *ip; 870 u_int16_t csum, offset; 871 872 ip = mtod(m, struct ip *); 873 offset = ip->ip_hl << 2; 874 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 875 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 876 csum = 0xffff; 877 878 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 879 880 if ((offset + sizeof(u_int16_t)) > m->m_len) { 881 /* This happen when ip options were inserted 882 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 883 m->m_len, offset, ip->ip_p); 884 */ 885 m_copyback(m, offset, sizeof(csum), (void *) &csum); 886 } else 887 *(u_int16_t *)(mtod(m, char *) + offset) = csum; 888 } 889 890 /* 891 * Determine the maximum length of the options to be inserted; 892 * we would far rather allocate too much space rather than too little. 893 */ 894 895 u_int 896 ip_optlen(struct inpcb *inp) 897 { 898 struct mbuf *m = inp->inp_options; 899 900 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) { 901 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 902 } 903 return 0; 904 } 905 906 /* 907 * Insert IP options into preformed packet. 908 * Adjust IP destination as required for IP source routing, 909 * as indicated by a non-zero in_addr at the start of the options. 910 */ 911 static struct mbuf * 912 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 913 { 914 struct ipoption *p = mtod(opt, struct ipoption *); 915 struct mbuf *n; 916 struct ip *ip = mtod(m, struct ip *); 917 unsigned optlen; 918 919 optlen = opt->m_len - sizeof(p->ipopt_dst); 920 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 921 return (m); /* XXX should fail */ 922 if (!in_nullhost(p->ipopt_dst)) 923 ip->ip_dst = p->ipopt_dst; 924 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 925 MGETHDR(n, M_DONTWAIT, MT_HEADER); 926 if (n == 0) 927 return (m); 928 MCLAIM(n, m->m_owner); 929 M_MOVE_PKTHDR(n, m); 930 m->m_len -= sizeof(struct ip); 931 m->m_data += sizeof(struct ip); 932 n->m_next = m; 933 m = n; 934 m->m_len = optlen + sizeof(struct ip); 935 m->m_data += max_linkhdr; 936 bcopy((void *)ip, mtod(m, void *), sizeof(struct ip)); 937 } else { 938 m->m_data -= optlen; 939 m->m_len += optlen; 940 memmove(mtod(m, void *), ip, sizeof(struct ip)); 941 } 942 m->m_pkthdr.len += optlen; 943 ip = mtod(m, struct ip *); 944 bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen); 945 *phlen = sizeof(struct ip) + optlen; 946 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 947 return (m); 948 } 949 950 /* 951 * Copy options from ip to jp, 952 * omitting those not copied during fragmentation. 953 */ 954 int 955 ip_optcopy(struct ip *ip, struct ip *jp) 956 { 957 u_char *cp, *dp; 958 int opt, optlen, cnt; 959 960 cp = (u_char *)(ip + 1); 961 dp = (u_char *)(jp + 1); 962 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 963 for (; cnt > 0; cnt -= optlen, cp += optlen) { 964 opt = cp[0]; 965 if (opt == IPOPT_EOL) 966 break; 967 if (opt == IPOPT_NOP) { 968 /* Preserve for IP mcast tunnel's LSRR alignment. */ 969 *dp++ = IPOPT_NOP; 970 optlen = 1; 971 continue; 972 } 973 974 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp)); 975 optlen = cp[IPOPT_OLEN]; 976 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen < cnt); 977 978 /* Invalid lengths should have been caught by ip_dooptions. */ 979 if (optlen > cnt) 980 optlen = cnt; 981 if (IPOPT_COPIED(opt)) { 982 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 983 dp += optlen; 984 } 985 } 986 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 987 *dp++ = IPOPT_EOL; 988 return (optlen); 989 } 990 991 /* 992 * IP socket option processing. 