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_mbuf_stress_test.h" 38 #include "opt_mpath.h" 39 #include "opt_route.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/sdt.h> 51 #include <sys/socket.h> 52 #include <sys/socketvar.h> 53 #include <sys/sysctl.h> 54 #include <sys/ucred.h> 55 56 #include <net/if.h> 57 #include <net/if_var.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 counter_u64_add(ia->ia_ifa.ifa_opackets, 614 m->m_pkthdr.len / m->m_pkthdr.tso_segsz); 615 else 616 counter_u64_add(ia->ia_ifa.ifa_opackets, 1); 617 618 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len); 619 } 620 #ifdef MBUF_STRESS_TEST 621 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 622 m = m_fragment(m, M_NOWAIT, mbuf_frag_size); 623 #endif 624 /* 625 * Reset layer specific mbuf flags 626 * to avoid confusing lower layers. 627 */ 628 m_clrprotoflags(m); 629 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); 630 error = (*ifp->if_output)(ifp, m, 631 (const struct sockaddr *)gw, ro); 632 goto done; 633 } 634 635 /* Balk when DF bit is set or the interface didn't support TSO. */ 636 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { 637 error = EMSGSIZE; 638 IPSTAT_INC(ips_cantfrag); 639 goto bad; 640 } 641 642 /* 643 * Too large for interface; fragment if possible. If successful, 644 * on return, m will point to a list of packets to be sent. 645 */ 646 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist); 647 if (error) 648 goto bad; 649 for (; m; m = m0) { 650 m0 = m->m_nextpkt; 651 m->m_nextpkt = 0; 652 if (error == 0) { 653 /* Record statistics for this interface address. */ 654 if (ia != NULL) { 655 counter_u64_add(ia->ia_ifa.ifa_opackets, 1); 656 counter_u64_add(ia->ia_ifa.ifa_obytes, 657 m->m_pkthdr.len); 658 } 659 /* 660 * Reset layer specific mbuf flags 661 * to avoid confusing upper layers. 662 */ 663 m_clrprotoflags(m); 664 665 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); 666 error = (*ifp->if_output)(ifp, m, 667 (const struct sockaddr *)gw, ro); 668 } else 669 m_freem(m); 670 } 671 672 if (error == 0) 673 IPSTAT_INC(ips_fragmented); 674 675 done: 676 if (ro == &iproute) 677 RO_RTFREE(ro); 678 if (ia != NULL) 679 ifa_free(&ia->ia_ifa); 680 return (error); 681 bad: 682 m_freem(m); 683 goto done; 684 } 685 686 /* 687 * Create a chain of fragments which fit the given mtu. m_frag points to the 688 * mbuf to be fragmented; on return it points to the chain with the fragments. 689 * Return 0 if no error. If error, m_frag may contain a partially built 690 * chain of fragments that should be freed by the caller. 691 * 692 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 693 */ 694 int 695 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 696 u_long if_hwassist_flags) 697 { 698 int error = 0; 699 int hlen = ip->ip_hl << 2; 700 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 701 int off; 702 struct mbuf *m0 = *m_frag; /* the original packet */ 703 int firstlen; 704 struct mbuf **mnext; 705 int nfrags; 706 uint16_t ip_len, ip_off; 707 708 ip_len = ntohs(ip->ip_len); 709 ip_off = ntohs(ip->ip_off); 710 711 if (ip_off & IP_DF) { /* Fragmentation not allowed */ 712 IPSTAT_INC(ips_cantfrag); 713 return EMSGSIZE; 714 } 715 716 /* 717 * Must be able to put at least 8 bytes per fragment. 