ip_output.c revision 1.279
1 /* $NetBSD: ip_output.c,v 1.279 2017/05/12 17:53:54 ryo 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 1.279 2017/05/12 17:53:54 ryo 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> 130 131 #ifdef INET6 132 #include <netinet6/ip6_var.h> 133 #endif 134 135 #ifdef MROUTING 136 #include <netinet/ip_mroute.h> 137 #endif 138 139 #ifdef IPSEC 140 #include <netipsec/ipsec.h> 141 #include <netipsec/key.h> 142 #endif 143 144 #ifdef MPLS 145 #include <netmpls/mpls.h> 146 #include <netmpls/mpls_var.h> 147 #endif 148 149 static int ip_pcbopts(struct inpcb *, const struct sockopt *); 150 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 151 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 152 static void ip_mloopback(struct ifnet *, struct mbuf *, 153 const struct sockaddr_in *); 154 static int ip_ifaddrvalid(const struct in_ifaddr *); 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, 162 const struct rtentry *rt) 163 { 164 int error = 0; 165 #ifdef MPLS 166 union mpls_shim msh; 167 168 if (rt == NULL || rt_gettag(rt) == NULL || 169 rt_gettag(rt)->sa_family != AF_MPLS || 170 (m->m_flags & (M_MCAST | M_BCAST)) != 0 || 171 ifp->if_type != IFT_ETHER) 172 return 0; 173 174 msh.s_addr = MPLS_GETSADDR(rt); 175 if (msh.shim.label != MPLS_LABEL_IMPLNULL) { 176 struct m_tag *mtag; 177 /* 178 * XXX tentative solution to tell ether_output 179 * it's MPLS. Need some more efficient solution. 180 */ 181 mtag = m_tag_get(PACKET_TAG_MPLS, 182 sizeof(int) /* dummy */, 183 M_NOWAIT); 184 if (mtag == NULL) 185 return ENOMEM; 186 m_tag_prepend(m, mtag); 187 } 188 #endif 189 return error; 190 } 191 192 /* 193 * Send an IP packet to a host. 194 */ 195 int 196 ip_if_output(struct ifnet * const ifp, struct mbuf * const m, 197 const struct sockaddr * const dst, const struct rtentry *rt) 198 { 199 int error = 0; 200 201 if (rt != NULL) { 202 error = rt_check_reject_route(rt, ifp); 203 if (error != 0) { 204 m_freem(m); 205 return error; 206 } 207 } 208 209 error = ip_mark_mpls(ifp, m, rt); 210 if (error != 0) { 211 m_freem(m); 212 return error; 213 } 214 215 error = if_output_lock(ifp, ifp, m, dst, rt); 216 217 return error; 218 } 219 220 /* 221 * IP output. The packet in mbuf chain m contains a skeletal IP 222 * header (with len, off, ttl, proto, tos, src, dst). 223 * The mbuf chain containing the packet will be freed. 224 * The mbuf opt, if present, will not be freed. 225 */ 226 int 227 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, int flags, 228 struct ip_moptions *imo, struct inpcb *inp) 229 { 230 struct rtentry *rt; 231 struct ip *ip; 232 struct ifnet *ifp, *mifp = NULL; 233 struct mbuf *m = m0; 234 int hlen = sizeof (struct ip); 235 int len, error = 0; 236 struct route iproute; 237 const struct sockaddr_in *dst; 238 struct in_ifaddr *ia = NULL; 239 int isbroadcast; 240 int sw_csum; 241 u_long mtu; 242 bool natt_frag = false; 243 bool rtmtu_nolock; 244 union { 245 struct sockaddr dst; 246 struct sockaddr_in dst4; 247 } u; 248 struct sockaddr *rdst = &u.dst; /* real IP destination, as opposed 249 * to the nexthop 250 */ 251 struct psref psref, psref_ia; 252 int bound; 253 bool bind_need_restore = false; 254 255 len = 0; 256 257 MCLAIM(m, &ip_tx_mowner); 258 259 KASSERT((m->m_flags & M_PKTHDR) != 0); 260 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 0); 261 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) != 262 (M_CSUM_TCPv4|M_CSUM_UDPv4)); 263 264 if (opt) { 265 m = ip_insertoptions(m, opt, &len); 266 if (len >= sizeof(struct ip)) 267 hlen = len; 268 } 269 ip = mtod(m, struct ip *); 270 271 /* 272 * Fill in IP header. 273 */ 274 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 275 ip->ip_v = IPVERSION; 276 ip->ip_off = htons(0); 277 /* ip->ip_id filled in after we find out source ia */ 278 ip->ip_hl = hlen >> 2; 279 IP_STATINC(IP_STAT_LOCALOUT); 280 } else { 281 hlen = ip->ip_hl << 2; 282 } 283 284 /* 285 * Route packet. 286 */ 287 if (ro == NULL) { 288 memset(&iproute, 0, sizeof(iproute)); 289 ro = &iproute; 290 } 291 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 292 dst = satocsin(rtcache_getdst(ro)); 293 294 /* 295 * If there is a cached route, check that it is to the same 296 * destination and is still up. If not, free it and try again. 297 * The address family should also be checked in case of sharing 298 * the cache with IPv6. 299 */ 300 if (dst && (dst->sin_family != AF_INET || 301 !in_hosteq(dst->sin_addr, ip->ip_dst))) 302 rtcache_free(ro); 303 304 if ((rt = rtcache_validate(ro)) == NULL && 305 (rt = rtcache_update(ro, 1)) == NULL) { 306 dst = &u.dst4; 307 error = rtcache_setdst(ro, &u.dst); 308 if (error != 0) 309 goto bad; 310 } 311 312 bound = curlwp_bind(); 313 bind_need_restore = true; 314 /* 315 * If routing to interface only, short circuit routing lookup. 316 */ 317 if (flags & IP_ROUTETOIF) { 318 struct ifaddr *ifa; 319 320 ifa = ifa_ifwithladdr_psref(sintocsa(dst), &psref_ia); 321 if (ifa == NULL) { 322 IP_STATINC(IP_STAT_NOROUTE); 323 error = ENETUNREACH; 324 goto bad; 325 } 326 /* ia is already referenced by psref_ia */ 327 ia = ifatoia(ifa); 328 329 ifp = ia->ia_ifp; 330 mtu = ifp->if_mtu; 331 ip->ip_ttl = 1; 332 isbroadcast = in_broadcast(dst->sin_addr, ifp); 333 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 334 ip->ip_dst.s_addr == INADDR_BROADCAST) && 335 imo != NULL && imo->imo_multicast_if_index != 0) { 336 ifp = mifp = if_get_byindex(imo->imo_multicast_if_index, &psref); 337 if (ifp == NULL) { 338 IP_STATINC(IP_STAT_NOROUTE); 339 error = ENETUNREACH; 340 goto bad; 341 } 342 mtu = ifp->if_mtu; 343 ia = in_get_ia_from_ifp_psref(ifp, &psref_ia); 344 if (ia == NULL) { 345 error = EADDRNOTAVAIL; 346 goto bad; 347 } 348 isbroadcast = 0; 349 } else { 350 if (rt == NULL) 351 rt = rtcache_init(ro); 352 if (rt == NULL) { 