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_var.h> 59 #include <net/if_llatbl.h> 60 #include <net/netisr.h> 61 #include <net/pfil.h> 62 #include <net/route.h> 63 #include <net/flowtable.h> 64 #ifdef RADIX_MPATH 65 #include <net/radix_mpath.h> 66 #endif 67 #include <net/vnet.h> 68 69 #include <netinet/in.h> 70 #include <netinet/in_kdtrace.h> 71 #include <netinet/in_systm.h> 72 #include <netinet/ip.h> 73 #include <netinet/in_pcb.h> 74 #include <netinet/in_var.h> 75 #include <netinet/ip_var.h> 76 #include <netinet/ip_options.h> 77 #ifdef SCTP 78 #include <netinet/sctp.h> 79 #include <netinet/sctp_crc32.h> 80 #endif 81 82 #ifdef IPSEC 83 #include <netinet/ip_ipsec.h> 84 #include <netipsec/ipsec.h> 85 #endif /* IPSEC*/ 86 87 #include <machine/in_cksum.h> 88 89 #include <security/mac/mac_framework.h> 90 91 VNET_DEFINE(u_short, ip_id); 92 93 #ifdef MBUF_STRESS_TEST 94 static int mbuf_frag_size = 0; 95 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 96 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 97 #endif 98 99 static void ip_mloopback 100 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 101 102 103 extern int in_mcast_loop; 104 extern struct protosw inetsw[]; 105 106 /* 107 * IP output. The packet in mbuf chain m contains a skeletal IP 108 * header (with len, off, ttl, proto, tos, src, dst). 109 * The mbuf chain containing the packet will be freed. 110 * The mbuf opt, if present, will not be freed. 111 * If route ro is present and has ro_rt initialized, route lookup would be 112 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL, 113 * then result of route lookup is stored in ro->ro_rt. 114 * 115 * In the IP forwarding case, the packet will arrive with options already 116 * inserted, so must have a NULL opt pointer. 117 */ 118 int 119 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags, 120 struct ip_moptions *imo, struct inpcb *inp) 121 { 122 struct ip *ip; 123 struct ifnet *ifp = NULL; /* keep compiler happy */ 124 struct mbuf *m0; 125 int hlen = sizeof (struct ip); 126 int mtu; 127 int n; /* scratchpad */ 128 int error = 0; 129 struct sockaddr_in *dst; 130 const struct sockaddr_in *gw; 131 struct in_ifaddr *ia; 132 int isbroadcast; 133 uint16_t ip_len, ip_off; 134 struct route iproute; 135 struct rtentry *rte; /* cache for ro->ro_rt */ 136 struct in_addr odst; 137 struct m_tag *fwd_tag = NULL; 138 #ifdef IPSEC 139 int no_route_but_check_spd = 0; 140 #endif 141 M_ASSERTPKTHDR(m); 142 143 if (inp != NULL) { 144 INP_LOCK_ASSERT(inp); 145 M_SETFIB(m, inp->inp_inc.inc_fibnum); 146 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) { 147 m->m_pkthdr.flowid = inp->inp_flowid; 148 m->m_flags |= M_FLOWID; 149 } 150 } 151 152 if (ro == NULL) { 153 ro = &iproute; 154 bzero(ro, sizeof (*ro)); 155 } 156 157 #ifdef FLOWTABLE 158 if (ro->ro_rt == NULL) { 159 struct flentry *fle; 160 161 /* 162 * The flow table returns route entries valid for up to 30 163 * seconds; we rely on the remainder of ip_output() taking no 164 * longer than that long for the stability of ro_rt. The 165 * flow ID assignment must have happened before this point. 166 */ 167 fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET); 168 if (fle != NULL) 169 flow_to_route(fle, ro); 170 } 171 #endif 172 173 if (opt) { 174 int len = 0; 175 m = ip_insertoptions(m, opt, &len); 176 if (len != 0) 177 hlen = len; /* ip->ip_hl is updated above */ 178 } 179 ip = mtod(m, struct ip *); 180 ip_len = ntohs(ip->ip_len); 181 ip_off = ntohs(ip->ip_off); 182 183 /* 184 * Fill in IP header. If we are not allowing fragmentation, 185 * then the ip_id field is meaningless, but we don't set it 186 * to zero. Doing so causes various problems when devices along 187 * the path (routers, load balancers, firewalls, etc.) illegally 188 * disable DF on our packet. Note that a 16-bit counter 189 * will wrap around in less than 10 seconds at 100 Mbit/s on a 190 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 191 * for Counting NATted Hosts", Proc. IMW'02, available at 192 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>. 