993 */ 994 int 995 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt) 996 { 997 struct inpcb *inp = sotoinpcb(so); 998 struct ip *ip = &inp->inp_ip; 999 int inpflags = inp->inp_flags; 1000 int optval = 0, error = 0; 1001 1002 if (sopt->sopt_level != IPPROTO_IP) { 1003 if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER) 1004 return 0; 1005 return ENOPROTOOPT; 1006 } 1007 1008 switch (op) { 1009 case PRCO_SETOPT: 1010 switch (sopt->sopt_name) { 1011 case IP_OPTIONS: 1012 #ifdef notyet 1013 case IP_RETOPTS: 1014 #endif 1015 error = ip_pcbopts(inp, sopt); 1016 break; 1017 1018 case IP_TOS: 1019 case IP_TTL: 1020 case IP_MINTTL: 1021 case IP_PKTINFO: 1022 case IP_RECVOPTS: 1023 case IP_RECVRETOPTS: 1024 case IP_RECVDSTADDR: 1025 case IP_RECVIF: 1026 case IP_RECVPKTINFO: 1027 case IP_RECVTTL: 1028 error = sockopt_getint(sopt, &optval); 1029 if (error) 1030 break; 1031 1032 switch (sopt->sopt_name) { 1033 case IP_TOS: 1034 ip->ip_tos = optval; 1035 break; 1036 1037 case IP_TTL: 1038 ip->ip_ttl = optval; 1039 break; 1040 1041 case IP_MINTTL: 1042 if (optval > 0 && optval <= MAXTTL) 1043 inp->inp_ip_minttl = optval; 1044 else 1045 error = EINVAL; 1046 break; 1047 #define OPTSET(bit) \ 1048 if (optval) \ 1049 inpflags |= bit; \ 1050 else \ 1051 inpflags &= ~bit; 1052 1053 case IP_PKTINFO: 1054 OPTSET(INP_PKTINFO); 1055 break; 1056 1057 case IP_RECVOPTS: 1058 OPTSET(INP_RECVOPTS); 1059 break; 1060 1061 case IP_RECVPKTINFO: 1062 OPTSET(INP_RECVPKTINFO); 1063 break; 1064 1065 case IP_RECVRETOPTS: 1066 OPTSET(INP_RECVRETOPTS); 1067 break; 1068 1069 case IP_RECVDSTADDR: 1070 OPTSET(INP_RECVDSTADDR); 1071 break; 1072 1073 case IP_RECVIF: 1074 OPTSET(INP_RECVIF); 1075 break; 1076 1077 case IP_RECVTTL: 1078 OPTSET(INP_RECVTTL); 1079 break; 1080 } 1081 break; 1082 #undef OPTSET 1083 1084 case IP_MULTICAST_IF: 1085 case IP_MULTICAST_TTL: 1086 case IP_MULTICAST_LOOP: 1087 case IP_ADD_MEMBERSHIP: 1088 case IP_DROP_MEMBERSHIP: 1089 error = ip_setmoptions(&inp->inp_moptions, sopt); 1090 break; 1091 1092 case IP_PORTRANGE: 1093 error = sockopt_getint(sopt, &optval); 1094 if (error) 1095 break; 1096 1097 switch (optval) { 1098 case IP_PORTRANGE_DEFAULT: 1099 case IP_PORTRANGE_HIGH: 1100 inpflags &= ~(INP_LOWPORT); 1101 break; 1102 1103 case IP_PORTRANGE_LOW: 1104 inpflags |= INP_LOWPORT; 1105 break; 1106 1107 default: 1108 error = EINVAL; 1109 break; 1110 } 1111 break; 1112 1113 case IP_PORTALGO: 1114 error = sockopt_getint(sopt, &optval); 1115 if (error) 1116 break; 1117 1118 error = portalgo_algo_index_select( 1119 (struct inpcb_hdr *)inp, optval); 1120 break; 1121 1122 #if defined(IPSEC) 1123 case IP_IPSEC_POLICY: 1124 if (ipsec_enabled) { 1125 error = ipsec4_set_policy(inp, sopt->sopt_name, 1126 sopt->sopt_data, sopt->sopt_size, 1127 curlwp->l_cred); 1128 break; 1129 } 1130 /*FALLTHROUGH*/ 1131 #endif /* IPSEC */ 1132 1133 default: 1134 error = ENOPROTOOPT; 1135 break; 1136 } 1137 break; 1138 1139 case PRCO_GETOPT: 1140 switch (sopt->sopt_name) { 1141 case IP_OPTIONS: 1142 