718 */ 719 if (len < 8) 720 return EMSGSIZE; 721 722 /* 723 * If the interface will not calculate checksums on 724 * fragmented packets, then do it here. 725 */ 726 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 727 in_delayed_cksum(m0); 728 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 729 } 730 #ifdef SCTP 731 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) { 732 sctp_delayed_cksum(m0, hlen); 733 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP; 734 } 735 #endif 736 if (len > PAGE_SIZE) { 737 /* 738 * Fragment large datagrams such that each segment 739 * contains a multiple of PAGE_SIZE amount of data, 740 * plus headers. This enables a receiver to perform 741 * page-flipping zero-copy optimizations. 742 * 743 * XXX When does this help given that sender and receiver 744 * could have different page sizes, and also mtu could 745 * be less than the receiver's page size ? 746 */ 747 int newlen; 748 struct mbuf *m; 749 750 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 751 off += m->m_len; 752 753 /* 754 * firstlen (off - hlen) must be aligned on an 755 * 8-byte boundary 756 */ 757 if (off < hlen) 758 goto smart_frag_failure; 759 off = ((off - hlen) & ~7) + hlen; 760 newlen = (~PAGE_MASK) & mtu; 761 if ((newlen + sizeof (struct ip)) > mtu) { 762 /* we failed, go back the default */ 763 smart_frag_failure: 764 newlen = len; 765 off = hlen + len; 766 } 767 len = newlen; 768 769 } else { 770 off = hlen + len; 771 } 772 773 firstlen = off - hlen; 774 mnext = &m0->m_nextpkt; /* pointer to next packet */ 775 776 /* 777 * Loop through length of segment after first fragment, 778 * make new header and copy data of each part and link onto chain. 779 * Here, m0 is the original packet, m is the fragment being created. 780 * The fragments are linked off the m_nextpkt of the original 781 * packet, which after processing serves as the first fragment. 782 */ 783 for (nfrags = 1; off < ip_len; off += len, nfrags++) { 784 struct ip *mhip; /* ip header on the fragment */ 785 struct mbuf *m; 786 int mhlen = sizeof (struct ip); 787 788 m = m_gethdr(M_NOWAIT, MT_DATA); 789 if (m == NULL) { 790 error = ENOBUFS; 791 IPSTAT_INC(ips_odropped); 792 goto done; 793 } 794 m->m_flags |= (m0->m_flags & M_MCAST); 795 /* 796 * In the first mbuf, leave room for the link header, then 797 * copy the original IP header including options. The payload 798 * goes into an additional mbuf chain returned by m_copym(). 799 */ 800 m->m_data += max_linkhdr; 801 mhip = mtod(m, struct ip *); 802 *mhip = *ip; 803 if (hlen > sizeof (struct ip)) { 804 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 805 mhip->ip_v = IPVERSION; 806 mhip->ip_hl = mhlen >> 2; 807 } 808 m->m_len = mhlen; 809 /* XXX do we need to add ip_off below ? */ 810 mhip->ip_off = ((off - hlen) >> 3) + ip_off; 811 if (off + len >= ip_len) 812 len = ip_len - off; 813 else 814 mhip->ip_off |= IP_MF; 815 mhip->ip_len = htons((u_short)(len + mhlen)); 816 m->m_next = m_copym(m0, off, len, M_NOWAIT); 817 if (m->m_next == NULL) { /* copy failed */ 818 m_free(m); 819 error = ENOBUFS; /* ??? */ 820 IPSTAT_INC(ips_odropped); 821 goto done; 822 } 823 m->m_pkthdr.len = mhlen + len; 824 m->m_pkthdr.rcvif = NULL; 825 #ifdef MAC 826 mac_netinet_fragment(m0, m); 827 #endif 828 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 829 mhip->ip_off = htons(mhip->ip_off); 830 mhip->ip_sum = 0; 831 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { 832 mhip->ip_sum = in_cksum(m, mhlen); 833 m->m_pkthdr.csum_flags &= ~CSUM_IP; 834 } 835 *mnext = m; 836 mnext = &m->m_nextpkt; 837 } 838 IPSTAT_ADD(ips_ofragments, nfrags); 839 840 /* 841 * Update first fragment by trimming what's been copied out 842 * and updating header. 