ip_output.c revision 1.320
1 /* $NetBSD: ip_output.c,v 1.320 2020/09/08 14:12:57 christos Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 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 1.320 2020/09/08 14:12:57 christos 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 "arp.h" 105 106 #include <sys/param.h> 107 #include <sys/kmem.h> 108 #include <sys/mbuf.h> 109 #include <sys/socket.h> 110 #include <sys/socketvar.h> 111 #include <sys/kauth.h> 112 #include <sys/systm.h> 113 #include <sys/syslog.h> 114 115 #include <net/if.h> 116 #include <net/if_types.h> 117 #include <net/route.h> 118 #include <net/pfil.h> 119 120 #include <netinet/in.h> 121 #include <netinet/in_systm.h> 122 #include <netinet/ip.h> 123 #include <netinet/in_pcb.h> 124 #include <netinet/in_var.h> 125 #include <netinet/ip_var.h> 126 #include <netinet/ip_private.h> 127 #include <netinet/in_offload.h> 128 #include <netinet/portalgo.h> 129 #include <netinet/udp.h> 130 #include <netinet/udp_var.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 static int ip_ifaddrvalid(const struct in_ifaddr *); 156 157 extern pfil_head_t *inet_pfil_hook; /* XXX */ 158 159 int ip_do_loopback_cksum = 0; 160 161 static int 162 ip_mark_mpls(struct ifnet * const ifp, struct mbuf * const m, 163 const struct rtentry *rt) 164 { 165 int error = 0; 166 #ifdef MPLS 167 union mpls_shim msh; 168 169 if (rt == NULL || rt_gettag(rt) == NULL || 170 rt_gettag(rt)->sa_family != AF_MPLS || 171 (m->m_flags & (M_MCAST | M_BCAST)) != 0 || 172 ifp->if_type != IFT_ETHER) 173 return 0; 174 175 msh.s_addr = MPLS_GETSADDR(rt); 176 if (msh.shim.label != MPLS_LABEL_IMPLNULL) { 177 struct m_tag *mtag; 178 /* 179 * XXX tentative solution to tell ether_output 180 * it's MPLS. Need some more efficient solution. 181 */ 182 mtag = m_tag_get(PACKET_TAG_MPLS, 183 sizeof(int) /* dummy */, 184 M_NOWAIT); 185 if (mtag == NULL) 186 return ENOMEM; 187 m_tag_prepend(m, mtag); 188 } 189 #endif 190 return error; 191 } 192 193 /* 194 * Send an IP packet to a host. 195 */ 196 int 197 ip_if_output(struct ifnet * const ifp, struct mbuf * const m, 198 const struct sockaddr * const dst, const struct rtentry *rt) 199 { 200 int error = 0; 201 202 if (rt != NULL) { 203 error = rt_check_reject_route(rt, ifp); 204 if (error != 0) { 205 IP_STATINC(IP_STAT_RTREJECT); 206 m_freem(m); 207 return error; 208 } 209 } 210 211 error = ip_mark_mpls(ifp, m, rt); 212 if (error != 0) { 213 m_freem(m); 214 return error; 215 } 216 217 error = if_output_lock(ifp, ifp, m, dst, rt); 218 219 return error; 220 } 221 222 /* 223 * IP output. The packet in mbuf chain m contains a skeletal IP 224 * header (with len, off, ttl, proto, tos, src, dst). 225 * The mbuf chain containing the packet will be freed. 226 * The mbuf opt, if present, will not be freed. 227 */ 228 int 229 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, int flags, 230 struct ip_moptions *imo, struct inpcb *inp) 231 { 232 struct rtentry *rt; 233 struct ip *ip; 234 struct ifnet *ifp, *mifp = NULL; 235 struct mbuf *m = m0; 236 int len, hlen, error = 0; 237 struct route iproute; 238 const struct sockaddr_in *dst; 239 struct in_ifaddr *ia = NULL; 240 struct ifaddr *ifa; 241 int isbroadcast; 242 int sw_csum; 243 u_long mtu; 244 bool natt_frag = false; 245 bool rtmtu_nolock; 246 union { 247 struct sockaddr sa; 248 struct sockaddr_in sin; 249 } udst, usrc; 250 struct sockaddr *rdst = &udst.sa; /* real IP destination, as 251 * opposed to the nexthop 252 */ 253 struct psref psref, psref_ia; 254 int bound; 255 bool bind_need_restore = false; 256 const struct sockaddr *sa; 257 258 len = 0; 259 260 MCLAIM(m, &ip_tx_mowner); 261 262 KASSERT((m->m_flags & M_PKTHDR) != 0); 263 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 0); 264 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) != 265 (M_CSUM_TCPv4|M_CSUM_UDPv4)); 266 KASSERT(m->m_len >= sizeof(struct ip)); 267 268 hlen = sizeof(struct ip); 269 if (opt) { 270 m = ip_insertoptions(m, opt, &len); 271 hlen = len; 272 } 273 ip = mtod(m, struct ip *); 274 275 /* 276 * Fill in IP header. 277 */ 278 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 279 ip->ip_v = IPVERSION; 280 ip->ip_off = htons(0); 281 /* ip->ip_id filled in after we find out source ia */ 282 ip->ip_hl = hlen >> 2; 283 IP_STATINC(IP_STAT_LOCALOUT); 284 } else { 285 hlen = ip->ip_hl << 2; 286 } 287 288 /* 289 * Route packet. 290 */ 291 if (ro == NULL) { 292 memset(&iproute, 0, sizeof(iproute)); 293 ro = &iproute; 294 } 295 sockaddr_in_init(&udst.sin, &ip->ip_dst, 0); 296 dst = satocsin(rtcache_getdst(ro)); 297 298 /* 299 * If there is a cached route, check that it is to the same 300 * destination and is still up. If not, free it and try again. 301 * The address family should also be checked in case of sharing 302 * the cache with IPv6. 303 */ 304 if (dst && (dst->sin_family != AF_INET || 305 !in_hosteq(dst->sin_addr, ip->ip_dst))) 306 rtcache_free(ro); 307 308 /* XXX must be before rtcache operations */ 309 bound = curlwp_bind(); 310 bind_need_restore = true; 311 312 if ((rt = rtcache_validate(ro)) == NULL && 313 (rt = rtcache_update(ro, 1)) == NULL) { 314 dst = &udst.sin; 315 error = rtcache_setdst(ro, &udst.sa); 316 if (error != 0) { 317 IP_STATINC(IP_STAT_ODROPPED); 318 goto bad; 319 } 320 } 321 322 /* 323 * If routing to interface only, short circuit routing lookup. 324 */ 325 if (flags & IP_ROUTETOIF) { 326 ifa = ifa_ifwithladdr_psref(sintocsa(dst), &psref_ia); 327 if (ifa == NULL) { 328 IP_STATINC(IP_STAT_NOROUTE); 329 error = ENETUNREACH; 330 goto bad; 331 } 332 /* ia is already referenced by psref_ia */ 333 ia = ifatoia(ifa); 334 335 ifp = ia->ia_ifp; 336 mtu = ifp->if_mtu; 337 ip->ip_ttl = 1; 338 isbroadcast = in_broadcast(dst->sin_addr, ifp); 339 } else if (((IN_MULTICAST(ip->ip_dst.s_addr) || 340 ip->ip_dst.s_addr == INADDR_BROADCAST) || 341 (flags & IP_ROUTETOIFINDEX)) && 342 imo != NULL && imo->imo_multicast_if_index != 0) { 343 ifp = mifp = if_get_byindex(imo->imo_multicast_if_index, &psref); 344 if (ifp == NULL) { 345 IP_STATINC(IP_STAT_NOROUTE); 346 error = ENETUNREACH; 347 goto bad; 348 } 349 mtu = ifp->if_mtu; 350 ia = in_get_ia_from_ifp_psref(ifp, &psref_ia); 351 if (ia == NULL) { 352 IP_STATINC(IP_STAT_IFNOADDR); 353 error = EADDRNOTAVAIL; 354 goto bad; 355 } 356 if (IN_MULTICAST(ip->ip_dst.s_addr) || 357 ip->ip_dst.s_addr == INADDR_BROADCAST) { 358 isbroadcast = 0; 359 } else { 360 /* IP_ROUTETOIFINDEX */ 361 isbroadcast = in_broadcast(dst->sin_addr, ifp); 362 if ((isbroadcast == 0) && ((ifp->if_flags & 363 (IFF_LOOPBACK | IFF_POINTOPOINT)) == 0) && 364 (in_direct(dst->sin_addr, ifp) == 0)) { 365 /* gateway address required */ 366 if (rt == NULL) 367 rt = rtcache_init(ro); 368 if (rt == NULL || rt->rt_ifp != ifp) { 369 IP_STATINC(IP_STAT_NOROUTE); 370 error = EHOSTUNREACH; 371 goto bad; 372 } 373 rt->rt_use++; 374 if (rt->rt_flags & RTF_GATEWAY) 375 dst = satosin(rt->rt_gateway); 376 if (rt->rt_flags & RTF_HOST) 377 isbroadcast = 378 rt->rt_flags & RTF_BROADCAST; 379 } 380 } 381 } else { 382 if (rt == NULL) 383 rt = rtcache_init(ro); 384 if (rt == NULL) { 385 IP_STATINC(IP_STAT_NOROUTE); 386 error = EHOSTUNREACH; 387 goto bad; 388 } 389 if (ifa_is_destroying(rt->rt_ifa)) { 390 rtcache_unref(rt, ro); 391 rt = NULL; 392 IP_STATINC(IP_STAT_NOROUTE); 393 error = EHOSTUNREACH; 394 goto bad; 395 } 396 ifa_acquire(rt->rt_ifa, &psref_ia); 397 ia = ifatoia(rt->rt_ifa); 398 ifp = rt->rt_ifp; 399 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 400 mtu = ifp->if_mtu; 401 rt->rt_use++; 402 if (rt->rt_flags & RTF_GATEWAY) 403 dst = satosin(rt->rt_gateway); 404 if (rt->rt_flags & RTF_HOST) 405 isbroadcast = rt->rt_flags & RTF_BROADCAST; 406 else 407 isbroadcast = in_broadcast(dst->sin_addr, ifp); 408 } 409 rtmtu_nolock = rt && (rt->rt_rmx.rmx_locks & RTV_MTU) == 0; 410 411 if (IN_MULTICAST(ip->ip_dst.s_addr) || 412 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 413 bool inmgroup; 414 415 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 416 M_BCAST : M_MCAST; 417 /* 418 * See if the caller provided any multicast options 419 */ 420 if (imo != NULL) 421 ip->ip_ttl = imo->imo_multicast_ttl; 422 else 423 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 424 425 /* 426 * if we don't know the outgoing ifp yet, we can't generate 427 * output 428 */ 429 if (!ifp) { 430 IP_STATINC(IP_STAT_NOROUTE); 431 error = ENETUNREACH; 432 goto bad; 433 } 434 435 /* 436 * If the packet is multicast or broadcast, confirm that 437 * the outgoing interface can transmit it. 438 */ 439 if (((m->m_flags & M_MCAST) && 440 (ifp->if_flags & IFF_MULTICAST) == 0) || 441 ((m->m_flags & M_BCAST) && 442 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 443 IP_STATINC(IP_STAT_NOROUTE); 444 error = ENETUNREACH; 445 goto bad; 446 } 447 /* 448 * If source address not specified yet, use an address 449 * of outgoing interface. 450 */ 451 if (in_nullhost(ip->ip_src)) { 452 struct in_ifaddr *xia; 453 struct ifaddr *xifa; 454 struct psref _psref; 455 456 xia = in_get_ia_from_ifp_psref(ifp, &_psref); 457 if (!xia) { 458 IP_STATINC(IP_STAT_IFNOADDR); 459 error = EADDRNOTAVAIL; 460 goto bad; 461 } 462 xifa = &xia->ia_ifa; 463 if (xifa->ifa_getifa != NULL) { 464 ia4_release(xia, &_psref); 465 /* FIXME ifa_getifa is NOMPSAFE */ 466 xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 467 if (xia == NULL) { 468 IP_STATINC(IP_STAT_IFNOADDR); 469 error = EADDRNOTAVAIL; 470 goto bad; 471 } 472 ia4_acquire(xia, &_psref); 473 } 474 ip->ip_src = xia->ia_addr.sin_addr; 475 ia4_release(xia, &_psref); 476 } 477 478 inmgroup = in_multi_group(ip->ip_dst, ifp, flags); 479 if (inmgroup && (imo == NULL || imo->imo_multicast_loop)) { 480 /* 481 * If we belong to the destination multicast group 482 * on the outgoing interface, and the caller did not 483 * forbid loopback, loop back a copy. 484 */ 485 ip_mloopback(ifp, m, &udst.sin); 486 } 487 #ifdef MROUTING 488 else { 489 /* 490 * If we are acting as a multicast router, perform 491 * multicast forwarding as if the packet had just 492 * arrived on the interface to which we are about 493 * to send. The multicast forwarding function 494 * recursively calls this function, using the 495 * IP_FORWARDING flag to prevent infinite recursion. 496 * 497 * Multicasts that are looped back by ip_mloopback(), 498 * above, will be forwarded by the ip_input() routine, 499 * if necessary. 500 */ 501 extern struct socket *ip_mrouter; 502 503 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 504 if (ip_mforward(m, ifp) != 0) { 505 m_freem(m); 506 goto done; 507 } 508 } 509 } 510 #endif 511 /* 512 * Multicasts with a time-to-live of zero may be looped- 513 * back, above, but must not be transmitted on a network. 514 * Also, multicasts addressed to the loopback interface 515 * are not sent -- the above call to ip_mloopback() will 516 * loop back a copy if this host actually belongs to the 517 * destination group on the loopback interface. 518 */ 519 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 520 IP_STATINC(IP_STAT_ODROPPED); 521 m_freem(m); 522 goto done; 523 } 524 goto sendit; 525 } 526 527 /* 528 * If source address not specified yet, use address 529 * of outgoing interface. 530 */ 531 if (in_nullhost(ip->ip_src)) { 532 struct ifaddr *xifa; 533 534 xifa = &ia->ia_ifa; 535 if (xifa->ifa_getifa != NULL) { 536 ia4_release(ia, &psref_ia); 537 /* FIXME ifa_getifa is NOMPSAFE */ 538 ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 539 if (ia == NULL) { 540 error = EADDRNOTAVAIL; 541 goto bad; 542 } 543 ia4_acquire(ia, &psref_ia); 544 } 545 ip->ip_src = ia->ia_addr.sin_addr; 546 } 547 548 /* 549 * Packets with Class-D address as source are not valid per 550 * RFC1112. 551 */ 552 if (IN_MULTICAST(ip->ip_src.s_addr)) { 553 IP_STATINC(IP_STAT_ODROPPED); 554 error = EADDRNOTAVAIL; 555 goto bad; 556 } 557 558 /* 559 * Look for broadcast address and verify user is allowed to 560 * send such a packet. 561 */ 562 if (isbroadcast) { 563 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 564 IP_STATINC(IP_STAT_BCASTDENIED); 565 error = EADDRNOTAVAIL; 566 goto bad; 567 } 568 if ((flags & IP_ALLOWBROADCAST) == 0) { 569 IP_STATINC(IP_STAT_BCASTDENIED); 570 error = EACCES; 571 goto bad; 572 } 573 /* don't allow broadcast messages to be fragmented */ 574 if (ntohs(ip->ip_len) > ifp->if_mtu) { 575 IP_STATINC(IP_STAT_BCASTDENIED); 576 error = EMSGSIZE; 577 goto bad; 578 } 579 m->m_flags |= M_BCAST; 580 } else 581 m->m_flags &= ~M_BCAST; 582 583 sendit: 584 if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) { 585 if (m->m_pkthdr.len < IP_MINFRAGSIZE) { 586 ip->ip_id = 0; 587 } else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 588 ip->ip_id = ip_newid(ia); 589 } else { 590 /* 591 * TSO capable interfaces (typically?) increment 592 * ip_id for each segment. 593 * "allocate" enough ids here to increase the chance 594 * for them to be unique. 595 * 596 * note that the following calculation is not 597 * needed to be precise. wasting some ip_id is fine. 598 */ 599 600 unsigned int segsz = m->m_pkthdr.segsz; 601 unsigned int datasz = ntohs(ip->ip_len) - hlen; 602 unsigned int num = howmany(datasz, segsz); 603 604 ip->ip_id = ip_newid_range(ia, num); 605 } 606 } 607 if (ia != NULL) { 608 ia4_release(ia, &psref_ia); 609 ia = NULL; 610 } 611 612 /* 613 * If we're doing Path MTU Discovery, we need to set DF unless 614 * the route's MTU is locked. 