case IP_RETOPTS: { 1143 struct mbuf *mopts = inp->inp_options; 1144 1145 if (mopts) { 1146 struct mbuf *m; 1147 1148 m = m_copym(mopts, 0, M_COPYALL, M_DONTWAIT); 1149 if (m == NULL) { 1150 error = ENOBUFS; 1151 break; 1152 } 1153 error = sockopt_setmbuf(sopt, m); 1154 } 1155 break; 1156 } 1157 case IP_PKTINFO: 1158 case IP_TOS: 1159 case IP_TTL: 1160 case IP_MINTTL: 1161 case IP_RECVOPTS: 1162 case IP_RECVRETOPTS: 1163 case IP_RECVDSTADDR: 1164 case IP_RECVIF: 1165 case IP_RECVPKTINFO: 1166 case IP_RECVTTL: 1167 case IP_ERRORMTU: 1168 switch (sopt->sopt_name) { 1169 case IP_TOS: 1170 optval = ip->ip_tos; 1171 break; 1172 1173 case IP_TTL: 1174 optval = ip->ip_ttl; 1175 break; 1176 1177 case IP_MINTTL: 1178 optval = inp->inp_ip_minttl; 1179 break; 1180 1181 case IP_ERRORMTU: 1182 optval = inp->inp_errormtu; 1183 break; 1184 1185 #define OPTBIT(bit) (inpflags & bit ? 1 : 0) 1186 1187 case IP_PKTINFO: 1188 optval = OPTBIT(INP_PKTINFO); 1189 break; 1190 1191 case IP_RECVOPTS: 1192 optval = OPTBIT(INP_RECVOPTS); 1193 break; 1194 1195 case IP_RECVPKTINFO: 1196 optval = OPTBIT(INP_RECVPKTINFO); 1197 break; 1198 1199 case IP_RECVRETOPTS: 1200 optval = OPTBIT(INP_RECVRETOPTS); 1201 break; 1202 1203 case IP_RECVDSTADDR: 1204 optval = OPTBIT(INP_RECVDSTADDR); 1205 break; 1206 1207 case IP_RECVIF: 1208 optval = OPTBIT(INP_RECVIF); 1209 break; 1210 1211 case IP_RECVTTL: 1212 optval = OPTBIT(INP_RECVTTL); 1213 break; 1214 } 1215 error = sockopt_setint(sopt, optval); 1216 break; 1217 1218 #if 0 /* defined(IPSEC) */ 1219 case IP_IPSEC_POLICY: 1220 { 1221 struct mbuf *m = NULL; 1222 1223 /* XXX this will return EINVAL as sopt is empty */ 1224 error = ipsec4_get_policy(inp, sopt->sopt_data, 1225 sopt->sopt_size, &m); 1226 if (error == 0) 1227 error = sockopt_setmbuf(sopt, m); 1228 break; 1229 } 1230 #endif /*IPSEC*/ 1231 1232 case IP_MULTICAST_IF: 1233 case IP_MULTICAST_TTL: 1234 case IP_MULTICAST_LOOP: 1235 case IP_ADD_MEMBERSHIP: 1236 case IP_DROP_MEMBERSHIP: 1237 error = ip_getmoptions(inp->inp_moptions, sopt); 1238 break; 1239 1240 case IP_PORTRANGE: 1241 if (inpflags & INP_LOWPORT) 1242 optval = IP_PORTRANGE_LOW; 1243 else 1244 optval = IP_PORTRANGE_DEFAULT; 1245 error = sockopt_setint(sopt, optval); 1246 break; 1247 1248 case IP_PORTALGO: 1249 optval = inp->inp_portalgo; 1250 error = sockopt_setint(sopt, optval); 1251 break; 1252 1253 default: 1254 error = ENOPROTOOPT; 1255 break; 1256 } 1257 break; 1258 } 1259 1260 if (!error) { 1261 inp->inp_flags = inpflags; 1262 } 1263 return error; 1264 } 1265 1266 /* 1267 * Set up IP options in pcb for insertion in output packets. 1268 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1269 * with destination address if source routed. 1270 */ 1271 static int 1272 ip_pcbopts(struct inpcb *inp, const struct sockopt *sopt) 1273 { 1274 struct mbuf *m; 1275 const u_char *cp; 1276 u_char *dp; 1277 int cnt; 1278 1279 /* Turn off any old options. */ 1280 if (inp->inp_options) { 1281 m_free(inp->inp_options); 1282 } 1283 inp->inp_options = NULL; 1284 if ((cnt = sopt->sopt_size) == 0) { 1285 /* Only turning off any previous options. */ 1286 return 0; 1287 } 1288 cp = sopt->sopt_data; 1289 1290 #ifndef __vax__ 1291 if (cnt % sizeof(int32_t)) 1292 return (EINVAL); 1293 #endif 1294 1295 m = m_get(M_DONTWAIT, MT_SOOPTS); 