843 */ 844 m_adj(m0, hlen + firstlen - ip_len); 845 m0->m_pkthdr.len = hlen + firstlen; 846 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 847 ip->ip_off = htons(ip_off | IP_MF); 848 ip->ip_sum = 0; 849 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { 850 ip->ip_sum = in_cksum(m0, hlen); 851 m0->m_pkthdr.csum_flags &= ~CSUM_IP; 852 } 853 854 done: 855 *m_frag = m0; 856 return error; 857 } 858 859 void 860 in_delayed_cksum(struct mbuf *m) 861 { 862 struct ip *ip; 863 uint16_t csum, offset, ip_len; 864 865 ip = mtod(m, struct ip *); 866 offset = ip->ip_hl << 2 ; 867 ip_len = ntohs(ip->ip_len); 868 csum = in_cksum_skip(m, ip_len, offset); 869 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 870 csum = 0xffff; 871 offset += m->m_pkthdr.csum_data; /* checksum offset */ 872 873 if (offset + sizeof(u_short) > m->m_len) { 874 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 875 m->m_len, offset, ip->ip_p); 876 /* 877 * XXX 878 * this shouldn't happen, but if it does, the 879 * correct behavior may be to insert the checksum 880 * in the appropriate next mbuf in the chain. 881 */ 882 return; 883 } 884 *(u_short *)(m->m_data + offset) = csum; 885 } 886 887 /* 888 * IP socket option processing. 889 */ 890 int 891 ip_ctloutput(struct socket *so, struct sockopt *sopt) 892 { 893 struct inpcb *inp = sotoinpcb(so); 894 int error, optval; 895 896 error = optval = 0; 897 if (sopt->sopt_level != IPPROTO_IP) { 898 error = EINVAL; 899 900 if (sopt->sopt_level == SOL_SOCKET && 901 sopt->sopt_dir == SOPT_SET) { 902 switch (sopt->sopt_name) { 903 case SO_REUSEADDR: 904 INP_WLOCK(inp); 905 if ((so->so_options & SO_REUSEADDR) != 0) 906 inp->inp_flags2 |= INP_REUSEADDR; 907 else 908 inp->inp_flags2 &= ~INP_REUSEADDR; 909 INP_WUNLOCK(inp); 910 error = 0; 911 break; 912 case SO_REUSEPORT: 913 INP_WLOCK(inp); 914 if ((so->so_options & SO_REUSEPORT) != 0) 915 inp->inp_flags2 |= INP_REUSEPORT; 916 else 917 inp->inp_flags2 &= ~INP_REUSEPORT; 918 INP_WUNLOCK(inp); 919 error = 0; 920 break; 921 case SO_SETFIB: 922 INP_WLOCK(inp); 923 inp->inp_inc.inc_fibnum = so->so_fibnum; 924 INP_WUNLOCK(inp); 925 error = 0; 926 break; 927 default: 928 break; 929 } 930 } 931 return (error); 932 } 933 934 switch (sopt->sopt_dir) { 935 case SOPT_SET: 936 switch (sopt->sopt_name) { 937 case IP_OPTIONS: 938 #ifdef notyet 939 case IP_RETOPTS: 940 #endif 941 { 942 struct mbuf *m; 943 if (sopt->sopt_valsize > MLEN) { 944 error = EMSGSIZE; 945 break; 946 } 947 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA); 948 if (m == NULL) { 949 error = ENOBUFS; 950 break; 951 } 952 m->m_len = sopt->sopt_valsize; 953 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 954 m->m_len); 955 if (error) { 956 m_free(m); 957 break; 958 } 959 INP_WLOCK(inp); 960 error = ip_pcbopts(inp, sopt->sopt_name, m); 961 INP_WUNLOCK(inp); 962 return (error); 963 } 964 965 case IP_BINDANY: 966 if (sopt->sopt_td != NULL) { 967 error = priv_check(sopt->sopt_td, 968 PRIV_NETINET_BINDANY); 969 if (error) 970 break; 971 } 972 /* FALLTHROUGH */ 973 case IP_TOS: 974 case IP_TTL: 975 case IP_MINTTL: 976 case IP_RECVOPTS: 977 case IP_RECVRETOPTS: 978 case IP_RECVDSTADDR: 979 case IP_RECVTTL: 980 case IP_RECVIF: 981 case IP_FAITH: 982 case IP_ONESBCAST: 983 case IP_DONTFRAG: 984 case IP_RECVTOS: 985 error = sooptcopyin(sopt, &optval, sizeof optval, 986 sizeof optval); 987 if (error) 988 break; 989 990 switch (sopt->sopt_name) { 991 case IP_TOS: 992 inp->inp_ip_tos = optval; 993 break; 994 995 case IP_TTL: 996 inp->inp_ip_ttl = optval; 997 break; 998 999 case IP_MINTTL: 1000 if (optval >= 0 && optval <= MAXTTL) 1001 inp->inp_ip_minttl = optval; 1002 else 1003 error = EINVAL; 1004 break; 1005 