353 IP_STATINC(IP_STAT_NOROUTE); 354 error = EHOSTUNREACH; 355 goto bad; 356 } 357 if (ifa_is_destroying(rt->rt_ifa)) { 358 rtcache_unref(rt, ro); 359 rt = NULL; 360 IP_STATINC(IP_STAT_NOROUTE); 361 error = EHOSTUNREACH; 362 goto bad; 363 } 364 ifa_acquire(rt->rt_ifa, &psref_ia); 365 ia = ifatoia(rt->rt_ifa); 366 ifp = rt->rt_ifp; 367 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 368 mtu = ifp->if_mtu; 369 rt->rt_use++; 370 if (rt->rt_flags & RTF_GATEWAY) 371 dst = satosin(rt->rt_gateway); 372 if (rt->rt_flags & RTF_HOST) 373 isbroadcast = rt->rt_flags & RTF_BROADCAST; 374 else 375 isbroadcast = in_broadcast(dst->sin_addr, ifp); 376 } 377 rtmtu_nolock = rt && (rt->rt_rmx.rmx_locks & RTV_MTU) == 0; 378 379 if (IN_MULTICAST(ip->ip_dst.s_addr) || 380 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 381 bool inmgroup; 382 383 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 384 M_BCAST : M_MCAST; 385 /* 386 * See if the caller provided any multicast options 387 */ 388 if (imo != NULL) 389 ip->ip_ttl = imo->imo_multicast_ttl; 390 else 391 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 392 393 /* 394 * if we don't know the outgoing ifp yet, we can't generate 395 * output 396 */ 397 if (!ifp) { 398 IP_STATINC(IP_STAT_NOROUTE); 399 error = ENETUNREACH; 400 goto bad; 401 } 402 403 /* 404 * If the packet is multicast or broadcast, confirm that 405 * the outgoing interface can transmit it. 406 */ 407 if (((m->m_flags & M_MCAST) && 408 (ifp->if_flags & IFF_MULTICAST) == 0) || 409 ((m->m_flags & M_BCAST) && 410 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 411 IP_STATINC(IP_STAT_NOROUTE); 412 error = ENETUNREACH; 413 goto bad; 414 } 415 /* 416 * If source address not specified yet, use an address 417 * of outgoing interface. 418 */ 419 if (in_nullhost(ip->ip_src)) { 420 struct in_ifaddr *xia; 421 struct ifaddr *xifa; 422 struct psref _psref; 423 424 xia = in_get_ia_from_ifp_psref(ifp, &_psref); 425 if (!xia) { 426 error = EADDRNOTAVAIL; 427 goto bad; 428 } 429 xifa = &xia->ia_ifa; 430 if (xifa->ifa_getifa != NULL) { 431 ia4_release(xia, &_psref); 432 /* FIXME ifa_getifa is NOMPSAFE */ 433 xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 434 if (xia == NULL) { 435 error = EADDRNOTAVAIL; 436 goto bad; 437 } 438 ia4_acquire(xia, &_psref); 439 } 440 ip->ip_src = xia->ia_addr.sin_addr; 441 ia4_release(xia, &_psref); 442 } 443 444 inmgroup = in_multi_group(ip->ip_dst, ifp, flags); 445 if (inmgroup && (imo == NULL || imo->imo_multicast_loop)) { 446 /* 447 * If we belong to the destination multicast group 448 * on the outgoing interface, and the caller did not 449 * forbid loopback, loop back a copy. 450 */ 451 ip_mloopback(ifp, m, &u.dst4); 452 } 453 #ifdef MROUTING 454 else { 455 /* 456 * If we are acting as a multicast router, perform 457 * multicast forwarding as if the packet had just 458 * arrived on the interface to which we are about 459 * to send. The multicast forwarding function 460 * recursively calls this function, using the 461 * IP_FORWARDING flag to prevent infinite recursion. 462 * 463 * Multicasts that are looped back by ip_mloopback(), 464 * above, will be forwarded by the ip_input() routine, 465 * if necessary. 466 */ 467 extern struct socket *ip_mrouter; 468 469 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 470 if (ip_mforward(m, ifp) != 0) { 471 m_freem(m); 472 goto done; 473 } 474 } 475 } 476 #endif 477 /* 478 * Multicasts with a time-to-live of zero may be looped- 479 * back, above, but must not be transmitted on a network. 480 * Also, multicasts addressed to the loopback interface 481 * are not sent -- the above call to ip_mloopback() will 482 * loop back a copy if this host actually belongs to the 483 * destination group on the loopback interface. 484 */ 485 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 486 m_freem(m); 487 goto done; 488 } 489 goto sendit; 490 } 491 492 /* 493 * If source address not specified yet, use address 494 * of outgoing interface. 495 */ 496 if (in_nullhost(ip->ip_src)) { 497 struct ifaddr *xifa; 498 499 xifa = &ia->ia_ifa; 500 if (xifa->ifa_getifa != NULL) { 501 ia4_release(ia, &psref_ia); 502 /* FIXME ifa_getifa is NOMPSAFE */ 503 ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 504 if (ia == NULL) { 505 error = EADDRNOTAVAIL; 506 goto bad; 507 } 508 ia4_acquire(ia, &psref_ia); 509 } 510 ip->ip_src = ia->ia_addr.sin_addr; 511 } 512 513 /* 514 * packets with Class-D address as source are not valid per 515 * RFC 1112 516 */ 517 if (IN_MULTICAST(ip->ip_src.s_addr)) { 518 IP_STATINC(IP_STAT_ODROPPED); 519 error = EADDRNOTAVAIL; 520 goto bad; 521 } 522 523 /* 524 * Look for broadcast address and and verify user is allowed to 525 * send such a packet. 526 */ 527 if (isbroadcast) { 528 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 529 error = EADDRNOTAVAIL; 530 goto bad; 531 } 532 if ((flags & IP_ALLOWBROADCAST) == 0) { 533 error = EACCES; 534 goto bad; 535 } 536 /* don't allow broadcast messages to be fragmented */ 537 if (ntohs(ip->ip_len) > ifp->if_mtu) { 538 error = EMSGSIZE; 539 goto bad; 540 } 541 m->m_flags |= M_BCAST; 542 } else 543 m->m_flags &= ~M_BCAST; 544 545 sendit: 546 if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) { 547 if (m->m_pkthdr.len < IP_MINFRAGSIZE) { 548 ip->ip_id = 0; 549 } else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 550 ip->ip_id = ip_newid(ia); 551 } else { 552 553 /* 554 * TSO capable interfaces (typically?) increment 555 * ip_id for each segment. 556 * "allocate" enough ids here to increase the chance 557 * for them to be unique. 