1296 if (m == NULL) 1297 return (ENOBUFS); 1298 1299 dp = mtod(m, u_char *); 1300 memset(dp, 0, sizeof(struct in_addr)); 1301 dp += sizeof(struct in_addr); 1302 m->m_len = sizeof(struct in_addr); 1303 1304 /* 1305 * IP option list according to RFC791. Each option is of the form 1306 * 1307 * [optval] [olen] [(olen - 2) data bytes] 1308 * 1309 * We validate the list and copy options to an mbuf for prepending 1310 * to data packets. The IP first-hop destination address will be 1311 * stored before actual options and is zero if unset. 1312 */ 1313 while (cnt > 0) { 1314 uint8_t optval, olen, offset; 1315 1316 optval = cp[IPOPT_OPTVAL]; 1317 1318 if (optval == IPOPT_EOL || optval == IPOPT_NOP) { 1319 olen = 1; 1320 } else { 1321 if (cnt < IPOPT_OLEN + 1) 1322 goto bad; 1323 1324 olen = cp[IPOPT_OLEN]; 1325 if (olen < IPOPT_OLEN + 1 || olen > cnt) 1326 goto bad; 1327 } 1328 1329 if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) { 1330 /* 1331 * user process specifies route as: 1332 * ->A->B->C->D 1333 * D must be our final destination (but we can't 1334 * check that since we may not have connected yet). 1335 * A is first hop destination, which doesn't appear in 1336 * actual IP option, but is stored before the options. 1337 */ 1338 if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr)) 1339 goto bad; 1340 1341 offset = cp[IPOPT_OFFSET]; 1342 memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1, 1343 sizeof(struct in_addr)); 1344 1345 cp += sizeof(struct in_addr); 1346 cnt -= sizeof(struct in_addr); 1347 olen -= sizeof(struct in_addr); 1348 1349 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1350 goto bad; 1351 1352 memcpy(dp, cp, olen); 1353 dp[IPOPT_OPTVAL] = optval; 1354 dp[IPOPT_OLEN] = olen; 1355 dp[IPOPT_OFFSET] = offset; 1356 break; 1357 } else { 1358 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1359 goto bad; 1360 1361 memcpy(dp, cp, olen); 1362 break; 1363 } 1364 1365 dp += olen; 1366 m->m_len += olen; 1367 1368 if (optval == IPOPT_EOL) 1369 break; 1370 1371 cp += olen; 1372 cnt -= olen; 1373 } 1374 1375 inp->inp_options = m; 1376 return 0; 1377 bad: 1378 (void)m_free(m); 1379 return EINVAL; 1380 } 1381 1382 /* 1383 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1384 */ 1385 static struct ifnet * 1386 ip_multicast_if(struct in_addr *a, int *ifindexp) 1387 { 1388 int ifindex; 1389 struct ifnet *ifp = NULL; 1390 struct in_ifaddr *ia; 1391 1392 if (ifindexp) 1393 *ifindexp = 0; 1394 if (ntohl(a->s_addr) >> 24 == 0) { 1395 ifindex = ntohl(a->s_addr) & 0xffffff; 1396 ifp = if_byindex(ifindex); 1397 if (!ifp) 1398 return NULL; 1399 if (ifindexp) 1400 *ifindexp = ifindex; 1401 } else { 1402 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1403 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1404 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1405 ifp = ia->ia_ifp; 1406 break; 1407 } 1408 } 1409 } 1410 return ifp; 1411 } 1412 1413 static int 1414 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval) 1415 { 1416 u_int tval; 1417 u_char cval; 1418 int error; 1419 1420 if (sopt == NULL) 1421 return EINVAL; 1422 1423 switch (sopt->sopt_size) { 1424 case sizeof(u_char): 1425 error = sockopt_get(sopt, &cval, sizeof(u_char)); 1426 tval = cval; 1427 break; 1428 1429 case sizeof(u_int): 