1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_ipfw.h" 36 #include "opt_ipsec.h" 37 #include "opt_route.h" 38 #include "opt_mbuf_stress_test.h" 39 #include "opt_mpath.h" 40 #include "opt_sctp.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/kernel.h> 45 #include <sys/malloc.h> 46 #include <sys/mbuf.h> 47 #include <sys/priv.h> 48 #include <sys/proc.h> 49 #include <sys/protosw.h> 50 #include <sys/socket.h> 51 #include <sys/socketvar.h> 52 #include <sys/sysctl.h> 53 #include <sys/ucred.h> 54 55 #include <net/if.h> 56 #include <net/if_llatbl.h> 57 #include <net/netisr.h> 58 #include <net/pfil.h> 59 #include <net/route.h> 60 #include <net/flowtable.h> 61 #ifdef RADIX_MPATH 62 #include <net/radix_mpath.h> 63 #endif 64 #include <net/vnet.h> 65 66 #include <netinet/in.h> 67 #include <netinet/in_systm.h> 68 #include <netinet/ip.h> 69 #include <netinet/in_pcb.h> 70 #include <netinet/in_var.h> 71 #include <netinet/ip_var.h> 72 #include <netinet/ip_options.h> 73 #ifdef SCTP 74 #include <netinet/sctp.h> 75 #include <netinet/sctp_crc32.h> 76 #endif 77 78 #ifdef IPSEC 79 #include <netinet/ip_ipsec.h> 80 #include <netipsec/ipsec.h> 81 #endif /* IPSEC*/ 82 83 #include <machine/in_cksum.h> 84 85 #include <security/mac/mac_framework.h> 86 87 VNET_DEFINE(u_short, ip_id); 88 89 #ifdef MBUF_STRESS_TEST 90 static int mbuf_frag_size = 0; 91 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 92 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 93 #endif 94 95 static void ip_mloopback 96 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 97 98 99 extern int in_mcast_loop; 100 extern struct protosw inetsw[]; 101 102 /* 103 * IP output. The packet in mbuf chain m contains a skeletal IP 104 * header (with len, off, ttl, proto, tos, src, dst). 105 * The mbuf chain containing the packet will be freed. 106 * The mbuf opt, if present, will not be freed. 107 * If route ro is present and has ro_rt initialized, route lookup would be 108 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL, 109 * then result of route lookup is stored in ro->ro_rt. 110 * 111 * In the IP forwarding case, the packet will arrive with options already 112 * inserted, so must have a NULL opt pointer. 113 */ 114 int 115 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags, 116 struct ip_moptions *imo, struct inpcb *inp) 117 { 118 struct ip *ip; 119 struct ifnet *ifp = NULL; /* keep compiler happy */ 120 struct mbuf *m0; 121 int hlen = sizeof (struct ip); 122 int mtu; 123 int n; /* scratchpad */ 124 int error = 0; 125 struct sockaddr_in *dst; 126 const struct sockaddr_in *gw; 127 struct in_ifaddr *ia; 128 int isbroadcast; 129 uint16_t ip_len, ip_off; 130 struct route iproute; 131 struct rtentry *rte; /* cache for ro->ro_rt */ 132 struct in_addr odst; 133 struct m_tag *fwd_tag = NULL; 134 #ifdef IPSEC 135 int no_route_but_check_spd = 0; 136 #endif 137 M_ASSERTPKTHDR(m); 138 139 if (inp != NULL) { 140 INP_LOCK_ASSERT(inp); 141 M_SETFIB(m, inp->inp_inc.inc_fibnum); 142 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) { 143 m->m_pkthdr.flowid = inp->inp_flowid; 144 m->m_flags |= M_FLOWID; 145 } 146 } 147 148 if (ro == NULL) { 149 ro = &iproute; 150 bzero(ro, sizeof (*ro)); 151 } 152 153 #ifdef FLOWTABLE 154 if (ro->ro_rt == NULL) { 155 struct flentry *fle; 156 157 /* 158 * The flow table returns route entries valid for up to 30 159 * seconds; we rely on the remainder of ip_output() taking no 160 * longer than that long for the stability of ro_rt. The 161 * flow ID assignment must have happened before this point. 162 */ 163 fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET); 164 if (fle != NULL) 165 flow_to_route(fle, ro); 166 } 167 #endif 168 169 if (opt) { 170 int len = 0; 171 m = ip_insertoptions(m, opt, &len); 172 if (len != 0) 173 hlen = len; /* ip->ip_hl is updated above */ 174 } 175 ip = mtod(m, struct ip *); 176 ip_len = ntohs(ip->ip_len); 177 ip_off = ntohs(ip->ip_off); 178 179 /* 180 * Fill in IP header. If we are not allowing fragmentation, 181 * then the ip_id field is meaningless, but we don't set it 182 * to zero. Doing so causes various problems when devices along 183 * the path (routers, load balancers, firewalls, etc.) illegally 184 * disable DF on our packet. Note that a 16-bit counter 185 * will wrap around in less than 10 seconds at 100 Mbit/s on a 186 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 187 * for Counting NATted Hosts", Proc. IMW'02, available at 188 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>. 