558 * 559 * note that the following calculation is not 560 * needed to be precise. wasting some ip_id is fine. 561 */ 562 563 unsigned int segsz = m->m_pkthdr.segsz; 564 unsigned int datasz = ntohs(ip->ip_len) - hlen; 565 unsigned int num = howmany(datasz, segsz); 566 567 ip->ip_id = ip_newid_range(ia, num); 568 } 569 } 570 if (ia != NULL) { 571 ia4_release(ia, &psref_ia); 572 ia = NULL; 573 } 574 575 /* 576 * If we're doing Path MTU Discovery, we need to set DF unless 577 * the route's MTU is locked. 578 */ 579 if ((flags & IP_MTUDISC) != 0 && rtmtu_nolock) { 580 ip->ip_off |= htons(IP_DF); 581 } 582 583 #ifdef IPSEC 584 if (ipsec_used) { 585 bool ipsec_done = false; 586 587 /* Perform IPsec processing, if any. */ 588 error = ipsec4_output(m, inp, flags, &mtu, &natt_frag, 589 &ipsec_done); 590 if (error || ipsec_done) 591 goto done; 592 } 593 #endif 594 595 /* 596 * Run through list of hooks for output packets. 597 */ 598 error = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_OUT); 599 if (error) 600 goto done; 601 if (m == NULL) 602 goto done; 603 604 ip = mtod(m, struct ip *); 605 hlen = ip->ip_hl << 2; 606 607 m->m_pkthdr.csum_data |= hlen << 16; 608 609 /* 610 * search for the source address structure to 611 * maintain output statistics, and verify address 612 * validity 613 */ 614 KASSERT(ia == NULL); 615 ia = in_get_ia_psref(ip->ip_src, &psref_ia); 616 617 /* 618 * Ensure we only send from a valid address. 619 * A NULL address is valid because the packet could be 620 * generated from a packet filter. 621 */ 622 if (ia != NULL && (flags & IP_FORWARDING) == 0 && 623 (error = ip_ifaddrvalid(ia)) != 0) 624 { 625 ARPLOG(LOG_ERR, 626 "refusing to send from invalid address %s (pid %d)\n", 627 ARPLOGADDR(&ip->ip_src), curproc->p_pid); 628 IP_STATINC(IP_STAT_ODROPPED); 629 if (error == 1) 630 /* 631 * Address exists, but is tentative or detached. 632 * We can't send from it because it's invalid, 633 * so we drop the packet. 634 */ 635 error = 0; 636 else 637 error = EADDRNOTAVAIL; 638 goto bad; 639 } 640 641 /* Maybe skip checksums on loopback interfaces. */ 642 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 643 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 644 } 645 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 646 /* 647 * If small enough for mtu of path, or if using TCP segmentation 648 * offload, can just send directly. 649 */ 650 if (ntohs(ip->ip_len) <= mtu || 651 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 652 const struct sockaddr *sa; 653 654 #if IFA_STATS 655 if (ia) 656 ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len); 657 #endif 658 /* 659 * Always initialize the sum to 0! Some HW assisted 660 * checksumming requires this. 661 */ 662 ip->ip_sum = 0; 663 664 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 665 /* 666 * Perform any checksums that the hardware can't do 667 * for us. 668 * 669 * XXX Does any hardware require the {th,uh}_sum 670 * XXX fields to be 0? 671 */ 672 if (sw_csum & M_CSUM_IPv4) { 673 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 674 ip->ip_sum = in_cksum(m, hlen); 675 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 676 } 677 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 678 if (IN_NEED_CHECKSUM(ifp, 679 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 680 in_delayed_cksum(m); 681 } 682 m->m_pkthdr.csum_flags &= 683 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 684 } 685 } 686 687 sa = (m->m_flags & M_MCAST) ? sintocsa(rdst) : sintocsa(dst); 688 if (__predict_true( 689 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 || 690 (ifp->if_capenable & IFCAP_TSOv4) != 0)) { 691 error = ip_if_output(ifp, m, sa, rt); 692 } else { 693 error = ip_tso_output(ifp, m, sa, rt); 694 } 695 goto done; 696 } 697 698 /* 699 * We can't use HW checksumming if we're about to 700 * to fragment the packet. 701 * 702 * XXX Some hardware can do this. 703 */ 704 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 705 if (IN_NEED_CHECKSUM(ifp, 706 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 707 in_delayed_cksum(m); 708 } 709 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 710 } 711 712 /* 713 * Too large for interface; fragment if possible. 714 * Must be able to put at least 8 bytes per fragment. 715 */ 716 if (ntohs(ip->ip_off) & IP_DF) { 717 if (flags & IP_RETURNMTU) { 718 KASSERT(inp != NULL); 719 inp->inp_errormtu = mtu; 720 } 721 error = EMSGSIZE; 722 IP_STATINC(IP_STAT_CANTFRAG); 723 goto bad; 724 } 725 726 error = ip_fragment(m, ifp, mtu); 727 if (error) { 728 m = NULL; 729 goto bad; 730 } 731 732 for (; m; m = m0) { 733 m0 = m->m_nextpkt; 734 m->m_nextpkt = 0; 735 if (error) { 736 m_freem(m); 737 continue; 738 } 739 #if IFA_STATS 740 if (ia) 741 ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len); 742 #endif 743 /* 744 * If we get there, the packet has not been handled by 745 * IPsec whereas it should have. Now that it has been 746 * fragmented, re-inject it in ip_output so that IPsec 747 * processing can occur. 748 */ 749 if (natt_frag) { 750 error = ip_output(m, opt, ro, 751 flags | IP_RAWOUTPUT | IP_NOIPNEWID, 752 imo, inp); 753 } else { 754 KASSERT((m->m_pkthdr.csum_flags & 755 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 756 error = ip_if_output(ifp, m, 757 (m->m_flags & M_MCAST) ? 758 sintocsa(rdst) : sintocsa(dst), rt); 759 } 760 } 761 if (error == 0) { 762 IP_STATINC(IP_STAT_FRAGMENTED); 763 } 764 done: 765 ia4_release(ia, &psref_ia); 766 rtcache_unref(rt, ro); 767 if (ro == &iproute) { 768 rtcache_free(&iproute); 769 } 770 if (mifp != NULL) { 771 if_put(mifp, &psref); 772 } 773 if (bind_need_restore) 774 curlwp_bindx(bound); 775 return error; 776 bad: 777 m_freem(m); 778 goto done; 779 } 780 781 int 782 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 783 { 784 struct ip *ip, *mhip; 785 struct mbuf *m0; 786 int len, hlen, off; 787 int mhlen, firstlen; 788 struct mbuf **mnext; 789 int sw_csum = m->m_pkthdr.csum_flags; 790 int fragments = 0; 791 int error = 0; 792 793 ip = mtod(m, struct ip *); 794 hlen = ip->ip_hl << 2; 795 if (ifp != NULL) 796 sw_csum &= ~ifp->if_csum_flags_tx; 797 798 len = (mtu - hlen) &~ 7; 799 if (len < 8) { 800 m_freem(m); 801 return (EMSGSIZE); 802 } 803 804 firstlen = len; 805 mnext = &m->m_nextpkt; 806 807 /* 808 * Loop through length of segment after first fragment, 809 * make new header and copy data of each part and link onto chain. 