193 */ 194 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 195 ip->ip_v = IPVERSION; 196 ip->ip_hl = hlen >> 2; 197 ip->ip_id = ip_newid(); 198 IPSTAT_INC(ips_localout); 199 } else { 200 /* Header already set, fetch hlen from there */ 201 hlen = ip->ip_hl << 2; 202 } 203 204 gw = dst = (struct sockaddr_in *)&ro->ro_dst; 205 again: 206 ia = NULL; 207 /* 208 * If there is a cached route, 209 * check that it is to the same destination 210 * and is still up. If not, free it and try again. 211 * The address family should also be checked in case of sharing the 212 * cache with IPv6. 213 */ 214 rte = ro->ro_rt; 215 if (rte && ((rte->rt_flags & RTF_UP) == 0 || 216 rte->rt_ifp == NULL || 217 !RT_LINK_IS_UP(rte->rt_ifp) || 218 dst->sin_family != AF_INET || 219 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 220 RO_RTFREE(ro); 221 ro->ro_lle = NULL; 222 rte = NULL; 223 } 224 if (rte == NULL && fwd_tag == NULL) { 225 bzero(dst, sizeof(*dst)); 226 dst->sin_family = AF_INET; 227 dst->sin_len = sizeof(*dst); 228 dst->sin_addr = ip->ip_dst; 229 } 230 /* 231 * If routing to interface only, short circuit routing lookup. 232 * The use of an all-ones broadcast address implies this; an 233 * interface is specified by the broadcast address of an interface, 234 * or the destination address of a ptp interface. 235 */ 236 if (flags & IP_SENDONES) { 237 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL && 238 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) { 239 IPSTAT_INC(ips_noroute); 240 error = ENETUNREACH; 241 goto bad; 242 } 243 ip->ip_dst.s_addr = INADDR_BROADCAST; 244 dst->sin_addr = ip->ip_dst; 245 ifp = ia->ia_ifp; 246 ip->ip_ttl = 1; 247 isbroadcast = 1; 248 } else if (flags & IP_ROUTETOIF) { 249 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 250 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) { 251 IPSTAT_INC(ips_noroute); 252 error = ENETUNREACH; 253 goto bad; 254 } 255 ifp = ia->ia_ifp; 256 ip->ip_ttl = 1; 257 isbroadcast = in_broadcast(dst->sin_addr, ifp); 258 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 259 imo != NULL && imo->imo_multicast_ifp != NULL) { 260 /* 261 * Bypass the normal routing lookup for multicast 262 * packets if the interface is specified. 263 */ 264 ifp = imo->imo_multicast_ifp; 265 IFP_TO_IA(ifp, ia); 266 isbroadcast = 0; /* fool gcc */ 267 } else { 268 /* 269 * We want to do any cloning requested by the link layer, 270 * as this is probably required in all cases for correct 271 * operation (as it is for ARP). 272 */ 273 if (rte == NULL) { 274 #ifdef RADIX_MPATH 275 rtalloc_mpath_fib(ro, 276 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), 277 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 278 #else 279 in_rtalloc_ign(ro, 0, 280 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 281 #endif 282 rte = ro->ro_rt; 283 } 284 if (rte == NULL || 285 rte->rt_ifp == NULL || 286 !RT_LINK_IS_UP(rte->rt_ifp)) { 287 #ifdef IPSEC 288 /* 289 * There is no route for this packet, but it is 290 * possible that a matching SPD entry exists. 291 */ 292 no_route_but_check_spd = 1; 293 mtu = 0; /* Silence GCC warning. */ 294 goto sendit; 295 #endif 296 IPSTAT_INC(ips_noroute); 297 error = EHOSTUNREACH; 298 goto bad; 299 } 300 ia = ifatoia(rte->rt_ifa); 301 ifa_ref(&ia->ia_ifa); 302 ifp = rte->rt_ifp; 303 rte->rt_rmx.rmx_pksent++; 304 if (rte->rt_flags & RTF_GATEWAY) 305 gw = (struct sockaddr_in *)rte->rt_gateway; 306 if (rte->rt_flags & RTF_HOST) 307 isbroadcast = (rte->rt_flags & RTF_BROADCAST); 308 else 309 isbroadcast = in_broadcast(gw->sin_addr, ifp); 310 } 311 /* 312 * Calculate MTU. If we have a route that is up, use that, 313 * otherwise use the interface's MTU. 314 */ 315 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) { 316 /* 317 * This case can happen if the user changed the MTU 318 * of an interface after enabling IP on it. Because 319 * most netifs don't keep track of routes pointing to 320 * them, there is no way for one to update all its 321 * routes when the MTU is changed. 