189 */ 190 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 191 ip->ip_v = IPVERSION; 192 ip->ip_hl = hlen >> 2; 193 ip->ip_id = ip_newid(); 194 IPSTAT_INC(ips_localout); 195 } else { 196 /* Header already set, fetch hlen from there */ 197 hlen = ip->ip_hl << 2; 198 } 199 200 gw = dst = (struct sockaddr_in *)&ro->ro_dst; 201 again: 202 ia = NULL; 203 /* 204 * If there is a cached route, 205 * check that it is to the same destination 206 * and is still up. If not, free it and try again. 207 * The address family should also be checked in case of sharing the 208 * cache with IPv6. 209 */ 210 rte = ro->ro_rt; 211 if (rte && ((rte->rt_flags & RTF_UP) == 0 || 212 rte->rt_ifp == NULL || 213 !RT_LINK_IS_UP(rte->rt_ifp) || 214 dst->sin_family != AF_INET || 215 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 216 RO_RTFREE(ro); 217 ro->ro_lle = NULL; 218 rte = NULL; 219 } 220 if (rte == NULL && fwd_tag == NULL) { 221 bzero(dst, sizeof(*dst)); 222 dst->sin_family = AF_INET; 223 dst->sin_len = sizeof(*dst); 224 dst->sin_addr = ip->ip_dst; 225 } 226 /* 227 * If routing to interface only, short circuit routing lookup. 228 * The use of an all-ones broadcast address implies this; an 229 * interface is specified by the broadcast address of an interface, 230 * or the destination address of a ptp interface. 231 */ 232 if (flags & IP_SENDONES) { 233 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL && 234 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) { 235 IPSTAT_INC(ips_noroute); 236 error = ENETUNREACH; 237 goto bad; 238 } 239 ip->ip_dst.s_addr = INADDR_BROADCAST; 240 dst->sin_addr = ip->ip_dst; 241 ifp = ia->ia_ifp; 242 ip->ip_ttl = 1; 243 isbroadcast = 1; 244 } else if (flags & IP_ROUTETOIF) { 245 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 246 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) { 247 IPSTAT_INC(ips_noroute); 248 error = ENETUNREACH; 249 goto bad; 250 } 251 ifp = ia->ia_ifp; 252 ip->ip_ttl = 1; 253 isbroadcast = in_broadcast(dst->sin_addr, ifp); 254 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 255 imo != NULL && imo->imo_multicast_ifp != NULL) { 256 /* 257 * Bypass the normal routing lookup for multicast 258 * packets if the interface is specified. 259 */ 260 ifp = imo->imo_multicast_ifp; 261 IFP_TO_IA(ifp, ia); 262 isbroadcast = 0; /* fool gcc */ 263 } else { 264 /* 265 * We want to do any cloning requested by the link layer, 266 * as this is probably required in all cases for correct 267 * operation (as it is for ARP). 268 */ 269 if (rte == NULL) { 270 #ifdef RADIX_MPATH 271 rtalloc_mpath_fib(ro, 272 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), 273 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 274 #else 275 in_rtalloc_ign(ro, 0, 276 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 277 #endif 278 rte = ro->ro_rt; 279 } 280 if (rte == NULL || 281 rte->rt_ifp == NULL || 282 !RT_LINK_IS_UP(rte->rt_ifp)) { 283 #ifdef IPSEC 284 /* 285 * There is no route for this packet, but it is 286 * possible that a matching SPD entry exists. 287 */ 288 no_route_but_check_spd = 1; 289 mtu = 0; /* Silence GCC warning. */ 290 goto sendit; 291 #endif 292 IPSTAT_INC(ips_noroute); 293 error = EHOSTUNREACH; 294 goto bad; 295 } 296 ia = ifatoia(rte->rt_ifa); 297 ifa_ref(&ia->ia_ifa); 298 ifp = rte->rt_ifp; 299 rte->rt_rmx.rmx_pksent++; 300 if (rte->rt_flags & RTF_GATEWAY) 301 gw = (struct sockaddr_in *)rte->rt_gateway; 302 if (rte->rt_flags & RTF_HOST) 303 isbroadcast = (rte->rt_flags & RTF_BROADCAST); 304 else 305 isbroadcast = in_broadcast(gw->sin_addr, ifp); 306 } 307 /* 308 * Calculate MTU. If we have a route that is up, use that, 309 * otherwise use the interface's MTU. 310 */ 311 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) { 312 /* 313 * This case can happen if the user changed the MTU 314 * of an interface after enabling IP on it. Because 315 * most netifs don't keep track of routes pointing to 316 * them, there is no way for one to update all its 317 * routes when the MTU is changed. 