810 */ 811 m0 = m; 812 mhlen = sizeof (struct ip); 813 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 814 MGETHDR(m, M_DONTWAIT, MT_HEADER); 815 if (m == 0) { 816 error = ENOBUFS; 817 IP_STATINC(IP_STAT_ODROPPED); 818 goto sendorfree; 819 } 820 MCLAIM(m, m0->m_owner); 821 *mnext = m; 822 mnext = &m->m_nextpkt; 823 m->m_data += max_linkhdr; 824 mhip = mtod(m, struct ip *); 825 *mhip = *ip; 826 /* we must inherit MCAST and BCAST flags */ 827 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 828 if (hlen > sizeof (struct ip)) { 829 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 830 mhip->ip_hl = mhlen >> 2; 831 } 832 m->m_len = mhlen; 833 mhip->ip_off = ((off - hlen) >> 3) + 834 (ntohs(ip->ip_off) & ~IP_MF); 835 if (ip->ip_off & htons(IP_MF)) 836 mhip->ip_off |= IP_MF; 837 if (off + len >= ntohs(ip->ip_len)) 838 len = ntohs(ip->ip_len) - off; 839 else 840 mhip->ip_off |= IP_MF; 841 HTONS(mhip->ip_off); 842 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 843 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 844 if (m->m_next == 0) { 845 error = ENOBUFS; /* ??? */ 846 IP_STATINC(IP_STAT_ODROPPED); 847 goto sendorfree; 848 } 849 m->m_pkthdr.len = mhlen + len; 850 m_reset_rcvif(m); 851 mhip->ip_sum = 0; 852 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 853 if (sw_csum & M_CSUM_IPv4) { 854 mhip->ip_sum = in_cksum(m, mhlen); 855 } else { 856 /* 857 * checksum is hw-offloaded or not necessary. 858 */ 859 m->m_pkthdr.csum_flags |= 860 m0->m_pkthdr.csum_flags & M_CSUM_IPv4; 861 m->m_pkthdr.csum_data |= mhlen << 16; 862 KASSERT(!(ifp != NULL && 863 IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) || 864 (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 865 } 866 IP_STATINC(IP_STAT_OFRAGMENTS); 867 fragments++; 868 } 869 /* 870 * Update first fragment by trimming what's been copied out 871 * and updating header, then send each fragment (in order). 872 */ 873 m = m0; 874 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 875 m->m_pkthdr.len = hlen + firstlen; 876 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 877 ip->ip_off |= htons(IP_MF); 878 ip->ip_sum = 0; 879 if (sw_csum & M_CSUM_IPv4) { 880 ip->ip_sum = in_cksum(m, hlen); 881 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 882 } else { 883 /* 884 * checksum is hw-offloaded or not necessary. 885 */ 886 KASSERT(!(ifp != NULL && IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) || 887 (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 888 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 889 sizeof(struct ip)); 890 } 891 sendorfree: 892 /* 893 * If there is no room for all the fragments, don't queue 894 * any of them. 895 */ 896 if (ifp != NULL) { 897 IFQ_LOCK(&ifp->if_snd); 898 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 899 error == 0) { 900 error = ENOBUFS; 901 IP_STATINC(IP_STAT_ODROPPED); 902 IFQ_INC_DROPS(&ifp->if_snd); 903 } 904 IFQ_UNLOCK(&ifp->if_snd); 905 } 906 if (error) { 907 for (m = m0; m; m = m0) { 908 m0 = m->m_nextpkt; 909 m->m_nextpkt = NULL; 910 m_freem(m); 911 } 912 } 913 return (error); 914 } 915 916 /* 917 * Process a delayed payload checksum calculation. 918 */ 919 void 920 in_delayed_cksum(struct mbuf *m) 921 { 922 struct ip *ip; 923 u_int16_t csum, offset; 924 925 ip = mtod(m, struct ip *); 926 offset = ip->ip_hl << 2; 927 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 928 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 929 csum = 0xffff; 930 931 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 932 933 if ((offset + sizeof(u_int16_t)) > m->m_len) { 934 /* This happen when ip options were inserted 935 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 936 m->m_len, offset, ip->ip_p); 937 */ 938 m_copyback(m, offset, sizeof(csum), (void *) &csum); 939 } else 940 *(u_int16_t *)(mtod(m, char *) + offset) = csum; 941 } 942 943 /* 944 * Determine the maximum length of the options to be inserted; 945 * we would far rather allocate too much space rather than too little. 946 */ 947 948 u_int 949 ip_optlen(struct inpcb *inp) 950 { 951 struct mbuf *m = inp->inp_options; 952 953 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) { 954 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 955 } 956 return 0; 957 } 958 959 /* 960 * Insert IP options into preformed packet. 961 * Adjust IP destination as required for IP source routing, 962 * as indicated by a non-zero in_addr at the start of the options. 963 */ 964 static struct mbuf * 965 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 966 { 967 struct ipoption *p = mtod(opt, struct ipoption *); 968 struct mbuf *n; 969 struct ip *ip = mtod(m, struct ip *); 970 unsigned optlen; 971 972 optlen = opt->m_len - sizeof(p->ipopt_dst); 973 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 974 return (m); /* XXX should fail */ 975 if (!in_nullhost(p->ipopt_dst)) 976 ip->ip_dst = p->ipopt_dst; 977 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 978 MGETHDR(n, M_DONTWAIT, MT_HEADER); 979 if (n == 0) 980 return (m); 981 MCLAIM(n, m->m_owner); 982 M_MOVE_PKTHDR(n, m); 983 m->m_len -= sizeof(struct ip); 984 m->m_data += sizeof(struct ip); 985 n->m_next = m; 986 m = n; 987 m->m_len = optlen + sizeof(struct ip); 988 m->m_data += max_linkhdr; 989 bcopy((void *)ip, mtod(m, void *), sizeof(struct ip)); 990 } else { 991 m->m_data -= optlen; 992 m->m_len += optlen; 993 memmove(mtod(m, void *), ip, sizeof(struct ip)); 994 } 995 m->m_pkthdr.len += optlen; 996 ip = mtod(m, struct ip *); 997 bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen); 998 *phlen = sizeof(struct ip) + optlen; 999 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1000 return (m); 1001 } 1002 1003 /* 1004 * Copy options from ip to jp, 1005 * omitting those not copied during fragmentation. 