322 */ 323 if (rte->rt_rmx.rmx_mtu > ifp->if_mtu) 324 rte->rt_rmx.rmx_mtu = ifp->if_mtu; 325 mtu = rte->rt_rmx.rmx_mtu; 326 } else { 327 mtu = ifp->if_mtu; 328 } 329 /* Catch a possible divide by zero later. */ 330 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p", 331 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp)); 332 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 333 m->m_flags |= M_MCAST; 334 /* 335 * See if the caller provided any multicast options 336 */ 337 if (imo != NULL) { 338 ip->ip_ttl = imo->imo_multicast_ttl; 339 if (imo->imo_multicast_vif != -1) 340 ip->ip_src.s_addr = 341 ip_mcast_src ? 342 ip_mcast_src(imo->imo_multicast_vif) : 343 INADDR_ANY; 344 } else 345 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 346 /* 347 * Confirm that the outgoing interface supports multicast. 348 */ 349 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 350 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 351 IPSTAT_INC(ips_noroute); 352 error = ENETUNREACH; 353 goto bad; 354 } 355 } 356 /* 357 * If source address not specified yet, use address 358 * of outgoing interface. 359 */ 360 if (ip->ip_src.s_addr == INADDR_ANY) { 361 /* Interface may have no addresses. */ 362 if (ia != NULL) 363 ip->ip_src = IA_SIN(ia)->sin_addr; 364 } 365 366 if ((imo == NULL && in_mcast_loop) || 367 (imo && imo->imo_multicast_loop)) { 368 /* 369 * Loop back multicast datagram if not expressly 370 * forbidden to do so, even if we are not a member 371 * of the group; ip_input() will filter it later, 372 * thus deferring a hash lookup and mutex acquisition 373 * at the expense of a cheap copy using m_copym(). 374 */ 375 ip_mloopback(ifp, m, dst, hlen); 376 } else { 377 /* 378 * If we are acting as a multicast router, perform 379 * multicast forwarding as if the packet had just 380 * arrived on the interface to which we are about 381 * to send. The multicast forwarding function 382 * recursively calls this function, using the 383 * IP_FORWARDING flag to prevent infinite recursion. 384 * 385 * Multicasts that are looped back by ip_mloopback(), 386 * above, will be forwarded by the ip_input() routine, 387 * if necessary. 388 */ 389 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) { 390 /* 391 * If rsvp daemon is not running, do not 392 * set ip_moptions. This ensures that the packet 393 * is multicast and not just sent down one link 394 * as prescribed by rsvpd. 395 */ 396 if (!V_rsvp_on) 397 imo = NULL; 398 if (ip_mforward && 399 ip_mforward(ip, ifp, m, imo) != 0) { 400 m_freem(m); 401 goto done; 402 } 403 } 404 } 405 406 /* 407 * Multicasts with a time-to-live of zero may be looped- 408 * back, above, but must not be transmitted on a network. 409 * Also, multicasts addressed to the loopback interface 410 * are not sent -- the above call to ip_mloopback() will 411 * loop back a copy. ip_input() will drop the copy if 412 * this host does not belong to the destination group on 413 * the loopback interface. 414 */ 415 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 416 m_freem(m); 417 goto done; 418 } 419 420 goto sendit; 421 } 422 423 /* 424 * If the source address is not specified yet, use the address 425 * of the outoing interface. 426 */ 427 if (ip->ip_src.s_addr == INADDR_ANY) { 428 /* Interface may have no addresses. */ 429 if (ia != NULL) { 430 ip->ip_src = IA_SIN(ia)->sin_addr; 431 } 432 } 433 434 /* 435 * Verify that we have any chance at all of being able to queue the 436 * packet or packet fragments, unless ALTQ is enabled on the given 437 * interface in which case packetdrop should be done by queueing. 