615 */ 616 if ((flags & IP_MTUDISC) != 0 && rtmtu_nolock) { 617 ip->ip_off |= htons(IP_DF); 618 } 619 620 #ifdef IPSEC 621 if (ipsec_used) { 622 bool ipsec_done = false; 623 bool count_drop = false; 624 625 /* Perform IPsec processing, if any. */ 626 error = ipsec4_output(m, inp, flags, &mtu, &natt_frag, 627 &ipsec_done, &count_drop); 628 if (count_drop) 629 IP_STATINC(IP_STAT_IPSECDROP_OUT); 630 if (error || ipsec_done) 631 goto done; 632 } 633 634 if (!ipsec_used || !natt_frag) 635 #endif 636 { 637 /* 638 * Run through list of hooks for output packets. 639 */ 640 error = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_OUT); 641 if (error || m == NULL) { 642 IP_STATINC(IP_STAT_PFILDROP_OUT); 643 goto done; 644 } 645 } 646 647 ip = mtod(m, struct ip *); 648 hlen = ip->ip_hl << 2; 649 650 m->m_pkthdr.csum_data |= hlen << 16; 651 652 /* 653 * search for the source address structure to 654 * maintain output statistics, and verify address 655 * validity 656 */ 657 KASSERT(ia == NULL); 658 sockaddr_in_init(&usrc.sin, &ip->ip_src, 0); 659 ifa = ifaof_ifpforaddr_psref(&usrc.sa, ifp, &psref_ia); 660 if (ifa != NULL) 661 ia = ifatoia(ifa); 662 663 /* 664 * Ensure we only send from a valid address. 665 * A NULL address is valid because the packet could be 666 * generated from a packet filter. 667 */ 668 if (ia != NULL && (flags & IP_FORWARDING) == 0 && 669 (error = ip_ifaddrvalid(ia)) != 0) 670 { 671 ARPLOG(LOG_ERR, 672 "refusing to send from invalid address %s (pid %d)\n", 673 ARPLOGADDR(&ip->ip_src), curproc->p_pid); 674 IP_STATINC(IP_STAT_ODROPPED); 675 if (error == 1) 676 /* 677 * Address exists, but is tentative or detached. 678 * We can't send from it because it's invalid, 679 * so we drop the packet. 680 */ 681 error = 0; 682 else 683 error = EADDRNOTAVAIL; 684 goto bad; 685 } 686 687 /* Maybe skip checksums on loopback interfaces. */ 688 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 689 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 690 } 691 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 692 693 /* Need to fragment the packet */ 694 if (ntohs(ip->ip_len) > mtu && 695 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 696 goto fragment; 697 } 698 699 #if IFA_STATS 700 if (ia) 701 ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len); 702 #endif 703 /* 704 * Always initialize the sum to 0! Some HW assisted 705 * checksumming requires this. 706 */ 707 ip->ip_sum = 0; 708 709 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 710 /* 711 * Perform any checksums that the hardware can't do 712 * for us. 713 * 714 * XXX Does any hardware require the {th,uh}_sum 715 * XXX fields to be 0? 716 */ 717 if (sw_csum & M_CSUM_IPv4) { 718 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 719 ip->ip_sum = in_cksum(m, hlen); 720 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 721 } 722 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 723 if (IN_NEED_CHECKSUM(ifp, 724 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 725 in_undefer_cksum_tcpudp(m); 726 } 727 m->m_pkthdr.csum_flags &= 728 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 729 } 730 } 731 732 sa = (m->m_flags & M_MCAST) ? sintocsa(rdst) : sintocsa(dst); 733 734 /* Send it */ 735 if (__predict_false(sw_csum & M_CSUM_TSOv4)) { 736 /* 737 * TSO4 is required by a packet, but disabled for 738 * the interface. 739 */ 740 error = ip_tso_output(ifp, m, sa, rt); 741 } else 742 error = ip_if_output(ifp, m, sa, rt); 743 goto done; 744 745 fragment: 746 /* 747 * We can't use HW checksumming if we're about to fragment the packet. 748 * 749 * XXX Some hardware can do this. 750 */ 751 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 752 if (IN_NEED_CHECKSUM(ifp, 753 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 754 in_undefer_cksum_tcpudp(m); 755 } 756 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 757 } 758 759 /* 760 * Too large for interface; fragment if possible. 761 * Must be able to put at least 8 bytes per fragment. 762 */ 763 if (ntohs(ip->ip_off) & IP_DF) { 764 if (flags & IP_RETURNMTU) { 765 KASSERT(inp != NULL); 766 inp->inp_errormtu = mtu; 767 } 768 error = EMSGSIZE; 769 IP_STATINC(IP_STAT_CANTFRAG); 770 goto bad; 771 } 772 773 error = ip_fragment(m, ifp, mtu); 774 if (error) { 775 m = NULL; 776 goto bad; 777 } 778 779 for (; m; m = m0) { 780 m0 = m->m_nextpkt; 781 m->m_nextpkt = NULL; 782 if (error) { 783 m_freem(m); 784 continue; 785 } 786 #if IFA_STATS 787 if (ia) 788 ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len); 789 #endif 790 /* 791 * If we get there, the packet has not been handled by 792 * IPsec whereas it should have. Now that it has been 793 * fragmented, re-inject it in ip_output so that IPsec 794 * processing can occur. 795 */ 796 if (natt_frag) { 797 error = ip_output(m, opt, NULL, 798 flags | IP_RAWOUTPUT | IP_NOIPNEWID, 799 imo, inp); 800 } else { 801 KASSERT((m->m_pkthdr.csum_flags & 802 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 803 error = ip_if_output(ifp, m, (m->m_flags & M_MCAST) ? 804 sintocsa(rdst) : sintocsa(dst), rt); 805 } 806 } 807 if (error == 0) { 808 IP_STATINC(IP_STAT_FRAGMENTED); 809 } 810 811 done: 812 ia4_release(ia, &psref_ia); 813 rtcache_unref(rt, ro); 814 if (ro == &iproute) { 815 rtcache_free(&iproute); 816 } 817 if (mifp != NULL) { 818 if_put(mifp, &psref); 819 } 820 if (bind_need_restore) 821 curlwp_bindx(bound); 822 return error; 823 824 bad: 825 m_freem(m); 826 goto done; 827 } 828 829 int 830 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 831 { 832 struct ip *ip, *mhip; 833 struct mbuf *m0; 834 int len, hlen, off; 835 int mhlen, firstlen; 836 struct mbuf **mnext; 837 int sw_csum = m->m_pkthdr.csum_flags; 838 int fragments = 0; 839 int error = 0; 840 int ipoff, ipflg; 841 842 ip = mtod(m, struct ip *); 843 hlen = ip->ip_hl << 2; 844 845 /* Preserve the offset and flags. */ 846 ipoff = ntohs(ip->ip_off) & IP_OFFMASK; 847 ipflg = ntohs(ip->ip_off) & (IP_RF|IP_DF|IP_MF); 848 849 if (ifp != NULL) 850 sw_csum &= ~ifp->if_csum_flags_tx; 851 852 len = (mtu - hlen) &~ 7; 853 if (len < 8) { 854 IP_STATINC(IP_STAT_CANTFRAG); 855 m_freem(m); 856 return EMSGSIZE; 857 } 858 859 firstlen = len; 860 mnext = &m->m_nextpkt; 861 862 /* 863 * Loop through length of segment after first fragment, 864 * make new header and copy data of each part and link onto chain. 