1430 error = sockopt_get(sopt, &tval, sizeof(u_int)); 1431 break; 1432 1433 default: 1434 error = EINVAL; 1435 } 1436 1437 if (error) 1438 return error; 1439 1440 if (tval > maxval) 1441 return EINVAL; 1442 1443 *val = tval; 1444 return 0; 1445 } 1446 1447 static int 1448 ip_get_membership(const struct sockopt *sopt, struct ifnet **ifp, 1449 struct in_addr *ia, bool add) 1450 { 1451 int error; 1452 struct ip_mreq mreq; 1453 1454 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 1455 if (error) 1456 return error; 1457 1458 if (!IN_MULTICAST(mreq.imr_multiaddr.s_addr)) 1459 return EINVAL; 1460 1461 memcpy(ia, &mreq.imr_multiaddr, sizeof(*ia)); 1462 1463 if (in_nullhost(mreq.imr_interface)) { 1464 union { 1465 struct sockaddr dst; 1466 struct sockaddr_in dst4; 1467 } u; 1468 struct route ro; 1469 1470 if (!add) { 1471 *ifp = NULL; 1472 return 0; 1473 } 1474 /* 1475 * If no interface address was provided, use the interface of 1476 * the route to the given multicast address. 1477 */ 1478 struct rtentry *rt; 1479 memset(&ro, 0, sizeof(ro)); 1480 1481 sockaddr_in_init(&u.dst4, ia, 0); 1482 error = rtcache_setdst(&ro, &u.dst); 1483 if (error != 0) 1484 return error; 1485 *ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL; 1486 rtcache_free(&ro); 1487 } else { 1488 *ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1489 if (!add && *ifp == NULL) 1490 return EADDRNOTAVAIL; 1491 } 1492 return 0; 1493 } 1494 1495 /* 1496 * Add a multicast group membership. 1497 * Group must be a valid IP multicast address. 1498 */ 1499 static int 1500 ip_add_membership(struct ip_moptions *imo, const struct sockopt *sopt) 1501 { 1502 struct ifnet *ifp; 1503 struct in_addr ia; 1504 int i, error; 1505 1506 if (sopt->sopt_size == sizeof(struct ip_mreq)) 1507 error = ip_get_membership(sopt, &ifp, &ia, true); 1508 else 1509 #ifdef INET6 1510 error = ip6_get_membership(sopt, &ifp, &ia, sizeof(ia)); 1511 #else 1512 return EINVAL; 1513 #endif 1514 1515 if (error) 1516 return error; 1517 1518 /* 1519 * See if we found an interface, and confirm that it 1520 * supports multicast. 1521 */ 1522 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) 1523 return EADDRNOTAVAIL; 1524 1525 /* 1526 * See if the membership already exists or if all the 1527 * membership slots are full. 1528 */ 1529 for (i = 0; i < imo->imo_num_memberships; ++i) { 1530 if (imo->imo_membership[i]->inm_ifp == ifp && 1531 in_hosteq(imo->imo_membership[i]->inm_addr, ia)) 1532 break; 1533 } 1534 if (i < imo->imo_num_memberships) 1535 return EADDRINUSE; 1536 1537 if (i == IP_MAX_MEMBERSHIPS) 1538 return ETOOMANYREFS; 1539 1540 /* 1541 * Everything looks good; add a new record to the multicast 1542 * address list for the given interface. 1543 */ 1544 if ((imo->imo_membership[i] = in_addmulti(&ia, ifp)) == NULL) 1545 return ENOBUFS; 1546 1547 ++imo->imo_num_memberships; 1548 return 0; 1549 } 1550 1551 /* 1552 * Drop a multicast group membership. 1553 * Group must be a valid IP multicast address. 1554 */ 1555 static int 1556 ip_drop_membership(struct ip_moptions *imo, const struct sockopt *sopt) 1557 { 1558 struct in_addr ia; 1559 struct ifnet *ifp; 1560 int i, error; 1561 1562 if (sopt->sopt_size == sizeof(struct ip_mreq)) 1563 error = ip_get_membership(sopt, &ifp, &ia, false); 1564 else 1565 #ifdef INET6 1566 error = ip6_get_membership(sopt, &ifp, &ia, sizeof(ia)); 1567 #else 1568 return EINVAL; 1569 #endif 1570 1571 if (error) 1572 return error; 1573 1574 /* 1575 * Find the membership in the membership array. 