438 */ 439 n = ip_len / mtu + 1; /* how many fragments ? */ 440 if ( 441 #ifdef ALTQ 442 (!ALTQ_IS_ENABLED(&ifp->if_snd)) && 443 #endif /* ALTQ */ 444 (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) { 445 error = ENOBUFS; 446 IPSTAT_INC(ips_odropped); 447 ifp->if_snd.ifq_drops += n; 448 goto bad; 449 } 450 451 /* 452 * Look for broadcast address and 453 * verify user is allowed to send 454 * such a packet. 455 */ 456 if (isbroadcast) { 457 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 458 error = EADDRNOTAVAIL; 459 goto bad; 460 } 461 if ((flags & IP_ALLOWBROADCAST) == 0) { 462 error = EACCES; 463 goto bad; 464 } 465 /* don't allow broadcast messages to be fragmented */ 466 if (ip_len > mtu) { 467 error = EMSGSIZE; 468 goto bad; 469 } 470 m->m_flags |= M_BCAST; 471 } else { 472 m->m_flags &= ~M_BCAST; 473 } 474 475 sendit: 476 #ifdef IPSEC 477 switch(ip_ipsec_output(&m, inp, &flags, &error)) { 478 case 1: 479 goto bad; 480 case -1: 481 goto done; 482 case 0: 483 default: 484 break; /* Continue with packet processing. */ 485 } 486 /* 487 * Check if there was a route for this packet; return error if not. 488 */ 489 if (no_route_but_check_spd) { 490 IPSTAT_INC(ips_noroute); 491 error = EHOSTUNREACH; 492 goto bad; 493 } 494 /* Update variables that are affected by ipsec4_output(). */ 495 ip = mtod(m, struct ip *); 496 hlen = ip->ip_hl << 2; 497 #endif /* IPSEC */ 498 499 /* Jump over all PFIL processing if hooks are not active. */ 500 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 501 goto passout; 502 503 /* Run through list of hooks for output packets. */ 504 odst.s_addr = ip->ip_dst.s_addr; 505 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 506 if (error != 0 || m == NULL) 507 goto done; 508 509 ip = mtod(m, struct ip *); 510 511 /* See if destination IP address was changed by packet filter. */ 512 if (odst.s_addr != ip->ip_dst.s_addr) { 513 m->m_flags |= M_SKIP_FIREWALL; 514 /* If destination is now ourself drop to ip_input(). */ 515 if (in_localip(ip->ip_dst)) { 516 m->m_flags |= M_FASTFWD_OURS; 517 if (m->m_pkthdr.rcvif == NULL) 518 m->m_pkthdr.rcvif = V_loif; 519 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 520 m->m_pkthdr.csum_flags |= 521 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 522 m->m_pkthdr.csum_data = 0xffff; 523 } 524 m->m_pkthdr.csum_flags |= 525 CSUM_IP_CHECKED | CSUM_IP_VALID; 526 #ifdef SCTP 527 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 528 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 529 #endif 530 error = netisr_queue(NETISR_IP, m); 531 goto done; 532 } else { 533 if (ia != NULL) 534 ifa_free(&ia->ia_ifa); 535 goto again; /* Redo the routing table lookup. */ 536 } 537 } 538 539 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ 540 if (m->m_flags & M_FASTFWD_OURS) { 541 if (m->m_pkthdr.rcvif == NULL) 542 m->m_pkthdr.rcvif = V_loif; 543 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 544 m->m_pkthdr.csum_flags |= 545 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 546 m->m_pkthdr.csum_data = 0xffff; 547 } 548 #ifdef SCTP 549 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 550 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 551 #endif 552 m->m_pkthdr.csum_flags |= 553 CSUM_IP_CHECKED | CSUM_IP_VALID; 554 555 error = netisr_queue(NETISR_IP, m); 556 goto done; 557 } 558 /* Or forward to some other address? */ 559 if ((m->m_flags & M_IP_NEXTHOP) && 560 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 561 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 562 m->m_flags |= M_SKIP_FIREWALL; 563 m->m_flags &= ~M_IP_NEXTHOP; 564 m_tag_delete(m, fwd_tag); 565 if (ia != NULL) 566 ifa_free(&ia->ia_ifa); 567 goto again; 568 } 569 570 passout: 571 /* 127/8 must not appear on wire - RFC1122. */ 572 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 573 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 574 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 575 IPSTAT_INC(ips_badaddr); 576 error = EADDRNOTAVAIL; 577 goto bad; 578 } 579 } 580 581 m->m_pkthdr.csum_flags |= CSUM_IP; 582 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) { 583 in_delayed_cksum(m); 584 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 585 } 586 #ifdef SCTP 587 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) { 588 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); 589 m->m_pkthdr.csum_flags &= ~CSUM_SCTP; 590 } 591 #endif 592 593 /* 594 * If small enough for interface, or the interface will take 595 * care of the fragmentation for us, we can just send directly. 