865 */ 866 m0 = m; 867 mhlen = sizeof(struct ip); 868 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 869 MGETHDR(m, M_DONTWAIT, MT_HEADER); 870 if (m == NULL) { 871 error = ENOBUFS; 872 IP_STATINC(IP_STAT_ODROPPED); 873 goto sendorfree; 874 } 875 MCLAIM(m, m0->m_owner); 876 877 *mnext = m; 878 mnext = &m->m_nextpkt; 879 880 m->m_data += max_linkhdr; 881 mhip = mtod(m, struct ip *); 882 *mhip = *ip; 883 884 /* we must inherit the flags */ 885 m->m_flags |= m0->m_flags & M_COPYFLAGS; 886 887 if (hlen > sizeof(struct ip)) { 888 mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip); 889 mhip->ip_hl = mhlen >> 2; 890 } 891 m->m_len = mhlen; 892 893 mhip->ip_off = ((off - hlen) >> 3) + ipoff; 894 mhip->ip_off |= ipflg; 895 if (off + len >= ntohs(ip->ip_len)) 896 len = ntohs(ip->ip_len) - off; 897 else 898 mhip->ip_off |= IP_MF; 899 HTONS(mhip->ip_off); 900 901 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 902 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 903 if (m->m_next == NULL) { 904 error = ENOBUFS; 905 IP_STATINC(IP_STAT_ODROPPED); 906 goto sendorfree; 907 } 908 909 m->m_pkthdr.len = mhlen + len; 910 m_reset_rcvif(m); 911 912 mhip->ip_sum = 0; 913 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 914 if (sw_csum & M_CSUM_IPv4) { 915 mhip->ip_sum = in_cksum(m, mhlen); 916 } else { 917 /* 918 * checksum is hw-offloaded or not necessary. 919 */ 920 m->m_pkthdr.csum_flags |= 921 m0->m_pkthdr.csum_flags & M_CSUM_IPv4; 922 m->m_pkthdr.csum_data |= mhlen << 16; 923 KASSERT(!(ifp != NULL && 924 IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) || 925 (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 926 } 927 IP_STATINC(IP_STAT_OFRAGMENTS); 928 fragments++; 929 } 930 931 /* 932 * Update first fragment by trimming what's been copied out 933 * and updating header, then send each fragment (in order). 934 */ 935 m = m0; 936 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 937 m->m_pkthdr.len = hlen + firstlen; 938 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 939 ip->ip_off |= htons(IP_MF); 940 ip->ip_sum = 0; 941 if (sw_csum & M_CSUM_IPv4) { 942 ip->ip_sum = in_cksum(m, hlen); 943 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 944 } else { 945 /* 946 * checksum is hw-offloaded or not necessary. 947 */ 948 KASSERT(!(ifp != NULL && IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) || 949 (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 950 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 951 sizeof(struct ip)); 952 } 953 954 sendorfree: 955 /* 956 * If there is no room for all the fragments, don't queue 957 * any of them. 958 */ 959 if (ifp != NULL) { 960 IFQ_LOCK(&ifp->if_snd); 961 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 962 error == 0) { 963 error = ENOBUFS; 964 IP_STATINC(IP_STAT_ODROPPED); 965 IFQ_INC_DROPS(&ifp->if_snd); 966 } 967 IFQ_UNLOCK(&ifp->if_snd); 968 } 969 if (error) { 970 for (m = m0; m; m = m0) { 971 m0 = m->m_nextpkt; 972 m->m_nextpkt = NULL; 973 m_freem(m); 974 } 975 } 976 977 return error; 978 } 979 980 /* 981 * Determine the maximum length of the options to be inserted; 982 * we would far rather allocate too much space rather than too little. 983 */ 984 u_int 985 ip_optlen(struct inpcb *inp) 986 { 987 struct mbuf *m = inp->inp_options; 988 989 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) { 990 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 991 } 992 return 0; 993 } 994 995 /* 996 * Insert IP options into preformed packet. 997 * Adjust IP destination as required for IP source routing, 998 * as indicated by a non-zero in_addr at the start of the options. 999 */ 1000 static struct mbuf * 1001 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 1002 { 1003 struct ipoption *p = mtod(opt, struct ipoption *); 1004 struct mbuf *n; 1005 struct ip *ip = mtod(m, struct ip *); 1006 unsigned optlen; 1007 1008 optlen = opt->m_len - sizeof(p->ipopt_dst); 1009 KASSERT(optlen % 4 == 0); 1010 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 1011 return m; /* XXX should fail */ 1012 if (!in_nullhost(p->ipopt_dst)) 1013 ip->ip_dst = p->ipopt_dst; 1014 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 1015 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1016 if (n == NULL) 1017 return m; 1018 MCLAIM(n, m->m_owner); 1019 m_move_pkthdr(n, m); 1020 m->m_len -= sizeof(struct ip); 1021 m->m_data += sizeof(struct ip); 1022 n->m_next = m; 1023 n->m_len = optlen + sizeof(struct ip); 1024 n->m_data += max_linkhdr; 1025 memcpy(mtod(n, void *), ip, sizeof(struct ip)); 1026 m = n; 1027 } else { 1028 m->m_data -= optlen; 1029 m->m_len += optlen; 1030 memmove(mtod(m, void *), ip, sizeof(struct ip)); 1031 } 1032 m->m_pkthdr.len += optlen; 1033 ip = mtod(m, struct ip *); 1034 memcpy(ip + 1, p->ipopt_list, optlen); 1035 *phlen = sizeof(struct ip) + optlen; 1036 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1037 return m; 1038 } 1039 1040 /* 1041 * Copy options from ipsrc to ipdst, omitting those not copied during 1042 * fragmentation. 1043 */ 1044 int 1045 ip_optcopy(struct ip *ipsrc, struct ip *ipdst) 1046 { 1047 u_char *cp, *dp; 1048 int opt, optlen, cnt; 1049 1050 cp = (u_char *)(ipsrc + 1); 1051 dp = (u_char *)(ipdst + 1); 1052 cnt = (ipsrc->ip_hl << 2) - sizeof(struct ip); 1053 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1054 opt = cp[0]; 1055 if (opt == IPOPT_EOL) 1056 break; 1057 if (opt == IPOPT_NOP) { 1058 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1059 *dp++ = IPOPT_NOP; 1060 optlen = 1; 1061 continue; 1062 } 1063 1064 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp)); 1065 optlen = cp[IPOPT_OLEN]; 1066 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen < cnt); 1067 1068 /* Invalid lengths should have been caught by ip_dooptions. */ 1069 if (optlen > cnt) 1070 optlen = cnt; 1071 if (IPOPT_COPIED(opt)) { 1072 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 1073 dp += optlen; 1074 } 1075 } 1076 1077 for (optlen = dp - (u_char *)(ipdst+1); optlen & 0x3; optlen++) { 1078 *dp++ = IPOPT_EOL; 1079 } 1080 1081 return optlen; 1082 } 1083 1084 /* 1085 * IP socket option processing. 