1006 */ 1007 int 1008 ip_optcopy(struct ip *ip, struct ip *jp) 1009 { 1010 u_char *cp, *dp; 1011 int opt, optlen, cnt; 1012 1013 cp = (u_char *)(ip + 1); 1014 dp = (u_char *)(jp + 1); 1015 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1016 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1017 opt = cp[0]; 1018 if (opt == IPOPT_EOL) 1019 break; 1020 if (opt == IPOPT_NOP) { 1021 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1022 *dp++ = IPOPT_NOP; 1023 optlen = 1; 1024 continue; 1025 } 1026 1027 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp)); 1028 optlen = cp[IPOPT_OLEN]; 1029 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen < cnt); 1030 1031 /* Invalid lengths should have been caught by ip_dooptions. */ 1032 if (optlen > cnt) 1033 optlen = cnt; 1034 if (IPOPT_COPIED(opt)) { 1035 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 1036 dp += optlen; 1037 } 1038 } 1039 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1040 *dp++ = IPOPT_EOL; 1041 return (optlen); 1042 } 1043 1044 /* 1045 * IP socket option processing. 1046 */ 1047 int 1048 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1049 { 1050 struct inpcb *inp = sotoinpcb(so); 1051 struct ip *ip = &inp->inp_ip; 1052 int inpflags = inp->inp_flags; 1053 int optval = 0, error = 0; 1054 1055 KASSERT(solocked(so)); 1056 1057 if (sopt->sopt_level != IPPROTO_IP) { 1058 if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER) 1059 return 0; 1060 return ENOPROTOOPT; 1061 } 1062 1063 switch (op) { 1064 case PRCO_SETOPT: 1065 switch (sopt->sopt_name) { 1066 case IP_OPTIONS: 1067 #ifdef notyet 1068 case IP_RETOPTS: 1069 #endif 1070 error = ip_pcbopts(inp, sopt); 1071 break; 1072 1073 case IP_TOS: 1074 case IP_TTL: 1075 case IP_MINTTL: 1076 case IP_PKTINFO: 1077 case IP_RECVOPTS: 1078 case IP_RECVRETOPTS: 1079 case IP_RECVDSTADDR: 1080 case IP_RECVIF: 1081 case IP_RECVPKTINFO: 1082 case IP_RECVTTL: 1083 error = sockopt_getint(sopt, &optval); 1084 if (error) 1085 break; 1086 1087 switch (sopt->sopt_name) { 1088 case IP_TOS: 1089 ip->ip_tos = optval; 1090 break; 1091 1092 case IP_TTL: 1093 ip->ip_ttl = optval; 1094 break; 1095 1096 case IP_MINTTL: 1097 if (optval > 0 && optval <= MAXTTL) 1098 inp->inp_ip_minttl = optval; 1099 else 1100 error = EINVAL; 1101 break; 1102 #define OPTSET(bit) \ 1103 if (optval) \ 1104 inpflags |= bit; \ 1105 else \ 1106 inpflags &= ~bit; 1107 1108 case IP_PKTINFO: 1109 OPTSET(INP_PKTINFO); 1110 break; 1111 1112 case IP_RECVOPTS: 1113 OPTSET(INP_RECVOPTS); 1114 break; 1115 1116 case IP_RECVPKTINFO: 1117 OPTSET(INP_RECVPKTINFO); 1118 break; 1119 1120 case IP_RECVRETOPTS: 1121 OPTSET(INP_RECVRETOPTS); 1122 break; 1123 1124 case IP_RECVDSTADDR: 1125 OPTSET(INP_RECVDSTADDR); 1126 break; 1127 1128 case IP_RECVIF: 1129 OPTSET(INP_RECVIF); 1130 break; 1131 1132 case IP_RECVTTL: 1133 OPTSET(INP_RECVTTL); 1134 break; 1135 } 1136 break; 1137 #undef OPTSET 1138 1139 case IP_MULTICAST_IF: 1140 case IP_MULTICAST_TTL: 1141 case IP_MULTICAST_LOOP: 1142 case IP_ADD_MEMBERSHIP: 1143 case IP_DROP_MEMBERSHIP: 1144 error = ip_setmoptions(&inp->inp_moptions, sopt); 1145 break; 1146 1147 case IP_PORTRANGE: 1148 error = sockopt_getint(sopt, &optval); 1149 if (error) 1150 break; 1151 1152 switch (optval) { 1153 case IP_PORTRANGE_DEFAULT: 1154 case IP_PORTRANGE_HIGH: 1155 inpflags &= ~(INP_LOWPORT); 1156 break; 1157 1158 case IP_PORTRANGE_LOW: 1159 inpflags |= INP_LOWPORT; 1160 break; 1161 1162 default: 1163 error = EINVAL; 1164 break; 1165 } 1166 break; 1167 1168 case IP_PORTALGO: 1169 error = sockopt_getint(sopt, &optval); 1170 if (error) 1171 break; 1172 1173 error = portalgo_algo_index_select( 1174 (struct inpcb_hdr *)inp, optval); 1175 break; 1176 1177 #if defined(IPSEC) 1178 case IP_IPSEC_POLICY: 1179 if (ipsec_enabled) { 1180 error = ipsec4_set_policy(inp, sopt->sopt_name, 1181 sopt->sopt_data, sopt->sopt_size, 1182 curlwp->l_cred); 1183 break; 1184 } 1185 /*FALLTHROUGH*/ 1186 #endif /* IPSEC */ 1187 1188 default: 1189 error = ENOPROTOOPT; 1190 break; 1191 } 1192 break; 1193 1194 case PRCO_GETOPT: 1195 switch (sopt->sopt_name) { 1196 case IP_OPTIONS: 1197 case IP_RETOPTS: { 1198 struct mbuf *mopts = inp->inp_options; 1199 1200 if (mopts) { 1201 struct mbuf *m; 1202 1203 m = m_copym(mopts, 0, M_COPYALL, M_DONTWAIT); 1204 if (m == NULL) { 1205 error = ENOBUFS; 1206 break; 1207 } 1208 error = sockopt_setmbuf(sopt, m); 1209 } 1210 break; 1211 } 1212 case IP_PKTINFO: 1213 case IP_TOS: 1214 case IP_TTL: 1215 case IP_MINTTL: 1216 case IP_RECVOPTS: 1217 case IP_RECVRETOPTS: 1218 case IP_RECVDSTADDR: 1219 case IP_RECVIF: 1220 case IP_RECVPKTINFO: 1221 case IP_RECVTTL: 1222 case IP_ERRORMTU: 1223 switch (sopt->sopt_name) { 1224 case IP_TOS: 1225 optval = ip->ip_tos; 1226 break; 1227 1228 case IP_TTL: 1229 optval = ip->ip_ttl; 1230 break; 1231 1232 case IP_MINTTL: 1233 optval = inp->inp_ip_minttl; 1234 break; 1235 1236 case IP_ERRORMTU: 1237 optval = inp->inp_errormtu; 1238 break; 1239 1240 #define OPTBIT(bit) (inpflags & bit ? 