318 */ 319 if (rte->rt_rmx.rmx_mtu > ifp->if_mtu) 320 rte->rt_rmx.rmx_mtu = ifp->if_mtu; 321 mtu = rte->rt_rmx.rmx_mtu; 322 } else { 323 mtu = ifp->if_mtu; 324 } 325 /* Catch a possible divide by zero later. */ 326 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p", 327 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp)); 328 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 329 m->m_flags |= M_MCAST; 330 /* 331 * See if the caller provided any multicast options 332 */ 333 if (imo != NULL) { 334 ip->ip_ttl = imo->imo_multicast_ttl; 335 if (imo->imo_multicast_vif != -1) 336 ip->ip_src.s_addr = 337 ip_mcast_src ? 338 ip_mcast_src(imo->imo_multicast_vif) : 339 INADDR_ANY; 340 } else 341 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 342 /* 343 * Confirm that the outgoing interface supports multicast. 344 */ 345 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 346 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 347 IPSTAT_INC(ips_noroute); 348 error = ENETUNREACH; 349 goto bad; 350 } 351 } 352 /* 353 * If source address not specified yet, use address 354 * of outgoing interface. 355 */ 356 if (ip->ip_src.s_addr == INADDR_ANY) { 357 /* Interface may have no addresses. */ 358 if (ia != NULL) 359 ip->ip_src = IA_SIN(ia)->sin_addr; 360 } 361 362 if ((imo == NULL && in_mcast_loop) || 363 (imo && imo->imo_multicast_loop)) { 364 /* 365 * Loop back multicast datagram if not expressly 366 * forbidden to do so, even if we are not a member 367 * of the group; ip_input() will filter it later, 368 * thus deferring a hash lookup and mutex acquisition 369 * at the expense of a cheap copy using m_copym(). 370 */ 371 ip_mloopback(ifp, m, dst, hlen); 372 } else { 373 /* 374 * If we are acting as a multicast router, perform 375 * multicast forwarding as if the packet had just 376 * arrived on the interface to which we are about 377 * to send. The multicast forwarding function 378 * recursively calls this function, using the 379 * IP_FORWARDING flag to prevent infinite recursion. 380 * 381 * Multicasts that are looped back by ip_mloopback(), 382 * above, will be forwarded by the ip_input() routine, 383 * if necessary. 384 */ 385 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) { 386 /* 387 * If rsvp daemon is not running, do not 388 * set ip_moptions. This ensures that the packet 389 * is multicast and not just sent down one link 390 * as prescribed by rsvpd. 391 */ 392 if (!V_rsvp_on) 393 imo = NULL; 394 if (ip_mforward && 395 ip_mforward(ip, ifp, m, imo) != 0) { 396 m_freem(m); 397 goto done; 398 } 399 } 400 } 401 402 /* 403 * Multicasts with a time-to-live of zero may be looped- 404 * back, above, but must not be transmitted on a network. 405 * Also, multicasts addressed to the loopback interface 406 * are not sent -- the above call to ip_mloopback() will 407 * loop back a copy. ip_input() will drop the copy if 408 * this host does not belong to the destination group on 409 * the loopback interface. 410 */ 411 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 412 m_freem(m); 413 goto done; 414 } 415 416 goto sendit; 417 } 418 419 /* 420 * If the source address is not specified yet, use the address 421 * of the outoing interface. 422 */ 423 if (ip->ip_src.s_addr == INADDR_ANY) { 424 /* Interface may have no addresses. */ 425 if (ia != NULL) { 426 ip->ip_src = IA_SIN(ia)->sin_addr; 427 } 428 } 429 430 /* 431 * Verify that we have any chance at all of being able to queue the 432 * packet or packet fragments, unless ALTQ is enabled on the given 433 * interface in which case packetdrop should be done by queueing. 434 */ 435 n = ip_len / mtu + 1; /* how many fragments ? */ 436 if ( 437 #ifdef ALTQ 438 (!ALTQ_IS_ENABLED(&ifp->if_snd)) && 439 #endif /* ALTQ */ 440 (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) { 441 error = ENOBUFS; 442 IPSTAT_INC(ips_odropped); 443 ifp->if_snd.ifq_drops += n; 444 goto bad; 445 } 446 447 /* 448 * Look for broadcast address and 449 * verify user is allowed to send 450 * such a packet. 