1086 */ 1087 int 1088 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1089 { 1090 struct inpcb *inp = sotoinpcb(so); 1091 struct ip *ip = &inp->inp_ip; 1092 int inpflags = inp->inp_flags; 1093 int optval = 0, error = 0; 1094 struct in_pktinfo pktinfo; 1095 1096 KASSERT(solocked(so)); 1097 1098 if (sopt->sopt_level != IPPROTO_IP) { 1099 if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER) 1100 return 0; 1101 return ENOPROTOOPT; 1102 } 1103 1104 switch (op) { 1105 case PRCO_SETOPT: 1106 switch (sopt->sopt_name) { 1107 case IP_OPTIONS: 1108 #ifdef notyet 1109 case IP_RETOPTS: 1110 #endif 1111 error = ip_pcbopts(inp, sopt); 1112 break; 1113 1114 case IP_TOS: 1115 case IP_TTL: 1116 case IP_MINTTL: 1117 case IP_RECVOPTS: 1118 case IP_RECVRETOPTS: 1119 case IP_RECVDSTADDR: 1120 case IP_RECVIF: 1121 case IP_RECVPKTINFO: 1122 case IP_RECVTTL: 1123 case IP_BINDANY: 1124 error = sockopt_getint(sopt, &optval); 1125 if (error) 1126 break; 1127 1128 switch (sopt->sopt_name) { 1129 case IP_TOS: 1130 ip->ip_tos = optval; 1131 break; 1132 1133 case IP_TTL: 1134 ip->ip_ttl = optval; 1135 break; 1136 1137 case IP_MINTTL: 1138 if (optval > 0 && optval <= MAXTTL) 1139 inp->inp_ip_minttl = optval; 1140 else 1141 error = EINVAL; 1142 break; 1143 #define OPTSET(bit) \ 1144 if (optval) \ 1145 inpflags |= bit; \ 1146 else \ 1147 inpflags &= ~bit; 1148 1149 case IP_RECVOPTS: 1150 OPTSET(INP_RECVOPTS); 1151 break; 1152 1153 case IP_RECVPKTINFO: 1154 OPTSET(INP_RECVPKTINFO); 1155 break; 1156 1157 case IP_RECVRETOPTS: 1158 OPTSET(INP_RECVRETOPTS); 1159 break; 1160 1161 case IP_RECVDSTADDR: 1162 OPTSET(INP_RECVDSTADDR); 1163 break; 1164 1165 case IP_RECVIF: 1166 OPTSET(INP_RECVIF); 1167 break; 1168 1169 case IP_RECVTTL: 1170 OPTSET(INP_RECVTTL); 1171 break; 1172 1173 case IP_BINDANY: 1174 error = kauth_authorize_network( 1175 kauth_cred_get(), KAUTH_NETWORK_BIND, 1176 KAUTH_REQ_NETWORK_BIND_ANYADDR, so, 1177 NULL, NULL); 1178 if (error == 0) { 1179 OPTSET(INP_BINDANY); 1180 } 1181 break; 1182 } 1183 break; 1184 case IP_PKTINFO: 1185 error = sockopt_getint(sopt, &optval); 1186 if (!error) { 1187 /* Linux compatibility */ 1188 OPTSET(INP_RECVPKTINFO); 1189 break; 1190 } 1191 error = sockopt_get(sopt, &pktinfo, sizeof(pktinfo)); 1192 if (error) 1193 break; 1194 1195 if (pktinfo.ipi_ifindex == 0) { 1196 inp->inp_prefsrcip = pktinfo.ipi_addr; 1197 break; 1198 } 1199 1200 /* Solaris compatibility */ 1201 struct ifnet *ifp; 1202 struct in_ifaddr *ia; 1203 int s; 1204 1205 /* pick up primary address */ 1206 s = pserialize_read_enter(); 1207 ifp = if_byindex(pktinfo.ipi_ifindex); 1208 if (ifp == NULL) { 1209 pserialize_read_exit(s); 1210 error = EADDRNOTAVAIL; 1211 break; 1212 } 1213 ia = in_get_ia_from_ifp(ifp); 1214 if (ia == NULL) { 1215 pserialize_read_exit(s); 1216 error = EADDRNOTAVAIL; 1217 break; 1218 } 1219 inp->inp_prefsrcip = IA_SIN(ia)->sin_addr; 1220 pserialize_read_exit(s); 1221 break; 1222 break; 1223 #undef OPTSET 1224 1225 case IP_MULTICAST_IF: 1226 case IP_MULTICAST_TTL: 1227 case IP_MULTICAST_LOOP: 1228 case IP_ADD_MEMBERSHIP: 1229 case IP_DROP_MEMBERSHIP: 1230 error = ip_setmoptions(&inp->inp_moptions, sopt); 1231 break; 1232 1233 case IP_PORTRANGE: 1234 error = sockopt_getint(sopt, &optval); 1235 if (error) 1236 break; 1237 1238 switch (optval) { 1239 case IP_PORTRANGE_DEFAULT: 1240 case IP_PORTRANGE_HIGH: 1241 inpflags &= ~(INP_LOWPORT); 1242 break; 1243 1244 case IP_PORTRANGE_LOW: 1245 inpflags |= INP_LOWPORT; 1246 break; 1247 1248 default: 1249 error = EINVAL; 1250 break; 1251 } 1252 break; 1253 1254 case IP_PORTALGO: 1255 error = sockopt_getint(sopt, &optval); 1256 if (error) 1257 break; 1258 1259 error = portalgo_algo_index_select( 1260 (struct inpcb_hdr *)inp, optval); 1261 break; 1262 1263 #if defined(IPSEC) 1264 case IP_IPSEC_POLICY: 1265 if (ipsec_enabled) { 1266 error = ipsec_set_policy(inp, 1267 sopt->sopt_data, sopt->sopt_size, 1268 curlwp->l_cred); 1269 } else 1270 error = ENOPROTOOPT; 1271 break; 1272 #endif /* IPSEC */ 1273 1274 default: 1275 error = ENOPROTOOPT; 1276 break; 1277 } 1278 break; 1279 1280 case PRCO_GETOPT: 1281 switch (sopt->sopt_name) { 1282 case IP_OPTIONS: 1283 case IP_RETOPTS: { 1284 struct mbuf *mopts = inp->inp_options; 1285 1286 if (mopts) { 1287 struct mbuf *m; 1288 1289 m = m_copym(mopts, 0, M_COPYALL, M_DONTWAIT); 1290 if (m == NULL) { 1291 error = ENOBUFS; 1292 break; 1293 } 1294 error = sockopt_setmbuf(sopt, m); 1295 } 1296 break; 1297 } 1298 case IP_TOS: 1299 case IP_TTL: 1300 case IP_MINTTL: 1301 case IP_RECVOPTS: 1302 case IP_RECVRETOPTS: 1303 case IP_RECVDSTADDR: 1304 case IP_RECVIF: 1305 case IP_RECVPKTINFO: 1306 case IP_RECVTTL: 1307 case IP_ERRORMTU: 1308 case IP_BINDANY: 1309 switch (sopt->sopt_name) { 1310 case IP_TOS: 1311 optval = ip->ip_tos; 1312 break; 1313 1314 case IP_TTL: 1315 optval = ip->ip_ttl; 1316 break; 1317 1318 case IP_MINTTL: 1319 optval = inp->inp_ip_minttl; 1320 break; 1321 1322 case IP_ERRORMTU: 1323 optval = inp->inp_errormtu; 1324 break; 1325 1326 #define OPTBIT(bit) (inpflags & bit ? 1 : 0) 1327 1328 case IP_RECVOPTS: 1329 optval = OPTBIT(INP_RECVOPTS); 1330 break; 1331 1332 case IP_RECVPKTINFO: 1333 optval = OPTBIT(INP_RECVPKTINFO); 1334 break; 1335 1336 case IP_RECVRETOPTS: 1337 optval = OPTBIT(INP_RECVRETOPTS); 1338 break; 1339 1340 case IP_RECVDSTADDR: 1341 optval = OPTBIT(INP_RECVDSTADDR); 1342 break; 1343 1344 case IP_RECVIF: 1345 optval = OPTBIT(INP_RECVIF); 1346 break; 1347 1348 case IP_RECVTTL: 1349 optval = OPTBIT(INP_RECVTTL); 1350 break; 1351 1352 case IP_BINDANY: 1353 optval = OPTBIT(INP_BINDANY); 1354 break; 1355 } 1356 error = sockopt_setint(sopt, optval); 1357 break; 1358 1359 case IP_PKTINFO: 1360 switch (sopt->sopt_size) { 1361 case sizeof(int): 1362 /* Linux compatibility */ 1363 optval = OPTBIT(INP_RECVPKTINFO); 1364 error = sockopt_setint(sopt, optval); 1365 break; 1366 case sizeof(struct in_pktinfo): 1367 /* Solaris compatibility */ 1368 pktinfo.ipi_ifindex = 0; 1369 pktinfo.ipi_addr = inp->inp_prefsrcip; 1370 error = sockopt_set(sopt, &pktinfo, 1371 sizeof(pktinfo)); 1372 break; 1373 default: 1374 /* 1375 * While size is stuck at 0, and, later, if 1376 * the caller doesn't use an exactly sized 1377 * recipient for the data, default to Linux 1378 * compatibility 1379 */ 1380 optval = OPTBIT(INP_RECVPKTINFO); 1381 error = sockopt_setint(sopt, optval); 1382 break; 1383 } 1384 break; 1385 1386 #if 0 /* defined(IPSEC) */ 1387 case IP_IPSEC_POLICY: 1388 { 1389 struct mbuf *m = NULL; 1390 1391 /* XXX this will return EINVAL as sopt is empty */ 1392 error = ipsec_get_policy(inp, sopt->sopt_data, 1393 sopt->sopt_size, &m); 1394 if (error == 0) 1395 error = sockopt_setmbuf(sopt, m); 1396 break; 1397 } 1398 #endif /*IPSEC*/ 1399 1400 case IP_MULTICAST_IF: 1401 case IP_MULTICAST_TTL: 1402 case IP_MULTICAST_LOOP: 1403 case IP_ADD_MEMBERSHIP: 1404 case IP_DROP_MEMBERSHIP: 1405 error = ip_getmoptions(inp->inp_moptions, sopt); 1406 break; 1407 1408 case IP_PORTRANGE: 1409 if (inpflags & INP_LOWPORT) 1410 optval = IP_PORTRANGE_LOW; 1411 else 1412 optval = IP_PORTRANGE_DEFAULT; 1413 error = sockopt_setint(sopt, optval); 1414 break; 1415 1416 case IP_PORTALGO: 1417 optval = inp->inp_portalgo; 1418 error = sockopt_setint(sopt, optval); 1419 break; 1420 1421 default: 1422 error = ENOPROTOOPT; 1423 break; 1424 } 1425 break; 1426 } 1427 1428 if (!error) { 1429 inp->inp_flags = inpflags; 1430 } 1431 return error; 1432 } 1433 1434 static int 1435 ip_pktinfo_prepare(const struct inpcb *inp, const struct in_pktinfo *pktinfo, 1436 struct ip_pktopts *pktopts, int *flags, kauth_cred_t cred) 1437 { 1438 struct ip_moptions *imo; 1439 int error = 0; 1440 bool addrset = false; 1441 1442 if (!in_nullhost(pktinfo->ipi_addr)) { 1443 pktopts->ippo_laddr.sin_addr = pktinfo->ipi_addr; 1444 /* EADDRNOTAVAIL? */ 1445 error = in_pcbbindableaddr(inp, &pktopts->ippo_laddr, cred); 1446 if (error != 0) 1447 return error; 1448 addrset = true; 1449 } 1450 1451 if (pktinfo->ipi_ifindex != 0) { 1452 if (!addrset) { 1453 struct ifnet *ifp; 1454 struct in_ifaddr *ia; 1455 int s; 1456 1457 /* pick up primary address */ 1458 s = pserialize_read_enter(); 1459 ifp = if_byindex(pktinfo->ipi_ifindex); 1460 if (ifp == NULL) { 1461 pserialize_read_exit(s); 1462 return EADDRNOTAVAIL; 1463 } 1464 ia = in_get_ia_from_ifp(ifp); 1465 if (ia == NULL) { 1466 pserialize_read_exit(s); 1467 return EADDRNOTAVAIL; 1468 } 1469 pktopts->ippo_laddr.sin_addr = IA_SIN(ia)->sin_addr; 1470 pserialize_read_exit(s); 1471 } 1472 1473 /* 1474 * If specified ipi_ifindex, 1475 * use copied or locally initialized ip_moptions. 1476 * Original ip_moptions must not be modified. 1477 */ 1478 imo = &pktopts->ippo_imobuf; /* local buf in pktopts */ 1479 if (pktopts->ippo_imo != NULL) { 1480 memcpy(imo, pktopts->ippo_imo, sizeof(*imo)); 1481 } else { 1482 memset(imo, 0, sizeof(*imo)); 1483 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1484 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1485 } 1486 imo->imo_multicast_if_index = pktinfo->ipi_ifindex; 1487 pktopts->ippo_imo = imo; 1488 *flags |= IP_ROUTETOIFINDEX; 1489 } 1490 return error; 1491 } 1492 1493 /* 1494 * Set up IP outgoing packet options. Even if control is NULL, 1495 * pktopts->ippo_laddr and pktopts->ippo_imo are set and used. 1496 */ 1497 int 1498 ip_setpktopts(struct mbuf *control, struct ip_pktopts *pktopts, int *flags, 1499 struct inpcb *inp, kauth_cred_t cred) 1500 { 1501 struct cmsghdr *cm; 1502 struct in_pktinfo pktinfo; 1503 int error; 1504 1505 pktopts->ippo_imo = inp->inp_moptions; 1506 1507 struct in_addr *ia = in_nullhost(inp->inp_prefsrcip) ? &inp->inp_laddr : 1508 &inp->inp_prefsrcip; 1509 sockaddr_in_init(&pktopts->ippo_laddr, ia, 0); 1510 1511 if (control == NULL) 1512 return 0; 1513 1514 /* 1515 * XXX: Currently, we assume all the optional information is 1516 * stored in a single mbuf. 1517 */ 1518 if (control->m_next) 1519 return EINVAL; 1520 1521 for (; control->m_len > 0; 1522 control->m_data += CMSG_ALIGN(cm->cmsg_len), 1523 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 1524 cm = mtod(control, struct cmsghdr *); 1525 if ((control->m_len < sizeof(*cm)) || 1526 (cm->cmsg_len == 0) || 1527 (cm->cmsg_len > control->m_len)) { 1528 return EINVAL; 1529 } 1530 if (cm->cmsg_level != IPPROTO_IP) 1531 continue; 1532 1533 switch (cm->cmsg_type) { 1534 case IP_PKTINFO: 1535 if (cm->cmsg_len != CMSG_LEN(sizeof(pktinfo))) 1536 return EINVAL; 1537 memcpy(&pktinfo, CMSG_DATA(cm), sizeof(pktinfo)); 1538 error = ip_pktinfo_prepare(inp, &pktinfo, pktopts, 1539 flags, cred); 1540 if (error) 1541 return error; 1542 break; 1543 case IP_SENDSRCADDR: /* FreeBSD compatibility */ 1544 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_addr))) 1545 return EINVAL; 1546 pktinfo.ipi_ifindex = 0; 1547 pktinfo.ipi_addr = 1548 ((struct in_pktinfo *)CMSG_DATA(cm))->ipi_addr; 1549 error = ip_pktinfo_prepare(inp, &pktinfo, pktopts, 1550 flags, cred); 1551 if (error) 1552 return error; 1553 break; 1554 default: 1555 return ENOPROTOOPT; 1556 } 1557 } 1558 return 0; 1559 } 1560 1561 /* 1562 * Set up IP options in pcb for insertion in output packets. 1563 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1564 * with destination address if source routed. 1565 */ 1566 static int 1567 ip_pcbopts(struct inpcb *inp, const struct sockopt *sopt) 1568 { 1569 struct mbuf *m; 1570 const u_char *cp; 1571 u_char *dp; 1572 int cnt; 1573 1574 KASSERT(inp_locked(inp)); 1575 1576 /* Turn off any old options. */ 1577 if (inp->inp_options) { 1578 m_free(inp->inp_options); 1579 } 1580 inp->inp_options = NULL; 1581 if ((cnt = sopt->sopt_size) == 0) { 1582 /* Only turning off any previous options. */ 1583 return 0; 1584 } 1585 cp = sopt->sopt_data; 1586 1587 if (cnt % 4) { 1588 /* Must be 4-byte aligned, because there's no padding. */ 1589 return EINVAL; 1590 } 1591 1592 m = m_get(M_DONTWAIT, MT_SOOPTS); 1593 if (m == NULL) 1594 return ENOBUFS; 1595 1596 dp = mtod(m, u_char *); 1597 memset(dp, 0, sizeof(struct in_addr)); 1598 dp += sizeof(struct in_addr); 1599 m->m_len = sizeof(struct in_addr); 1600 1601 /* 1602 * IP option list according to RFC791. Each option is of the form 1603 * 1604 * [optval] [olen] [(olen - 2) data bytes] 1605 * 1606 * We validate the list and copy options to an mbuf for prepending 1607 * to data packets. The IP first-hop destination address will be 1608 * stored before actual options and is zero if unset. 1609 */ 1610 while (cnt > 0) { 1611 uint8_t optval, olen, offset; 1612 1613 optval = cp[IPOPT_OPTVAL]; 1614 1615 if (optval == IPOPT_EOL || optval == IPOPT_NOP) { 1616 olen = 1; 1617 } else { 1618 if (cnt < IPOPT_OLEN + 1) 1619 goto bad; 1620 1621 olen = cp[IPOPT_OLEN]; 1622 if (olen < IPOPT_OLEN + 1 || olen > cnt) 1623 goto bad; 1624 } 1625 1626 if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) { 1627 /* 1628 * user process specifies route as: 1629 * ->A->B->C->D 1630 * D must be our final destination (but we can't 1631 * check that since we may not have connected yet). 1632 * A is first hop destination, which doesn't appear in 1633 * actual IP option, but is stored before the options. 1634 */ 1635 if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr)) 1636 goto bad; 1637 1638 offset = cp[IPOPT_OFFSET]; 1639 memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1, 1640 sizeof(struct in_addr)); 1641 1642 cp += sizeof(struct in_addr); 1643 cnt -= sizeof(struct in_addr); 1644 olen -= sizeof(struct in_addr); 1645 1646 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1647 goto bad; 1648 1649 memcpy(dp, cp, olen); 1650 dp[IPOPT_OPTVAL] = optval; 1651 dp[IPOPT_OLEN] = olen; 1652 dp[IPOPT_OFFSET] = offset; 1653 break; 1654 } else { 1655 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1656 goto bad; 1657 1658 memcpy(dp, cp, olen); 1659 break; 1660 } 1661 1662 dp += olen; 1663 m->m_len += olen; 1664 1665 if (optval == IPOPT_EOL) 1666 break; 1667 1668 cp += olen; 1669 cnt -= olen; 1670 } 1671 1672 inp->inp_options = m; 1673 return 0; 1674 1675 bad: 1676 (void)m_free(m); 1677 return EINVAL; 1678 } 1679 1680 /* 1681 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1682 * Must be called in a pserialize critical section. 