1 : 0) 1241 1242 case IP_PKTINFO: 1243 optval = OPTBIT(INP_PKTINFO); 1244 break; 1245 1246 case IP_RECVOPTS: 1247 optval = OPTBIT(INP_RECVOPTS); 1248 break; 1249 1250 case IP_RECVPKTINFO: 1251 optval = OPTBIT(INP_RECVPKTINFO); 1252 break; 1253 1254 case IP_RECVRETOPTS: 1255 optval = OPTBIT(INP_RECVRETOPTS); 1256 break; 1257 1258 case IP_RECVDSTADDR: 1259 optval = OPTBIT(INP_RECVDSTADDR); 1260 break; 1261 1262 case IP_RECVIF: 1263 optval = OPTBIT(INP_RECVIF); 1264 break; 1265 1266 case IP_RECVTTL: 1267 optval = OPTBIT(INP_RECVTTL); 1268 break; 1269 } 1270 error = sockopt_setint(sopt, optval); 1271 break; 1272 1273 #if 0 /* defined(IPSEC) */ 1274 case IP_IPSEC_POLICY: 1275 { 1276 struct mbuf *m = NULL; 1277 1278 /* XXX this will return EINVAL as sopt is empty */ 1279 error = ipsec4_get_policy(inp, sopt->sopt_data, 1280 sopt->sopt_size, &m); 1281 if (error == 0) 1282 error = sockopt_setmbuf(sopt, m); 1283 break; 1284 } 1285 #endif /*IPSEC*/ 1286 1287 case IP_MULTICAST_IF: 1288 case IP_MULTICAST_TTL: 1289 case IP_MULTICAST_LOOP: 1290 case IP_ADD_MEMBERSHIP: 1291 case IP_DROP_MEMBERSHIP: 1292 error = ip_getmoptions(inp->inp_moptions, sopt); 1293 break; 1294 1295 case IP_PORTRANGE: 1296 if (inpflags & INP_LOWPORT) 1297 optval = IP_PORTRANGE_LOW; 1298 else 1299 optval = IP_PORTRANGE_DEFAULT; 1300 error = sockopt_setint(sopt, optval); 1301 break; 1302 1303 case IP_PORTALGO: 1304 optval = inp->inp_portalgo; 1305 error = sockopt_setint(sopt, optval); 1306 break; 1307 1308 default: 1309 error = ENOPROTOOPT; 1310 break; 1311 } 1312 break; 1313 } 1314 1315 if (!error) { 1316 inp->inp_flags = inpflags; 1317 } 1318 return error; 1319 } 1320 1321 /* 1322 * Set up IP options in pcb for insertion in output packets. 1323 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1324 * with destination address if source routed. 1325 */ 1326 static int 1327 ip_pcbopts(struct inpcb *inp, const struct sockopt *sopt) 1328 { 1329 struct mbuf *m; 1330 const u_char *cp; 1331 u_char *dp; 1332 int cnt; 1333 1334 KASSERT(inp_locked(inp)); 1335 1336 /* Turn off any old options. */ 1337 if (inp->inp_options) { 1338 m_free(inp->inp_options); 1339 } 1340 inp->inp_options = NULL; 1341 if ((cnt = sopt->sopt_size) == 0) { 1342 /* Only turning off any previous options. */ 1343 return 0; 1344 } 1345 cp = sopt->sopt_data; 1346 1347 #ifndef __vax__ 1348 if (cnt % sizeof(int32_t)) 1349 return (EINVAL); 1350 #endif 1351 1352 m = m_get(M_DONTWAIT, MT_SOOPTS); 1353 if (m == NULL) 1354 return (ENOBUFS); 1355 1356 dp = mtod(m, u_char *); 1357 memset(dp, 0, sizeof(struct in_addr)); 1358 dp += sizeof(struct in_addr); 1359 m->m_len = sizeof(struct in_addr); 1360 1361 /* 1362 * IP option list according to RFC791. Each option is of the form 1363 * 1364 * [optval] [olen] [(olen - 2) data bytes] 1365 * 1366 * We validate the list and copy options to an mbuf for prepending 1367 * to data packets. The IP first-hop destination address will be 1368 * stored before actual options and is zero if unset. 1369 */ 1370 while (cnt > 0) { 1371 uint8_t optval, olen, offset; 1372 1373 optval = cp[IPOPT_OPTVAL]; 1374 1375 if (optval == IPOPT_EOL || optval == IPOPT_NOP) { 1376 olen = 1; 1377 } else { 1378 if (cnt < IPOPT_OLEN + 1) 1379 goto bad; 1380 1381 olen = cp[IPOPT_OLEN]; 1382 if (olen < IPOPT_OLEN + 1 || olen > cnt) 1383 goto bad; 1384 } 1385 1386 if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) { 1387 /* 1388 * user process specifies route as: 1389 * ->A->B->C->D 1390 * D must be our final destination (but we can't 1391 * check that since we may not have connected yet). 1392 * A is first hop destination, which doesn't appear in 1393 * actual IP option, but is stored before the options. 1394 */ 1395 if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr)) 1396 goto bad; 1397 1398 offset = cp[IPOPT_OFFSET]; 1399 memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1, 1400 sizeof(struct in_addr)); 1401 1402 cp += sizeof(struct in_addr); 1403 cnt -= sizeof(struct in_addr); 1404 olen -= sizeof(struct in_addr); 1405 1406 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1407 goto bad; 1408 1409 memcpy(dp, cp, olen); 1410 dp[IPOPT_OPTVAL] = optval; 1411 dp[IPOPT_OLEN] = olen; 1412 dp[IPOPT_OFFSET] = offset; 1413 break; 1414 } else { 1415 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1416 goto bad; 1417 1418 memcpy(dp, cp, olen); 1419 break; 1420 } 1421 1422 dp += olen; 1423 m->m_len += olen; 1424 1425 if (optval == IPOPT_EOL) 1426 break; 1427 1428 cp += olen; 1429 cnt -= olen; 1430 } 1431 1432 inp->inp_options = m; 1433 return 0; 1434 bad: 1435 (void)m_free(m); 1436 return EINVAL; 1437 } 1438 1439 /* 1440 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1441 * Must be called in a pserialize critical section. 1442 */ 1443 static struct ifnet * 1444 ip_multicast_if(struct in_addr *a, int *ifindexp) 1445 { 1446 int ifindex; 1447 struct ifnet *ifp = NULL; 1448 struct in_ifaddr *ia; 1449 1450 if (ifindexp) 1451 *ifindexp = 0; 1452 if (ntohl(a->s_addr) >> 24 == 0) { 1453 ifindex = ntohl(a->s_addr) & 0xffffff; 1454 ifp = if_byindex(ifindex); 1455 if (!ifp) 1456 return NULL; 1457 if (ifindexp) 1458 *ifindexp = ifindex; 1459 } else { 1460 IN_ADDRHASH_READER_FOREACH(ia, a->s_addr) { 1461 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1462 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1463 ifp = ia->ia_ifp; 1464 if (if_is_deactivated(ifp)) 1465 ifp = NULL; 1466 break; 1467 } 1468 } 1469 } 1470 return ifp; 1471 } 1472 1473 static int 1474 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval) 1475 { 1476 u_int tval; 1477 u_char cval; 1478 int error; 1479 1480 if (sopt == NULL) 1481 return EINVAL; 1482 1483 switch (sopt->sopt_size) { 1484 case sizeof(u_char): 1485 error = sockopt_get(sopt, &cval, sizeof(u_char)); 1486 tval = cval; 1487 break; 1488 1489 case sizeof(u_int): 1490 error = sockopt_get(sopt, &tval, sizeof(u_int)); 1491 break; 1492 1493 default: 1494 error = EINVAL; 1495 } 1496 1497 if (error) 1498 return error; 1499 1500 if (tval > maxval) 1501 return EINVAL; 1502 1503 *val = tval; 1504 return 0; 1505 } 1506 1507 static int 1508 ip_get_membership(const struct sockopt *sopt, struct ifnet **ifp, 1509 struct psref *psref, struct in_addr *ia, bool add) 1510 { 1511 int error; 1512 struct ip_mreq mreq; 1513 1514 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 1515 if (error) 1516 return error; 1517 1518 if (!IN_MULTICAST(mreq.imr_multiaddr.s_addr)) 1519 return EINVAL; 1520 1521 memcpy(ia, &mreq.imr_multiaddr, sizeof(*ia)); 1522 1523 if (in_nullhost(mreq.imr_interface)) { 1524 union { 1525 struct sockaddr dst; 1526 struct sockaddr_in dst4; 1527 } u; 1528 struct route ro; 1529 1530 if (!add) { 1531 *ifp = NULL; 1532 return 0; 1533 } 1534 /* 1535 * If no interface address was provided, use the interface of 1536 * the route to the given multicast address. 