1576 */ 1577 for (i = 0; i < imo->imo_num_memberships; ++i) { 1578 if ((ifp == NULL || 1579 imo->imo_membership[i]->inm_ifp == ifp) && 1580 in_hosteq(imo->imo_membership[i]->inm_addr, ia)) 1581 break; 1582 } 1583 if (i == imo->imo_num_memberships) 1584 return EADDRNOTAVAIL; 1585 1586 /* 1587 * Give up the multicast address record to which the 1588 * membership points. 1589 */ 1590 in_delmulti(imo->imo_membership[i]); 1591 1592 /* 1593 * Remove the gap in the membership array. 1594 */ 1595 for (++i; i < imo->imo_num_memberships; ++i) 1596 imo->imo_membership[i-1] = imo->imo_membership[i]; 1597 --imo->imo_num_memberships; 1598 return 0; 1599 } 1600 1601 /* 1602 * Set the IP multicast options in response to user setsockopt(). 1603 */ 1604 int 1605 ip_setmoptions(struct ip_moptions **pimo, const struct sockopt *sopt) 1606 { 1607 struct ip_moptions *imo = *pimo; 1608 struct in_addr addr; 1609 struct ifnet *ifp; 1610 int ifindex, error = 0; 1611 1612 if (!imo) { 1613 /* 1614 * No multicast option buffer attached to the pcb; 1615 * allocate one and initialize to default values. 1616 */ 1617 imo = kmem_intr_alloc(sizeof(*imo), KM_NOSLEEP); 1618 if (imo == NULL) 1619 return ENOBUFS; 1620 1621 imo->imo_multicast_ifp = NULL; 1622 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1623 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1624 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1625 imo->imo_num_memberships = 0; 1626 *pimo = imo; 1627 } 1628 1629 switch (sopt->sopt_name) { 1630 case IP_MULTICAST_IF: 1631 /* 1632 * Select the interface for outgoing multicast packets. 1633 */ 1634 error = sockopt_get(sopt, &addr, sizeof(addr)); 1635 if (error) 1636 break; 1637 1638 /* 1639 * INADDR_ANY is used to remove a previous selection. 1640 * When no interface is selected, a default one is 1641 * chosen every time a multicast packet is sent. 1642 */ 1643 if (in_nullhost(addr)) { 1644 imo->imo_multicast_ifp = NULL; 1645 break; 1646 } 1647 /* 1648 * The selected interface is identified by its local 1649 * IP address. Find the interface and confirm that 1650 * it supports multicasting. 1651 */ 1652 ifp = ip_multicast_if(&addr, &ifindex); 1653 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1654 error = EADDRNOTAVAIL; 1655 break; 1656 } 1657 imo->imo_multicast_ifp = ifp; 1658 if (ifindex) 1659 imo->imo_multicast_addr = addr; 1660 else 1661 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1662 break; 1663 1664 case IP_MULTICAST_TTL: 1665 /* 1666 * Set the IP time-to-live for outgoing multicast packets. 1667 */ 1668 error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL); 1669 break; 1670 1671 case IP_MULTICAST_LOOP: 1672 /* 1673 * Set the loopback flag for outgoing multicast packets. 