1683 */ 1684 static struct ifnet * 1685 ip_multicast_if(struct in_addr *a, int *ifindexp) 1686 { 1687 int ifindex; 1688 struct ifnet *ifp = NULL; 1689 struct in_ifaddr *ia; 1690 1691 if (ifindexp) 1692 *ifindexp = 0; 1693 if (ntohl(a->s_addr) >> 24 == 0) { 1694 ifindex = ntohl(a->s_addr) & 0xffffff; 1695 ifp = if_byindex(ifindex); 1696 if (!ifp) 1697 return NULL; 1698 if (ifindexp) 1699 *ifindexp = ifindex; 1700 } else { 1701 IN_ADDRHASH_READER_FOREACH(ia, a->s_addr) { 1702 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1703 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1704 ifp = ia->ia_ifp; 1705 if (if_is_deactivated(ifp)) 1706 ifp = NULL; 1707 break; 1708 } 1709 } 1710 } 1711 return ifp; 1712 } 1713 1714 static int 1715 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval) 1716 { 1717 u_int tval; 1718 u_char cval; 1719 int error; 1720 1721 if (sopt == NULL) 1722 return EINVAL; 1723 1724 switch (sopt->sopt_size) { 1725 case sizeof(u_char): 1726 error = sockopt_get(sopt, &cval, sizeof(u_char)); 1727 tval = cval; 1728 break; 1729 1730 case sizeof(u_int): 1731 error = sockopt_get(sopt, &tval, sizeof(u_int)); 1732 break; 1733 1734 default: 1735 error = EINVAL; 1736 } 1737 1738 if (error) 1739 return error; 1740 1741 if (tval > maxval) 1742 return EINVAL; 1743 1744 *val = tval; 1745 return 0; 1746 } 1747 1748 static int 1749 ip_get_membership(const struct sockopt *sopt, struct ifnet **ifp, 1750 struct psref *psref, struct in_addr *ia, bool add) 1751 { 1752 int error; 1753 struct ip_mreq mreq; 1754 1755 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 1756 if (error) 1757 return error; 1758 1759 if (!IN_MULTICAST(mreq.imr_multiaddr.s_addr)) 1760 return EINVAL; 1761 1762 memcpy(ia, &mreq.imr_multiaddr, sizeof(*ia)); 1763 1764 if (in_nullhost(mreq.imr_interface)) { 1765 union { 1766 struct sockaddr dst; 1767 struct sockaddr_in dst4; 1768 } u; 1769 struct route ro; 1770 1771 if (!add) { 1772 *ifp = NULL; 1773 return 0; 1774 } 1775 /* 1776 * If no interface address was provided, use the interface of 1777 * the route to the given multicast address. 1778 */ 1779 struct rtentry *rt; 1780 memset(&ro, 0, sizeof(ro)); 1781 1782 sockaddr_in_init(&u.dst4, ia, 0); 1783 error = rtcache_setdst(&ro, &u.dst); 1784 if (error != 0) 1785 return error; 1786 *ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL; 1787 if (*ifp != NULL) { 1788 if (if_is_deactivated(*ifp)) 1789 *ifp = NULL; 1790 else 1791 if_acquire(*ifp, psref); 1792 } 1793 rtcache_unref(rt, &ro); 1794 rtcache_free(&ro); 1795 } else { 1796 int s = pserialize_read_enter(); 1797 *ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1798 if (!add && *ifp == NULL) { 1799 pserialize_read_exit(s); 1800 return EADDRNOTAVAIL; 1801 } 1802 if (*ifp != NULL) { 1803 if (if_is_deactivated(*ifp)) 1804 *ifp = NULL; 1805 else 1806 if_acquire(*ifp, psref); 1807 } 1808 pserialize_read_exit(s); 1809 } 1810 return 0; 1811 } 1812 1813 /* 1814 * Add a multicast group membership. 1815 * Group must be a valid IP multicast address. 1816 */ 1817 static int 1818 ip_add_membership(struct ip_moptions *imo, const struct sockopt *sopt) 1819 { 1820 struct ifnet *ifp = NULL; // XXX: gcc [ppc] 1821 struct in_addr ia; 1822 int i, error, bound; 1823 struct psref psref; 1824 1825 /* imo is protected by solock or referenced only by the caller */ 1826 1827 bound = curlwp_bind(); 1828 if (sopt->sopt_size == sizeof(struct ip_mreq)) 1829 error = ip_get_membership(sopt, &ifp, &psref, &ia, true); 1830 else { 1831 #ifdef INET6 1832 error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia)); 1833 #else 1834 error = EINVAL; 1835 #endif 1836 } 1837 1838 if (error) 1839 goto out; 1840 1841 /* 1842 * See if we found an interface, and confirm that it 1843 * supports multicast. 1844 */ 1845 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1846 error = EADDRNOTAVAIL; 1847 goto out; 1848 } 1849 1850 /* 1851 * See if the membership already exists or if all the 1852 * membership slots are full. 1853 */ 1854 for (i = 0; i < imo->imo_num_memberships; ++i) { 1855 if (imo->imo_membership[i]->inm_ifp == ifp && 1856 in_hosteq(imo->imo_membership[i]->inm_addr, ia)) 1857 break; 1858 } 1859 if (i < imo->imo_num_memberships) { 1860 error = EADDRINUSE; 1861 goto out; 1862 } 1863 1864 if (i == IP_MAX_MEMBERSHIPS) { 1865 error = ETOOMANYREFS; 1866 goto out; 1867 } 1868 1869 /* 1870 * Everything looks good; add a new record to the multicast 1871 * address list for the given interface. 1872 */ 1873 imo->imo_membership[i] = in_addmulti(&ia, ifp); 1874 if (imo->imo_membership[i] == NULL) { 1875 error = ENOBUFS; 1876 goto out; 1877 } 1878 1879 ++imo->imo_num_memberships; 1880 error = 0; 1881 out: 1882 if_put(ifp, &psref); 1883 curlwp_bindx(bound); 1884 return error; 1885 } 1886 1887 /* 1888 * Drop a multicast group membership. 1889 * Group must be a valid IP multicast address. 1890 */ 1891 static int 1892 ip_drop_membership(struct ip_moptions *imo, const struct sockopt *sopt) 1893 { 1894 struct in_addr ia = { .s_addr = 0 }; // XXX: gcc [ppc] 1895 struct ifnet *ifp = NULL; // XXX: gcc [ppc] 1896 int i, error, bound; 1897 struct psref psref; 1898 1899 /* imo is protected by solock or referenced only by the caller */ 1900 1901 bound = curlwp_bind(); 1902 if (sopt->sopt_size == sizeof(struct ip_mreq)) 1903 error = ip_get_membership(sopt, &ifp, &psref, &ia, false); 1904 else { 1905 #ifdef INET6 1906 error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia)); 1907 #else 1908 error = EINVAL; 1909 #endif 1910 } 1911 1912 if (error) 1913 goto out; 1914 1915 /* 1916 * Find the membership in the membership array. 1917 */ 1918 for (i = 0; i < imo->imo_num_memberships; ++i) { 1919 if ((ifp == NULL || 1920 imo->imo_membership[i]->inm_ifp == ifp) && 1921 in_hosteq(imo->imo_membership[i]->inm_addr, ia)) 1922 break; 1923 } 1924 if (i == imo->imo_num_memberships) { 1925 error = EADDRNOTAVAIL; 1926 goto out; 1927 } 1928 1929 /* 1930 * Give up the multicast address record to which the 1931 * membership points. 1932 */ 1933 in_delmulti(imo->imo_membership[i]); 1934 1935 /* 1936 * Remove the gap in the membership array. 1937 */ 1938 for (++i; i < imo->imo_num_memberships; ++i) 1939 imo->imo_membership[i-1] = imo->imo_membership[i]; 1940 --imo->imo_num_memberships; 1941 error = 0; 1942 out: 1943 if_put(ifp, &psref); 1944 curlwp_bindx(bound); 1945 return error; 1946 } 1947 1948 /* 1949 * Set the IP multicast options in response to user setsockopt(). 