451 */ 452 if (isbroadcast) { 453 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 454 error = EADDRNOTAVAIL; 455 goto bad; 456 } 457 if ((flags & IP_ALLOWBROADCAST) == 0) { 458 error = EACCES; 459 goto bad; 460 } 461 /* don't allow broadcast messages to be fragmented */ 462 if (ip_len > mtu) { 463 error = EMSGSIZE; 464 goto bad; 465 } 466 m->m_flags |= M_BCAST; 467 } else { 468 m->m_flags &= ~M_BCAST; 469 } 470 471 sendit: 472 #ifdef IPSEC 473 switch(ip_ipsec_output(&m, inp, &flags, &error)) { 474 case 1: 475 goto bad; 476 case -1: 477 goto done; 478 case 0: 479 default: 480 break; /* Continue with packet processing. */ 481 } 482 /* 483 * Check if there was a route for this packet; return error if not. 484 */ 485 if (no_route_but_check_spd) { 486 IPSTAT_INC(ips_noroute); 487 error = EHOSTUNREACH; 488 goto bad; 489 } 490 /* Update variables that are affected by ipsec4_output(). */ 491 ip = mtod(m, struct ip *); 492 hlen = ip->ip_hl << 2; 493 #endif /* IPSEC */ 494 495 /* Jump over all PFIL processing if hooks are not active. */ 496 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 497 goto passout; 498 499 /* Run through list of hooks for output packets. */ 500 odst.s_addr = ip->ip_dst.s_addr; 501 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 502 if (error != 0 || m == NULL) 503 goto done; 504 505 ip = mtod(m, struct ip *); 506 507 /* See if destination IP address was changed by packet filter. */ 508 if (odst.s_addr != ip->ip_dst.s_addr) { 509 m->m_flags |= M_SKIP_FIREWALL; 510 /* If destination is now ourself drop to ip_input(). */ 511 if (in_localip(ip->ip_dst)) { 512 m->m_flags |= M_FASTFWD_OURS; 513 if (m->m_pkthdr.rcvif == NULL) 514 m->m_pkthdr.rcvif = V_loif; 515 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 516 m->m_pkthdr.csum_flags |= 517 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 518 m->m_pkthdr.csum_data = 0xffff; 519 } 520 m->m_pkthdr.csum_flags |= 521 CSUM_IP_CHECKED | CSUM_IP_VALID; 522 #ifdef SCTP 523 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 524 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 525 #endif 526 error = netisr_queue(NETISR_IP, m); 527 goto done; 528 } else { 529 if (ia != NULL) 530 ifa_free(&ia->ia_ifa); 531 goto again; /* Redo the routing table lookup. */ 532 } 533 } 534 535 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ 536 if (m->m_flags & M_FASTFWD_OURS) { 537 if (m->m_pkthdr.rcvif == NULL) 538 m->m_pkthdr.rcvif = V_loif; 539 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 540 m->m_pkthdr.csum_flags |= 541 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 542 m->m_pkthdr.csum_data = 0xffff; 543 } 544 #ifdef SCTP 545 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 546 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 547 #endif 548 m->m_pkthdr.csum_flags |= 549 CSUM_IP_CHECKED | CSUM_IP_VALID; 550 551 error = netisr_queue(NETISR_IP, m); 552 goto done; 553 } 554 /* Or forward to some other address? */ 555 if ((m->m_flags & M_IP_NEXTHOP) && 556 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 557 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 558 m->m_flags |= M_SKIP_FIREWALL; 559 m->m_flags &= ~M_IP_NEXTHOP; 560 m_tag_delete(m, fwd_tag); 561 if (ia != NULL) 562 ifa_free(&ia->ia_ifa); 563 goto again; 564 } 565 566 passout: 567 /* 127/8 must not appear on wire - RFC1122. */ 568 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 569 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 570 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 571 IPSTAT_INC(ips_badaddr); 572 error = EADDRNOTAVAIL; 573 goto bad; 574 } 575 } 576 577 m->m_pkthdr.csum_flags |= CSUM_IP; 578 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) { 579 in_delayed_cksum(m); 580 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 581 } 582 #ifdef SCTP 583 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) { 584 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); 585 m->m_pkthdr.csum_flags &= ~CSUM_SCTP; 586 } 587 #endif 588 589 /* 590 * If small enough for interface, or the interface will take 591 * care of the fragmentation for us, we can just send directly. 592 */ 593 if (ip_len <= mtu || 594 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 || 595 ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) { 596 ip->ip_sum = 0; 597 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) { 598 ip->ip_sum = in_cksum(m, hlen); 599 m->m_pkthdr.csum_flags &= ~CSUM_IP; 600 } 601 602 /* 603 * Record statistics for this interface address. 604 * With CSUM_TSO the byte/packet count will be slightly 605 * incorrect because we count the IP+TCP headers only 606 * once instead of for every generated packet. 607 */ 608 if (!