1537 */ 1538 struct rtentry *rt; 1539 memset(&ro, 0, sizeof(ro)); 1540 1541 sockaddr_in_init(&u.dst4, ia, 0); 1542 error = rtcache_setdst(&ro, &u.dst); 1543 if (error != 0) 1544 return error; 1545 *ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL; 1546 if (*ifp != NULL) { 1547 if (if_is_deactivated(*ifp)) 1548 *ifp = NULL; 1549 else 1550 if_acquire(*ifp, psref); 1551 } 1552 rtcache_unref(rt, &ro); 1553 rtcache_free(&ro); 1554 } else { 1555 int s = pserialize_read_enter(); 1556 *ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1557 if (!add && *ifp == NULL) { 1558 pserialize_read_exit(s); 1559 return EADDRNOTAVAIL; 1560 } 1561 if (*ifp != NULL) { 1562 if (if_is_deactivated(*ifp)) 1563 *ifp = NULL; 1564 else 1565 if_acquire(*ifp, psref); 1566 } 1567 pserialize_read_exit(s); 1568 } 1569 return 0; 1570 } 1571 1572 /* 1573 * Add a multicast group membership. 1574 * Group must be a valid IP multicast address. 1575 */ 1576 static int 1577 ip_add_membership(struct ip_moptions *imo, const struct sockopt *sopt) 1578 { 1579 struct ifnet *ifp = NULL; // XXX: gcc [ppc] 1580 struct in_addr ia; 1581 int i, error, bound; 1582 struct psref psref; 1583 1584 /* imo is protected by solock or referenced only by the caller */ 1585 1586 bound = curlwp_bind(); 1587 if (sopt->sopt_size == sizeof(struct ip_mreq)) 1588 error = ip_get_membership(sopt, &ifp, &psref, &ia, true); 1589 else 1590 #ifdef INET6 1591 error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia)); 1592 #else 1593 error = EINVAL; 1594 goto out; 1595 #endif 1596 1597 if (error) 1598 goto out; 1599 1600 /* 1601 * See if we found an interface, and confirm that it 1602 * supports multicast. 1603 */ 1604 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1605 error = EADDRNOTAVAIL; 1606 goto out; 1607 } 1608 1609 /* 1610 * See if the membership already exists or if all the 1611 * membership slots are full. 1612 */ 1613 for (i = 0; i < imo->imo_num_memberships; ++i) { 1614 if (imo->imo_membership[i]->inm_ifp == ifp && 1615 in_hosteq(imo->imo_membership[i]->inm_addr, ia)) 1616 break; 1617 } 1618 if (i < imo->imo_num_memberships) { 1619 error = EADDRINUSE; 1620 goto out; 1621 } 1622 1623 if (i == IP_MAX_MEMBERSHIPS) { 1624 error = ETOOMANYREFS; 1625 goto out; 1626 } 1627 1628 /* 1629 * Everything looks good; add a new record to the multicast 1630 * address list for the given interface. 1631 */ 1632 if ((imo->imo_membership[i] = in_addmulti(&ia, ifp)) == NULL) { 1633 error = ENOBUFS; 1634 goto out; 1635 } 1636 1637 ++imo->imo_num_memberships; 1638 error = 0; 1639 out: 1640 if_put(ifp, &psref); 1641 curlwp_bindx(bound); 1642 return error; 1643 } 1644 1645 /* 1646 * Drop a multicast group membership. 1647 * Group must be a valid IP multicast address. 1648 */ 1649 static int 1650 ip_drop_membership(struct ip_moptions *imo, const struct sockopt *sopt) 1651 { 1652 struct in_addr ia = { .s_addr = 0 }; // XXX: gcc [ppc] 1653 struct ifnet *ifp = NULL; // XXX: gcc [ppc] 1654 int i, error, bound; 1655 struct psref psref; 1656 1657 /* imo is protected by solock or referenced only by the caller */ 1658 1659 bound = curlwp_bind(); 1660 if (sopt->sopt_size == sizeof(struct ip_mreq)) 1661 error = ip_get_membership(sopt, &ifp, &psref, &ia, false); 1662 else 1663 #ifdef INET6 1664 error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia)); 1665 #else 1666 error = EINVAL; 1667 goto out; 1668 #endif 1669 1670 if (error) 1671 goto out; 1672 1673 /* 1674 * Find the membership in the membership array. 1675 */ 1676 for (i = 0; i < imo->imo_num_memberships; ++i) { 1677 if ((ifp == NULL || 1678 imo->imo_membership[i]->inm_ifp == ifp) && 1679 in_hosteq(imo->imo_membership[i]->inm_addr, ia)) 1680 break; 1681 } 1682 if (i == imo->imo_num_memberships) { 1683 error = EADDRNOTAVAIL; 1684 goto out; 1685 } 1686 1687 /* 1688 * Give up the multicast address record to which the 1689 * membership points. 1690 */ 1691 in_delmulti(imo->imo_membership[i]); 1692 1693 /* 1694 * Remove the gap in the membership array. 1695 */ 1696 for (++i; i < imo->imo_num_memberships; ++i) 1697 imo->imo_membership[i-1] = imo->imo_membership[i]; 1698 --imo->imo_num_memberships; 1699 error = 0; 1700 out: 1701 if_put(ifp, &psref); 1702 curlwp_bindx(bound); 1703 return error; 1704 } 1705 1706 /* 1707 * Set the IP multicast options in response to user setsockopt(). 1708 */ 1709 int 1710 ip_setmoptions(struct ip_moptions **pimo, const struct sockopt *sopt) 1711 { 1712 struct ip_moptions *imo = *pimo; 1713 struct in_addr addr; 1714 struct ifnet *ifp; 1715 int ifindex, error = 0; 1716 1717 /* The passed imo isn't NULL, it should be protected by solock */ 1718 1719 if (!imo) { 1720 /* 1721 * No multicast option buffer attached to the pcb; 1722 * allocate one and initialize to default values. 1723 */ 1724 imo = kmem_intr_alloc(sizeof(*imo), KM_NOSLEEP); 1725 if (imo == NULL) 1726 return ENOBUFS; 1727 1728 imo->imo_multicast_if_index = 0; 1729 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1730 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1731 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1732 imo->imo_num_memberships = 0; 1733 *pimo = imo; 1734 } 1735 1736 switch (sopt->sopt_name) { 1737 case IP_MULTICAST_IF: { 1738 int s; 1739 /* 1740 * Select the interface for outgoing multicast packets. 