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