(flags & IP_FORWARDING) && ia) { 609 if (m->m_pkthdr.csum_flags & CSUM_TSO) 610 ia->ia_ifa.if_opackets += 611 m->m_pkthdr.len / m->m_pkthdr.tso_segsz; 612 else 613 ia->ia_ifa.if_opackets++; 614 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 615 } 616 #ifdef MBUF_STRESS_TEST 617 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 618 m = m_fragment(m, M_NOWAIT, mbuf_frag_size); 619 #endif 620 /* 621 * Reset layer specific mbuf flags 622 * to avoid confusing lower layers. 623 */ 624 m->m_flags &= ~(M_PROTOFLAGS); 625 error = (*ifp->if_output)(ifp, m, 626 (const struct sockaddr *)gw, ro); 627 goto done; 628 } 629 630 /* Balk when DF bit is set or the interface didn't support TSO. */ 631 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { 632 error = EMSGSIZE; 633 IPSTAT_INC(ips_cantfrag); 634 goto bad; 635 } 636 637 /* 638 * Too large for interface; fragment if possible. If successful, 639 * on return, m will point to a list of packets to be sent. 640 */ 641 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist); 642 if (error) 643 goto bad; 644 for (; m; m = m0) { 645 m0 = m->m_nextpkt; 646 m->m_nextpkt = 0; 647 if (error == 0) { 648 /* Record statistics for this interface address. */ 649 if (ia != NULL) { 650 ia->ia_ifa.if_opackets++; 651 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 652 } 653 /* 654 * Reset layer specific mbuf flags 655 * to avoid confusing upper layers. 656 */ 657 m->m_flags &= ~(M_PROTOFLAGS); 658 659 error = (*ifp->if_output)(ifp, m, 660 (const struct sockaddr *)gw, ro); 661 } else 662 m_freem(m); 663 } 664 665 if (error == 0) 666 IPSTAT_INC(ips_fragmented); 667 668 done: 669 if (ro == &iproute) 670 RO_RTFREE(ro); 671 if (ia != NULL) 672 ifa_free(&ia->ia_ifa); 673 return (error); 674 bad: 675 m_freem(m); 676 goto done; 677 } 678 679 /* 680 * Create a chain of fragments which fit the given mtu. m_frag points to the 681 * mbuf to be fragmented; on return it points to the chain with the fragments. 682 * Return 0 if no error. If error, m_frag may contain a partially built 683 * chain of fragments that should be freed by the caller. 684 * 685 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 686 */ 687 int 688 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 689 u_long if_hwassist_flags) 690 { 691 int error = 0; 692 int hlen = ip->ip_hl << 2; 693 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 694 int off; 695 struct mbuf *m0 = *m_frag; /* the original packet */ 696 int firstlen; 697 struct mbuf **mnext; 698 int nfrags; 699 uint16_t ip_len, ip_off; 700 701 ip_len = ntohs(ip->ip_len); 702 ip_off = ntohs(ip->ip_off); 703 704 if (ip_off & IP_DF) { /* Fragmentation not allowed */ 705 IPSTAT_INC(ips_cantfrag); 706 return EMSGSIZE; 707 } 708 709 /* 710 * Must be able to put at least 8 bytes per fragment. 711 */ 712 if (len < 8) 713 return EMSGSIZE; 714 715 /* 716 * If the interface will not calculate checksums on 717 * fragmented packets, then do it here. 718 */ 719 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 720 in_delayed_cksum(m0); 721 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 722 } 723 #ifdef SCTP 724 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) { 725 sctp_delayed_cksum(m0, hlen); 726 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP; 727 } 728 #endif 729 if (len > PAGE_SIZE) { 730 /* 731 * Fragment large datagrams such that each segment 732 * contains a multiple of PAGE_SIZE amount of data, 733 * plus headers. This enables a receiver to perform 734 * page-flipping zero-copy optimizations. 735 * 736 * XXX When does this help given that sender and receiver 737 * could have different page sizes, and also mtu could 738 * be less than the receiver's page size ? 739 */ 740 int newlen; 741 struct mbuf *m; 742 743 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 744 off += m->m_len; 745 746 /* 747 * firstlen (off - hlen) must be aligned on an 748 * 8-byte boundary 749 */ 750 if (off < hlen) 751 goto smart_frag_failure; 752 off = ((off - hlen) & ~7) + hlen; 753 newlen = (~PAGE_MASK) & mtu; 754 if ((newlen + sizeof (struct ip)) > mtu) { 755 /* we failed, go back the default */ 756 smart_frag_failure: 757 newlen = len; 758 off = hlen + len; 759 } 760 len = newlen; 761 762 } else { 763 off = hlen + len; 764 } 765 766 firstlen = off - hlen; 767 mnext = &m0->m_nextpkt; /* pointer to next packet */ 768 769 /* 770 * Loop through length of segment after first fragment, 771 * make new header and copy data of each part and link onto chain. 772 * Here, m0 is the original packet, m is the fragment being created. 