1741 */ 1742 error = sockopt_get(sopt, &addr, sizeof(addr)); 1743 if (error) 1744 break; 1745 1746 /* 1747 * INADDR_ANY is used to remove a previous selection. 1748 * When no interface is selected, a default one is 1749 * chosen every time a multicast packet is sent. 1750 */ 1751 if (in_nullhost(addr)) { 1752 imo->imo_multicast_if_index = 0; 1753 break; 1754 } 1755 /* 1756 * The selected interface is identified by its local 1757 * IP address. Find the interface and confirm that 1758 * it supports multicasting. 1759 */ 1760 s = pserialize_read_enter(); 1761 ifp = ip_multicast_if(&addr, &ifindex); 1762 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1763 pserialize_read_exit(s); 1764 error = EADDRNOTAVAIL; 1765 break; 1766 } 1767 imo->imo_multicast_if_index = ifp->if_index; 1768 pserialize_read_exit(s); 1769 if (ifindex) 1770 imo->imo_multicast_addr = addr; 1771 else 1772 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1773 break; 1774 } 1775 1776 case IP_MULTICAST_TTL: 1777 /* 1778 * Set the IP time-to-live for outgoing multicast packets. 1779 */ 1780 error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL); 1781 break; 1782 1783 case IP_MULTICAST_LOOP: 1784 /* 1785 * Set the loopback flag for outgoing multicast packets. 1786 * Must be zero or one. 1787 */ 1788 error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1); 1789 break; 1790 1791 case IP_ADD_MEMBERSHIP: /* IPV6_JOIN_GROUP */ 1792 error = ip_add_membership(imo, sopt); 1793 break; 1794 1795 case IP_DROP_MEMBERSHIP: /* IPV6_LEAVE_GROUP */ 1796 error = ip_drop_membership(imo, sopt); 1797 break; 1798 1799 default: 1800 error = EOPNOTSUPP; 1801 break; 1802 } 1803 1804 /* 1805 * If all options have default values, no need to keep the mbuf. 1806 */ 1807 if (imo->imo_multicast_if_index == 0 && 1808 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1809 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1810 imo->imo_num_memberships == 0) { 1811 kmem_free(imo, sizeof(*imo)); 1812 *pimo = NULL; 1813 } 1814 1815 return error; 1816 } 1817 1818 /* 1819 * Return the IP multicast options in response to user getsockopt(). 1820 */ 1821 int 1822 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt) 1823 { 1824 struct in_addr addr; 1825 uint8_t optval; 1826 int error = 0; 1827 1828 /* imo is protected by solock or refereced only by the caller */ 1829 1830 switch (sopt->sopt_name) { 1831 case IP_MULTICAST_IF: 1832 if (imo == NULL || imo->imo_multicast_if_index == 0) 1833 addr = zeroin_addr; 1834 else if (imo->imo_multicast_addr.s_addr) { 1835 /* return the value user has set */ 1836 addr = imo->imo_multicast_addr; 1837 } else { 1838 struct ifnet *ifp; 1839 struct in_ifaddr *ia = NULL; 1840 int s = pserialize_read_enter(); 1841 1842 ifp = if_byindex(imo->imo_multicast_if_index); 1843 if (ifp != NULL) { 1844 ia = in_get_ia_from_ifp(ifp); 1845 } 1846 addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1847 pserialize_read_exit(s); 1848 } 1849 error = sockopt_set(sopt, &addr, sizeof(addr)); 1850 break; 1851 1852 case IP_MULTICAST_TTL: 1853 optval = imo ? imo->imo_multicast_ttl 1854 : IP_DEFAULT_MULTICAST_TTL; 1855 1856 error = sockopt_set(sopt, &optval, sizeof(optval)); 1857 break; 1858 1859 case IP_MULTICAST_LOOP: 1860 optval = imo ? imo->imo_multicast_loop 1861 : IP_DEFAULT_MULTICAST_LOOP; 1862 1863 error = sockopt_set(sopt, &optval, sizeof(optval)); 1864 break; 1865 1866 default: 1867 error = EOPNOTSUPP; 1868 } 1869 1870 return error; 1871 } 1872 1873 /* 1874 * Discard the IP multicast options. 1875 */ 1876 void 1877 ip_freemoptions(struct ip_moptions *imo) 1878 { 1879 int i; 1880 1881 /* The owner of imo (inp) should be protected by solock */ 1882 1883 if (imo != NULL) { 1884 for (i = 0; i < imo->imo_num_memberships; ++i) 1885 in_delmulti(imo->imo_membership[i]); 1886 kmem_free(imo, sizeof(*imo)); 1887 } 1888 } 1889 1890 /* 1891 * Routine called from ip_output() to loop back a copy of an IP multicast 1892 * packet to the input queue of a specified interface. Note that this 1893 * calls the output routine of the loopback "driver", but with an interface 1894 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1895 */ 1896 static void 1897 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 1898 { 1899 struct ip *ip; 1900 struct mbuf *copym; 1901 1902 copym = m_copypacket(m, M_DONTWAIT); 1903 if (copym != NULL && 1904 (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1905 copym = m_pullup(copym, sizeof(struct ip)); 1906 if (copym == NULL) 1907 return; 1908 /* 1909 * We don't bother to fragment if the IP length is greater 1910 * than the interface's MTU. Can this possibly matter? 1911 */ 1912 ip = mtod(copym, struct ip *); 1913 1914 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1915 in_delayed_cksum(copym); 1916 copym->m_pkthdr.csum_flags &= 1917 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1918 } 1919 1920 ip->ip_sum = 0; 1921 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1922 #ifndef NET_MPSAFE 1923 KERNEL_LOCK(1, NULL); 1924 #endif 1925 (void)looutput(ifp, copym, sintocsa(dst), NULL); 1926 #ifndef NET_MPSAFE 1927 KERNEL_UNLOCK_ONE(NULL); 1928 #endif 1929 } 1930 1931 /* 1932 * Ensure sending address is valid. 1933 * Returns 0 on success, -1 if an error should be sent back or 1 1934 * if the packet could be dropped without error (protocol dependent). 1935 */ 1936 static int 1937 ip_ifaddrvalid(const struct in_ifaddr *ia) 1938 { 1939 1940 if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY) 1941 return 0; 1942 1943 if (ia->ia4_flags & IN_IFF_DUPLICATED) 1944 return -1; 1945 else if (ia->ia4_flags & (IN_IFF_TENTATIVE | IN_IFF_DETACHED)) 1946 return 1; 1947 1948 return 0; 1949 } 1950
Indexes created Mon Sep 22 21:09:42 GMT 2025