1950 */ 1951 int 1952 ip_setmoptions(struct ip_moptions **pimo, const struct sockopt *sopt) 1953 { 1954 struct ip_moptions *imo = *pimo; 1955 struct in_addr addr; 1956 struct ifnet *ifp; 1957 int ifindex, error = 0; 1958 1959 /* The passed imo isn't NULL, it should be protected by solock */ 1960 1961 if (!imo) { 1962 /* 1963 * No multicast option buffer attached to the pcb; 1964 * allocate one and initialize to default values. 1965 */ 1966 imo = kmem_intr_alloc(sizeof(*imo), KM_NOSLEEP); 1967 if (imo == NULL) 1968 return ENOBUFS; 1969 1970 imo->imo_multicast_if_index = 0; 1971 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1972 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1973 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1974 imo->imo_num_memberships = 0; 1975 *pimo = imo; 1976 } 1977 1978 switch (sopt->sopt_name) { 1979 case IP_MULTICAST_IF: { 1980 int s; 1981 /* 1982 * Select the interface for outgoing multicast packets. 1983 */ 1984 error = sockopt_get(sopt, &addr, sizeof(addr)); 1985 if (error) 1986 break; 1987 1988 /* 1989 * INADDR_ANY is used to remove a previous selection. 1990 * When no interface is selected, a default one is 1991 * chosen every time a multicast packet is sent. 1992 */ 1993 if (in_nullhost(addr)) { 1994 imo->imo_multicast_if_index = 0; 1995 break; 1996 } 1997 /* 1998 * The selected interface is identified by its local 1999 * IP address. Find the interface and confirm that 2000 * it supports multicasting. 2001 */ 2002 s = pserialize_read_enter(); 2003 ifp = ip_multicast_if(&addr, &ifindex); 2004 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2005 pserialize_read_exit(s); 2006 error = EADDRNOTAVAIL; 2007 break; 2008 } 2009 imo->imo_multicast_if_index = ifp->if_index; 2010 pserialize_read_exit(s); 2011 if (ifindex) 2012 imo->imo_multicast_addr = addr; 2013 else 2014 imo->imo_multicast_addr.s_addr = INADDR_ANY; 2015 break; 2016 } 2017 2018 case IP_MULTICAST_TTL: 2019 /* 2020 * Set the IP time-to-live for outgoing multicast packets. 2021 */ 2022 error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL); 2023 break; 2024 2025 case IP_MULTICAST_LOOP: 2026 /* 2027 * Set the loopback flag for outgoing multicast packets. 2028 * Must be zero or one. 2029 */ 2030 error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1); 2031 break; 2032 2033 case IP_ADD_MEMBERSHIP: /* IPV6_JOIN_GROUP */ 2034 error = ip_add_membership(imo, sopt); 2035 break; 2036 2037 case IP_DROP_MEMBERSHIP: /* IPV6_LEAVE_GROUP */ 2038 error = ip_drop_membership(imo, sopt); 2039 break; 2040 2041 default: 2042 error = EOPNOTSUPP; 2043 break; 2044 } 2045 2046 /* 2047 * If all options have default values, no need to keep the mbuf. 2048 */ 2049 if (imo->imo_multicast_if_index == 0 && 2050 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 2051 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 2052 imo->imo_num_memberships == 0) { 2053 kmem_intr_free(imo, sizeof(*imo)); 2054 *pimo = NULL; 2055 } 2056 2057 return error; 2058 } 2059 2060 /* 2061 * Return the IP multicast options in response to user getsockopt(). 2062 */ 2063 int 2064 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt) 2065 { 2066 struct in_addr addr; 2067 uint8_t optval; 2068 int error = 0; 2069 2070 /* imo is protected by solock or referenced only by the caller */ 2071 2072 switch (sopt->sopt_name) { 2073 case IP_MULTICAST_IF: 2074 if (imo == NULL || imo->imo_multicast_if_index == 0) 2075 addr = zeroin_addr; 2076 else if (imo->imo_multicast_addr.s_addr) { 2077 /* return the value user has set */ 2078 addr = imo->imo_multicast_addr; 2079 } else { 2080 struct ifnet *ifp; 2081 struct in_ifaddr *ia = NULL; 2082 int s = pserialize_read_enter(); 2083 2084 ifp = if_byindex(imo->imo_multicast_if_index); 2085 if (ifp != NULL) { 2086 ia = in_get_ia_from_ifp(ifp); 2087 } 2088 addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 2089 pserialize_read_exit(s); 2090 } 2091 error = sockopt_set(sopt, &addr, sizeof(addr)); 2092 break; 2093 2094 case IP_MULTICAST_TTL: 2095 optval = imo ? imo->imo_multicast_ttl 2096 : IP_DEFAULT_MULTICAST_TTL; 2097 2098 error = sockopt_set(sopt, &optval, sizeof(optval)); 2099 break; 2100 2101 case IP_MULTICAST_LOOP: 2102 optval = imo ? imo->imo_multicast_loop 2103 : IP_DEFAULT_MULTICAST_LOOP; 2104 2105 error = sockopt_set(sopt, &optval, sizeof(optval)); 2106 break; 2107 2108 default: 2109 error = EOPNOTSUPP; 2110 } 2111 2112 return error; 2113 } 2114 2115 /* 2116 * Discard the IP multicast options. 2117 */ 2118 void 2119 ip_freemoptions(struct ip_moptions *imo) 2120 { 2121 int i; 2122 2123 /* The owner of imo (inp) should be protected by solock */ 2124 2125 if (imo != NULL) { 2126 for (i = 0; i < imo->imo_num_memberships; ++i) { 2127 struct in_multi *inm = imo->imo_membership[i]; 2128 in_delmulti(inm); 2129 /* ifp should not leave thanks to solock */ 2130 } 2131 2132 kmem_intr_free(imo, sizeof(*imo)); 2133 } 2134 } 2135 2136 /* 2137 * Routine called from ip_output() to loop back a copy of an IP multicast 2138 * packet to the input queue of a specified interface. Note that this 2139 * calls the output routine of the loopback "driver", but with an interface 2140 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 2141 */ 2142 static void 2143 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 2144 { 2145 struct ip *ip; 2146 struct mbuf *copym; 2147 2148 copym = m_copypacket(m, M_DONTWAIT); 2149 if (copym != NULL && 2150 (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 2151 copym = m_pullup(copym, sizeof(struct ip)); 2152 if (copym == NULL) 2153 return; 2154 /* 2155 * We don't bother to fragment if the IP length is greater 2156 * than the interface's MTU. Can this possibly matter? 2157 */ 2158 ip = mtod(copym, struct ip *); 2159 2160 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 2161 in_undefer_cksum_tcpudp(copym); 2162 copym->m_pkthdr.csum_flags &= 2163 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 2164 } 2165 2166 ip->ip_sum = 0; 2167 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 2168 KERNEL_LOCK_UNLESS_NET_MPSAFE(); 2169 (void)looutput(ifp, copym, sintocsa(dst), NULL); 2170 KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 2171 } 2172 2173 /* 2174 * Ensure sending address is valid. 2175 * Returns 0 on success, -1 if an error should be sent back or 1 2176 * if the packet could be dropped without error (protocol dependent). 2177 */ 2178 static int 2179 ip_ifaddrvalid(const struct in_ifaddr *ia) 2180 { 2181 2182 if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY) 2183 return 0; 2184 2185 if (ia->ia4_flags & IN_IFF_DUPLICATED) 2186 return -1; 2187 else if (ia->ia4_flags & (IN_IFF_TENTATIVE | IN_IFF_DETACHED)) 2188 return 1; 2189 2190 return 0; 2191 } 2192
Indexes created Mon Sep 22 21:09:42 GMT 2025