773 * The fragments are linked off the m_nextpkt of the original 774 * packet, which after processing serves as the first fragment. 775 */ 776 for (nfrags = 1; off < ip_len; off += len, nfrags++) { 777 struct ip *mhip; /* ip header on the fragment */ 778 struct mbuf *m; 779 int mhlen = sizeof (struct ip); 780 781 m = m_gethdr(M_NOWAIT, MT_DATA); 782 if (m == NULL) { 783 error = ENOBUFS; 784 IPSTAT_INC(ips_odropped); 785 goto done; 786 } 787 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 788 /* 789 * In the first mbuf, leave room for the link header, then 790 * copy the original IP header including options. The payload 791 * goes into an additional mbuf chain returned by m_copym(). 792 */ 793 m->m_data += max_linkhdr; 794 mhip = mtod(m, struct ip *); 795 *mhip = *ip; 796 if (hlen > sizeof (struct ip)) { 797 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 798 mhip->ip_v = IPVERSION; 799 mhip->ip_hl = mhlen >> 2; 800 } 801 m->m_len = mhlen; 802 /* XXX do we need to add ip_off below ? */ 803 mhip->ip_off = ((off - hlen) >> 3) + ip_off; 804 if (off + len >= ip_len) { /* last fragment */ 805 len = ip_len - off; 806 m->m_flags |= M_LASTFRAG; 807 } else 808 mhip->ip_off |= IP_MF; 809 mhip->ip_len = htons((u_short)(len + mhlen)); 810 m->m_next = m_copym(m0, off, len, M_NOWAIT); 811 if (m->m_next == NULL) { /* copy failed */ 812 m_free(m); 813 error = ENOBUFS; /* ??? */ 814 IPSTAT_INC(ips_odropped); 815 goto done; 816 } 817 m->m_pkthdr.len = mhlen + len; 818 m->m_pkthdr.rcvif = NULL; 819 #ifdef MAC 820 mac_netinet_fragment(m0, m); 821 #endif 822 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 823 mhip->ip_off = htons(mhip->ip_off); 824 mhip->ip_sum = 0; 825 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { 826 mhip->ip_sum = in_cksum(m, mhlen); 827 m->m_pkthdr.csum_flags &= ~CSUM_IP; 828 } 829 *mnext = m; 830 mnext = &m->m_nextpkt; 831 } 832 IPSTAT_ADD(ips_ofragments, nfrags); 833 834 /* set first marker for fragment chain */ 835 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 836 m0->m_pkthdr.csum_data = nfrags; 837 838 /* 839 * Update first fragment by trimming what's been copied out 840 * and updating header. 841 */ 842 m_adj(m0, hlen + firstlen - ip_len); 843 m0->m_pkthdr.len = hlen + firstlen; 844 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 845 ip->ip_off = htons(ip_off | IP_MF); 846 ip->ip_sum = 0; 847 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { 848 ip->ip_sum = in_cksum(m0, hlen); 849 m0->m_pkthdr.csum_flags &= ~CSUM_IP; 850 } 851 852 done: 853 *m_frag = m0; 854 return error; 855 } 856 857 void 858 in_delayed_cksum(struct mbuf *m) 859 { 860 struct ip *ip; 861 uint16_t csum, offset, ip_len; 862 863 ip = mtod(m, struct ip *); 864 offset = ip->ip_hl << 2 ; 865 ip_len = ntohs(ip->ip_len); 866 csum = in_cksum_skip(m, ip_len, offset); 867 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 868 csum = 0xffff; 869 offset += m->m_pkthdr.csum_data; /* checksum offset */ 870 871 if (offset + sizeof(u_short) > m->m_len) { 872 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 873 m->m_len, offset, ip->ip_p); 874 /* 875 * XXX 876 * this shouldn't happen, but if it does, the 877 * correct behavior may be to insert the checksum 878 * in the appropriate next mbuf in the chain. 879 */ 880 return; 881 } 882 *(u_short *)(m->m_data + offset) = csum; 883 } 884 885 /* 886 * IP socket option processing. 887 */ 888 int 889 ip_ctloutput(struct socket *so, struct sockopt *sopt) 890 { 891 struct inpcb *inp = sotoinpcb(so); 892 int error, optval; 893 894 error = optval = 0; 895 if (sopt->sopt_level != IPPROTO_IP) { 896 error = EINVAL; 897 898 if (sopt->sopt_level == SOL_SOCKET && 899 sopt->sopt_dir == SOPT_SET) { 900 switch (sopt->sopt_name) { 901 case SO_REUSEADDR: 902 INP_WLOCK(inp); 903 if (IN_MULTICAST(ntohl(inp->inp_laddr.s_addr))) { 904 if ((so->so_options & 905 (SO_REUSEADDR | SO_REUSEPORT)) != 0) 906 inp->inp_flags2 |= INP_REUSEPORT; 907 else 908 inp->inp_flags2 &= ~INP_REUSEPORT; 909 } 910 INP_WUNLOCK(inp); 911 error = 0; 912 break; 913 case SO_REUSEPORT: 914 INP_WLOCK(inp); 915 if ((so->so_options & SO_REUSEPORT) != 0) 916 inp->inp_flags2 |= INP_REUSEPORT; 917 else 918 inp->inp_flags2 &= ~INP_REUSEPORT; 919 INP_WUNLOCK(inp); 920 error = 0; 921 break; 922 case SO_SETFIB: 923 INP_WLOCK(inp); 924 inp->inp_inc.inc_fibnum = so->so_fibnum; 925 