1006 #define OPTSET(bit) do { \ 1007 INP_WLOCK(inp); \ 1008 if (optval) \ 1009 inp->inp_flags |= bit; \ 1010 else \ 1011 inp->inp_flags &= ~bit; \ 1012 INP_WUNLOCK(inp); \ 1013 } while (0) 1014 1015 case IP_RECVOPTS: 1016 OPTSET(INP_RECVOPTS); 1017 break; 1018 1019 case IP_RECVRETOPTS: 1020 OPTSET(INP_RECVRETOPTS); 1021 break; 1022 1023 case IP_RECVDSTADDR: 1024 OPTSET(INP_RECVDSTADDR); 1025 break; 1026 1027 case IP_RECVTTL: 1028 OPTSET(INP_RECVTTL); 1029 break; 1030 1031 case IP_RECVIF: 1032 OPTSET(INP_RECVIF); 1033 break; 1034 1035 case IP_FAITH: 1036 OPTSET(INP_FAITH); 1037 break; 1038 1039 case IP_ONESBCAST: 1040 OPTSET(INP_ONESBCAST); 1041 break; 1042 case IP_DONTFRAG: 1043 OPTSET(INP_DONTFRAG); 1044 break; 1045 case IP_BINDANY: 1046 OPTSET(INP_BINDANY); 1047 break; 1048 case IP_RECVTOS: 1049 OPTSET(INP_RECVTOS); 1050 break; 1051 } 1052 break; 1053 #undef OPTSET 1054 1055 /* 1056 * Multicast socket options are processed by the in_mcast 1057 * module. 1058 */ 1059 case IP_MULTICAST_IF: 1060 case IP_MULTICAST_VIF: 1061 case IP_MULTICAST_TTL: 1062 case IP_MULTICAST_LOOP: 1063 case IP_ADD_MEMBERSHIP: 1064 case IP_DROP_MEMBERSHIP: 1065 case IP_ADD_SOURCE_MEMBERSHIP: 1066 case IP_DROP_SOURCE_MEMBERSHIP: 1067 case IP_BLOCK_SOURCE: 1068 case IP_UNBLOCK_SOURCE: 1069 case IP_MSFILTER: 1070 case MCAST_JOIN_GROUP: 1071 case MCAST_LEAVE_GROUP: 1072 case MCAST_JOIN_SOURCE_GROUP: 1073 case MCAST_LEAVE_SOURCE_GROUP: 1074 case MCAST_BLOCK_SOURCE: 1075 case MCAST_UNBLOCK_SOURCE: 1076 error = inp_setmoptions(inp, sopt); 1077 break; 1078 1079 case IP_PORTRANGE: 1080 error = sooptcopyin(sopt, &optval, sizeof optval, 1081 sizeof optval); 1082 if (error) 1083 break; 1084 1085 INP_WLOCK(inp); 1086 switch (optval) { 1087 case IP_PORTRANGE_DEFAULT: 1088 inp->inp_flags &= ~(INP_LOWPORT); 1089 inp->inp_flags &= ~(INP_HIGHPORT); 1090 break; 1091 1092 case IP_PORTRANGE_HIGH: 1093 inp->inp_flags &= ~(INP_LOWPORT); 1094 inp->inp_flags |= INP_HIGHPORT; 1095 break; 1096 1097 case IP_PORTRANGE_LOW: 1098 inp->inp_flags &= ~(INP_HIGHPORT); 1099 inp->inp_flags |= INP_LOWPORT; 1100 break; 1101 1102 default: 1103 error = EINVAL; 1104 break; 1105 } 1106 INP_WUNLOCK(inp); 1107 break; 1108 1109 #ifdef IPSEC 1110 case IP_IPSEC_POLICY: 1111 { 1112 caddr_t req; 1113 struct mbuf *m; 1114 1115 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1116 break; 1117 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1118 break; 1119 req = mtod(m, caddr_t); 1120 error = ipsec_set_policy(inp, sopt->sopt_name, req, 1121 m->m_len, (sopt->sopt_td != NULL) ? 1122 sopt->sopt_td->td_ucred : NULL); 1123 m_freem(m); 1124 break; 1125 } 1126 #endif /* IPSEC */ 1127 1128 default: 1129 error = ENOPROTOOPT; 1130 break; 1131 } 1132 break; 1133 1134 case SOPT_GET: 1135 switch (sopt->sopt_name) { 1136 case IP_OPTIONS: 1137 case IP_RETOPTS: 1138 if (inp->inp_options) 1139 error = sooptcopyout(sopt, 1140 mtod(inp->inp_options, 1141 char *), 1142 inp->inp_options->m_len); 1143 else 1144 sopt->sopt_valsize = 0; 1145 break; 1146 1147 case IP_TOS: 1148 case IP_TTL: 1149 case IP_MINTTL: 1150 case IP_RECVOPTS: 1151 case IP_RECVRETOPTS: 1152 case IP_RECVDSTADDR: 1153 case IP_RECVTTL: 1154 case IP_RECVIF: 1155 case IP_PORTRANGE: 1156 case IP_FAITH: 1157 case IP_ONESBCAST: 1158 case IP_DONTFRAG: 1159 case IP_BINDANY: 1160 case IP_RECVTOS: 1161 switch (sopt->sopt_name) { 1162 1163 case IP_TOS: 1164 optval = inp->inp_ip_tos; 1165 break; 1166 1167 case IP_TTL: 1168 optval = inp->inp_ip_ttl; 1169 break; 1170 1171 case IP_MINTTL: 1172 optval = inp->inp_ip_minttl; 1173 break; 1174 