1674 * Must be zero or one. 1675 */ 1676 error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1); 1677 break; 1678 1679 case IP_ADD_MEMBERSHIP: /* IPV6_JOIN_GROUP */ 1680 error = ip_add_membership(imo, sopt); 1681 break; 1682 1683 case IP_DROP_MEMBERSHIP: /* IPV6_LEAVE_GROUP */ 1684 error = ip_drop_membership(imo, sopt); 1685 break; 1686 1687 default: 1688 error = EOPNOTSUPP; 1689 break; 1690 } 1691 1692 /* 1693 * If all options have default values, no need to keep the mbuf. 1694 */ 1695 if (imo->imo_multicast_ifp == NULL && 1696 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1697 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1698 imo->imo_num_memberships == 0) { 1699 kmem_free(imo, sizeof(*imo)); 1700 *pimo = NULL; 1701 } 1702 1703 return error; 1704 } 1705 1706 /* 1707 * Return the IP multicast options in response to user getsockopt(). 1708 */ 1709 int 1710 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt) 1711 { 1712 struct in_addr addr; 1713 struct in_ifaddr *ia; 1714 uint8_t optval; 1715 int error = 0; 1716 1717 switch (sopt->sopt_name) { 1718 case IP_MULTICAST_IF: 1719 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1720 addr = zeroin_addr; 1721 else if (imo->imo_multicast_addr.s_addr) { 1722 /* return the value user has set */ 1723 addr = imo->imo_multicast_addr; 1724 } else { 1725 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1726 addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1727 } 1728 error = sockopt_set(sopt, &addr, sizeof(addr)); 1729 break; 1730 1731 case IP_MULTICAST_TTL: 1732 optval = imo ? imo->imo_multicast_ttl 1733 : IP_DEFAULT_MULTICAST_TTL; 1734 1735 error = sockopt_set(sopt, &optval, sizeof(optval)); 1736 break; 1737 1738 case IP_MULTICAST_LOOP: 1739 optval = imo ? imo->imo_multicast_loop 1740 : IP_DEFAULT_MULTICAST_LOOP; 1741 1742 error = sockopt_set(sopt, &optval, sizeof(optval)); 1743 break; 1744 1745 default: 1746 error = EOPNOTSUPP; 1747 } 1748 1749 return error; 1750 } 1751 1752 /* 1753 * Discard the IP multicast options. 1754 */ 1755 void 1756 ip_freemoptions(struct ip_moptions *imo) 1757 { 1758 int i; 1759 1760 if (imo != NULL) { 1761 for (i = 0; i < imo->imo_num_memberships; ++i) 1762 in_delmulti(imo->imo_membership[i]); 1763 kmem_free(imo, sizeof(*imo)); 1764 } 1765 } 1766 1767 /* 1768 * Routine called from ip_output() to loop back a copy of an IP multicast 1769 * packet to the input queue of a specified interface. Note that this 1770 * calls the output routine of the loopback "driver", but with an interface 1771 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1772 */ 1773 static void 1774 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 1775 { 1776 struct ip *ip; 1777 struct mbuf *copym; 1778 1779 copym = m_copypacket(m, M_DONTWAIT); 1780 if (copym != NULL && 1781 (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1782 copym = m_pullup(copym, sizeof(struct ip)); 1783 if (copym == NULL) 1784 return; 1785 /* 1786 * We don't bother to fragment if the IP length is greater 1787 * than the interface's MTU. Can this possibly matter? 1788 */ 1789 ip = mtod(copym, struct ip *); 1790 1791 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1792 in_delayed_cksum(copym); 1793 copym->m_pkthdr.csum_flags &= 1794 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1795 } 1796 1797 ip->ip_sum = 0; 1798 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1799 #ifndef NET_MPSAFE 1800 KERNEL_LOCK(1, NULL); 1801 #endif 1802 (void)looutput(ifp, copym, sintocsa(dst), NULL); 1803 #ifndef NET_MPSAFE 1804 KERNEL_UNLOCK_ONE(NULL); 1805 #endif 1806 } 1807
Indexes created Thu Sep 25 16:09:42 GMT 2025