INP_WUNLOCK(inp); 926 error = 0; 927 break; 928 default: 929 break; 930 } 931 } 932 return (error); 933 } 934 935 switch (sopt->sopt_dir) { 936 case SOPT_SET: 937 switch (sopt->sopt_name) { 938 case IP_OPTIONS: 939 #ifdef notyet 940 case IP_RETOPTS: 941 #endif 942 { 943 struct mbuf *m; 944 if (sopt->sopt_valsize > MLEN) { 945 error = EMSGSIZE; 946 break; 947 } 948 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA); 949 if (m == NULL) { 950 error = ENOBUFS; 951 break; 952 } 953 m->m_len = sopt->sopt_valsize; 954 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 955 m->m_len); 956 if (error) { 957 m_free(m); 958 break; 959 } 960 INP_WLOCK(inp); 961 error = ip_pcbopts(inp, sopt->sopt_name, m); 962 INP_WUNLOCK(inp); 963 return (error); 964 } 965 966 case IP_BINDANY: 967 if (sopt->sopt_td != NULL) { 968 error = priv_check(sopt->sopt_td, 969 PRIV_NETINET_BINDANY); 970 if (error) 971 break; 972 } 973 /* FALLTHROUGH */ 974 case IP_TOS: 975 case IP_TTL: 976 case IP_MINTTL: 977 case IP_RECVOPTS: 978 case IP_RECVRETOPTS: 979 case IP_RECVDSTADDR: 980 case IP_RECVTTL: 981 case IP_RECVIF: 982 case IP_FAITH: 983 case IP_ONESBCAST: 984 case IP_DONTFRAG: 985 case IP_RECVTOS: 986 error = sooptcopyin(sopt, &optval, sizeof optval, 987 sizeof optval); 988 if (error) 989 break; 990 991 switch (sopt->sopt_name) { 992 case IP_TOS: 993 inp->inp_ip_tos = optval; 994 break; 995 996 case IP_TTL: 997 inp->inp_ip_ttl = optval; 998 break; 999 1000 case IP_MINTTL: 1001 if (optval >= 0 && optval <= MAXTTL) 1002 inp->inp_ip_minttl = optval; 1003 else 1004 error = EINVAL; 1005 break; 1006 1007 #define OPTSET(bit) do { \ 1008 INP_WLOCK(inp); \ 1009 if (optval) \ 1010 inp->inp_flags |= bit; \ 1011 else \ 1012 inp->inp_flags &= ~bit; \ 1013 INP_WUNLOCK(inp); \ 1014 } while (0) 1015 1016 case IP_RECVOPTS: 1017 OPTSET(INP_RECVOPTS); 1018 break; 1019 1020 case IP_RECVRETOPTS: 1021 OPTSET(INP_RECVRETOPTS); 1022 break; 1023 1024 case IP_RECVDSTADDR: 1025 OPTSET(INP_RECVDSTADDR); 1026 break; 1027 1028 case IP_RECVTTL: 1029 OPTSET(INP_RECVTTL); 1030 break; 1031 1032 case IP_RECVIF: 1033 OPTSET(INP_RECVIF); 1034 break; 1035 1036 case IP_FAITH: 1037 OPTSET(INP_FAITH); 1038 break; 1039 1040 case IP_ONESBCAST: 1041 OPTSET(INP_ONESBCAST); 1042 break; 1043 case IP_DONTFRAG: 1044 OPTSET(INP_DONTFRAG); 1045 break; 1046 case IP_BINDANY: 1047 OPTSET(INP_BINDANY); 1048 break; 1049 case IP_RECVTOS: 1050 OPTSET(INP_RECVTOS); 1051 break; 1052 } 1053 break; 1054 #undef OPTSET 1055 1056 /* 1057 * Multicast socket options are processed by the in_mcast 1058 * module. 1059 */ 1060 case IP_MULTICAST_IF: 1061 case IP_MULTICAST_VIF: 1062 case IP_MULTICAST_TTL: 1063 case IP_MULTICAST_LOOP: 1064 case IP_ADD_MEMBERSHIP: 1065 case IP_DROP_MEMBERSHIP: 1066 case IP_ADD_SOURCE_MEMBERSHIP: 1067 case IP_DROP_SOURCE_MEMBERSHIP: 1068 case IP_BLOCK_SOURCE: 1069 case IP_UNBLOCK_SOURCE: 1070 case IP_MSFILTER: 1071 case MCAST_JOIN_GROUP: 1072 case MCAST_LEAVE_GROUP: 1073 case MCAST_JOIN_SOURCE_GROUP: 1074 case MCAST_LEAVE_SOURCE_GROUP: 1075 case MCAST_BLOCK_SOURCE: 1076 case MCAST_UNBLOCK_SOURCE: 1077 error = inp_setmoptions(inp, sopt); 1078 break; 1079 1080 case IP_PORTRANGE: 1081 error = sooptcopyin(sopt, &optval, sizeof optval, 1082 sizeof optval); 1083 if (error) 1084 break; 1085 1086 INP_WLOCK(inp); 1087 switch (optval) { 1088 case IP_PORTRANGE_DEFAULT: 1089 inp->inp_flags &= ~(INP_LOWPORT); 1090 inp->inp_flags &= ~(INP_HIGHPORT); 1091 break; 1092 1093 case IP_PORTRANGE_HIGH: 1094 inp->inp_flags &= ~(INP_LOWPORT); 1095 inp->inp_flags |= INP_HIGHPORT; 1096 break; 1097 1098 case IP_PORTRANGE_LOW: 1099 inp->inp_flags &= ~(INP_HIGHPORT); 1100 inp->inp_flags |= INP_LOWPORT; 1101 break; 1102 1103 default: 1104 error = EINVAL; 1105 break; 1106 } 1107 INP_WUNLOCK(inp); 1108 break; 1109 1110 #ifdef IPSEC 1111 case IP_IPSEC_POLICY: 1112 { 1113 caddr_t req; 1114 struct mbuf *m; 1115 1116 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1117 break; 1118 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1119 break; 1120 req = mtod(m, caddr_t); 1121 error = ipsec_set_policy(inp, sopt->sopt_name, req, 1122 m->m_len, (sopt->sopt_td != NULL) ? 