1175 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1176 1177 case IP_RECVOPTS: 1178 optval = OPTBIT(INP_RECVOPTS); 1179 break; 1180 1181 case IP_RECVRETOPTS: 1182 optval = OPTBIT(INP_RECVRETOPTS); 1183 break; 1184 1185 case IP_RECVDSTADDR: 1186 optval = OPTBIT(INP_RECVDSTADDR); 1187 break; 1188 1189 case IP_RECVTTL: 1190 optval = OPTBIT(INP_RECVTTL); 1191 break; 1192 1193 case IP_RECVIF: 1194 optval = OPTBIT(INP_RECVIF); 1195 break; 1196 1197 case IP_PORTRANGE: 1198 if (inp->inp_flags & INP_HIGHPORT) 1199 optval = IP_PORTRANGE_HIGH; 1200 else if (inp->inp_flags & INP_LOWPORT) 1201 optval = IP_PORTRANGE_LOW; 1202 else 1203 optval = 0; 1204 break; 1205 1206 case IP_FAITH: 1207 optval = OPTBIT(INP_FAITH); 1208 break; 1209 1210 case IP_ONESBCAST: 1211 optval = OPTBIT(INP_ONESBCAST); 1212 break; 1213 case IP_DONTFRAG: 1214 optval = OPTBIT(INP_DONTFRAG); 1215 break; 1216 case IP_BINDANY: 1217 optval = OPTBIT(INP_BINDANY); 1218 break; 1219 case IP_RECVTOS: 1220 optval = OPTBIT(INP_RECVTOS); 1221 break; 1222 } 1223 error = sooptcopyout(sopt, &optval, sizeof optval); 1224 break; 1225 1226 /* 1227 * Multicast socket options are processed by the in_mcast 1228 * module. 1229 */ 1230 case IP_MULTICAST_IF: 1231 case IP_MULTICAST_VIF: 1232 case IP_MULTICAST_TTL: 1233 case IP_MULTICAST_LOOP: 1234 case IP_MSFILTER: 1235 error = inp_getmoptions(inp, sopt); 1236 break; 1237 1238 #ifdef IPSEC 1239 case IP_IPSEC_POLICY: 1240 { 1241 struct mbuf *m = NULL; 1242 caddr_t req = NULL; 1243 size_t len = 0; 1244 1245 if (m != 0) { 1246 req = mtod(m, caddr_t); 1247 len = m->m_len; 1248 } 1249 error = ipsec_get_policy(sotoinpcb(so), req, len, &m); 1250 if (error == 0) 1251 error = soopt_mcopyout(sopt, m); /* XXX */ 1252 if (error == 0) 1253 m_freem(m); 1254 break; 1255 } 1256 #endif /* IPSEC */ 1257 1258 default: 1259 error = ENOPROTOOPT; 1260 break; 1261 } 1262 break; 1263 } 1264 return (error); 1265 } 1266 1267 /* 1268 * Routine called from ip_output() to loop back a copy of an IP multicast 1269 * packet to the input queue of a specified interface. Note that this 1270 * calls the output routine of the loopback "driver", but with an interface 1271 * pointer that might NOT be a loopback interface -- evil, but easier than 1272 * replicating that code here. 1273 */ 1274 static void 1275 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst, 1276 int hlen) 1277 { 1278 register struct ip *ip; 1279 struct mbuf *copym; 1280 1281 /* 1282 * Make a deep copy of the packet because we're going to 1283 * modify the pack in order to generate checksums. 1284 */ 1285 copym = m_dup(m, M_NOWAIT); 1286 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1287 copym = m_pullup(copym, hlen); 1288 if (copym != NULL) { 1289 /* If needed, compute the checksum and mark it as valid. */ 1290 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1291 in_delayed_cksum(copym); 1292 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1293 copym->m_pkthdr.csum_flags |= 1294 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1295 copym->m_pkthdr.csum_data = 0xffff; 1296 } 1297 /* 1298 * We don't bother to fragment if the IP length is greater 1299 * than the interface's MTU. Can this possibly matter? 1300 */ 1301 ip = mtod(copym, struct ip *); 1302 ip->ip_sum = 0; 1303 ip->ip_sum = in_cksum(copym, hlen); 1304 #if 1 /* XXX */ 1305 if (dst->sin_family != AF_INET) { 1306 printf("ip_mloopback: bad address family %d\n", 1307 dst->sin_family); 1308 dst->sin_family = AF_INET; 1309 } 1310 #endif 1311 if_simloop(ifp, copym, dst->sin_family, 0); 1312 } 1313 } 1314