1123 sopt->sopt_td->td_ucred : NULL); 1124 m_freem(m); 1125 break; 1126 } 1127 #endif /* IPSEC */ 1128 1129 default: 1130 error = ENOPROTOOPT; 1131 break; 1132 } 1133 break; 1134 1135 case SOPT_GET: 1136 switch (sopt->sopt_name) { 1137 case IP_OPTIONS: 1138 case IP_RETOPTS: 1139 if (inp->inp_options) 1140 error = sooptcopyout(sopt, 1141 mtod(inp->inp_options, 1142 char *), 1143 inp->inp_options->m_len); 1144 else 1145 sopt->sopt_valsize = 0; 1146 break; 1147 1148 case IP_TOS: 1149 case IP_TTL: 1150 case IP_MINTTL: 1151 case IP_RECVOPTS: 1152 case IP_RECVRETOPTS: 1153 case IP_RECVDSTADDR: 1154 case IP_RECVTTL: 1155 case IP_RECVIF: 1156 case IP_PORTRANGE: 1157 case IP_FAITH: 1158 case IP_ONESBCAST: 1159 case IP_DONTFRAG: 1160 case IP_BINDANY: 1161 case IP_RECVTOS: 1162 switch (sopt->sopt_name) { 1163 1164 case IP_TOS: 1165 optval = inp->inp_ip_tos; 1166 break; 1167 1168 case IP_TTL: 1169 optval = inp->inp_ip_ttl; 1170 break; 1171 1172 case IP_MINTTL: 1173 optval = inp->inp_ip_minttl; 1174 break; 1175 1176 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1177 1178 case IP_RECVOPTS: 1179 optval = OPTBIT(INP_RECVOPTS); 1180 break; 1181 1182 case IP_RECVRETOPTS: 1183 optval = OPTBIT(INP_RECVRETOPTS); 1184 break; 1185 1186 case IP_RECVDSTADDR: 1187 optval = OPTBIT(INP_RECVDSTADDR); 1188 break; 1189 1190 case IP_RECVTTL: 1191 optval = OPTBIT(INP_RECVTTL); 1192 break; 1193 1194 case IP_RECVIF: 1195 optval = OPTBIT(INP_RECVIF); 1196 break; 1197 1198 case IP_PORTRANGE: 1199 if (inp->inp_flags & INP_HIGHPORT) 1200 optval = IP_PORTRANGE_HIGH; 1201 else if (inp->inp_flags & INP_LOWPORT) 1202 optval = IP_PORTRANGE_LOW; 1203 else 1204 optval = 0; 1205 break; 1206 1207 case IP_FAITH: 1208 optval = OPTBIT(INP_FAITH); 1209 break; 1210 1211 case IP_ONESBCAST: 1212 optval = OPTBIT(INP_ONESBCAST); 1213 break; 1214 case IP_DONTFRAG: 1215 optval = OPTBIT(INP_DONTFRAG); 1216 break; 1217 case IP_BINDANY: 1218 optval = OPTBIT(INP_BINDANY); 1219 break; 1220 case IP_RECVTOS: 1221 optval = OPTBIT(INP_RECVTOS); 1222 break; 1223 } 1224 error = sooptcopyout(sopt, &optval, sizeof optval); 1225 break; 1226 1227 /* 1228 * Multicast socket options are processed by the in_mcast 1229 * module. 1230 */ 1231 case IP_MULTICAST_IF: 1232 case IP_MULTICAST_VIF: 1233 case IP_MULTICAST_TTL: 1234 case IP_MULTICAST_LOOP: 1235 case IP_MSFILTER: 1236 error = inp_getmoptions(inp, sopt); 1237 break; 1238 1239 #ifdef IPSEC 1240 case IP_IPSEC_POLICY: 1241 { 1242 struct mbuf *m = NULL; 1243 caddr_t req = NULL; 1244 size_t len = 0; 1245 1246 if (m != 0) { 1247 req = mtod(m, caddr_t); 1248 len = m->m_len; 1249 } 1250 error = ipsec_get_policy(sotoinpcb(so), req, len, &m); 1251 if (error == 0) 1252 error = soopt_mcopyout(sopt, m); /* XXX */ 1253 if (error == 0) 1254 m_freem(m); 1255 break; 1256 } 1257 #endif /* IPSEC */ 1258 1259 default: 1260 error = ENOPROTOOPT; 1261 break; 1262 } 1263 break; 1264 } 1265 return (error); 1266 } 1267 1268 /* 1269 * Routine called from ip_output() to loop back a copy of an IP multicast 1270 * packet to the input queue of a specified interface. Note that this 1271 * calls the output routine of the loopback "driver", but with an interface 1272 * pointer that might NOT be a loopback interface -- evil, but easier than 1273 * replicating that code here. 1274 */ 1275 static void 1276 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst, 1277 int hlen) 1278 { 1279 register struct ip *ip; 1280 struct mbuf *copym; 1281 1282 /* 1283 * Make a deep copy of the packet because we're going to 1284 * modify the pack in order to generate checksums. 1285 */ 1286 copym = m_dup(m, M_NOWAIT); 1287 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1288 copym = m_pullup(copym, hlen); 1289 if (copym != NULL) { 1290 /* If needed, compute the checksum and mark it as valid. */ 1291 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1292 in_delayed_cksum(copym); 1293 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1294 copym->m_pkthdr.csum_flags |= 1295 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1296 copym->m_pkthdr.csum_data = 0xffff; 1297 } 1298 /* 1299 * We don't bother to fragment if the IP length is greater 1300 * than the interface's MTU. Can this possibly matter? 1301 */ 1302 ip = mtod(copym, struct ip *); 1303 ip->ip_sum = 0; 1304 ip->ip_sum = in_cksum(copym, hlen); 1305 #if 1 /* XXX */ 1306 if (dst->sin_family != AF_INET) { 1307 printf("ip_mloopback: bad address family %d\n", 1308 dst->sin_family); 1309 dst->sin_family = AF_INET; 1310 } 1311 #endif 1312 if_simloop(ifp, copym, dst->sin_family, 0); 1313 } 1314 } 1315