596 */ 597 if (ip_len <= mtu || 598 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 || 599 ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) { 600 ip->ip_sum = 0; 601 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) { 602 ip->ip_sum = in_cksum(m, hlen); 603 m->m_pkthdr.csum_flags &= ~CSUM_IP; 604 } 605 606 /* 607 * Record statistics for this interface address. 608 * With CSUM_TSO the byte/packet count will be slightly 609 * incorrect because we count the IP+TCP headers only 610 * once instead of for every generated packet. 611 */ 612 if (!(flags & IP_FORWARDING) && ia) { 613 if (m->m_pkthdr.csum_flags & CSUM_TSO) 614 counter_u64_add(ia->ia_ifa.ifa_opackets, 615 m->m_pkthdr.len / m->m_pkthdr.tso_segsz); 616 else 617 counter_u64_add(ia->ia_ifa.ifa_opackets, 1); 618 619 counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len); 620 } 621 #ifdef MBUF_STRESS_TEST 622 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 623 m = m_fragment(m, M_NOWAIT, mbuf_frag_size); 624 #endif 625 /* 626 * Reset layer specific mbuf flags 627 * to avoid confusing lower layers. 628 */ 629 m_clrprotoflags(m); 630 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); 631 error = (*ifp->if_output)(ifp, m, 632 (const struct sockaddr *)gw, ro); 633 goto done; 634 } 635 636 /* Balk when DF bit is set or the interface didn't support TSO. */ 637 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { 638 error = EMSGSIZE; 639 IPSTAT_INC(ips_cantfrag); 640 goto bad; 641 } 642 643 /* 644 * Too large for interface; fragment if possible. If successful, 645 * on return, m will point to a list of packets to be sent. 646 */ 647 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist); 648 if (error) 649 goto bad; 650 for (; m; m = m0) { 651 m0 = m->m_nextpkt; 652 m->m_nextpkt = 0; 653 if (error == 0) { 654 /* Record statistics for this interface address. */ 655 if (ia != NULL) { 656 counter_u64_add(ia->ia_ifa.ifa_opackets, 1); 657 counter_u64_add(ia->ia_ifa.ifa_obytes, 658 m->m_pkthdr.len); 659 } 660 /* 661 * Reset layer specific mbuf flags 662 * to avoid confusing upper layers. 663 */ 664 m_clrprotoflags(m); 665 666 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); 667 error = (*ifp->if_output)(ifp, m, 668 (const struct sockaddr *)gw, ro); 669 } else 670 m_freem(m); 671 } 672 673 if (error == 0) 674 IPSTAT_INC(ips_fragmented); 675 676 done: 677 if (ro == &iproute) 678 RO_RTFREE(ro); 679 if (ia != NULL) 680 ifa_free(&ia->ia_ifa); 681 return (error); 682 bad: 683 m_freem(m); 684 goto done; 685 } 686 687 /* 688 * Create a chain of fragments which fit the given mtu. m_frag points to the 689 * mbuf to be fragmented; on return it points to the chain with the fragments. 690 * Return 0 if no error. If error, m_frag may contain a partially built 691 * chain of fragments that should be freed by the caller. 692 * 693 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 694 */ 695 int 696 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 697 u_long if_hwassist_flags) 698 { 699 int error = 0; 700 int hlen = ip->ip_hl << 2; 701 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 702 int off; 703 struct mbuf *m0 = *m_frag; /* the original packet */ 704 int firstlen; 705 struct mbuf **mnext; 706 int nfrags; 707 uint16_t ip_len, ip_off; 708 709 ip_len = ntohs(ip->ip_len); 710 ip_off = ntohs(ip->ip_off); 711 712 if (ip_off & IP_DF) { /* Fragmentation not allowed */ 713 IPSTAT_INC(ips_cantfrag); 714 return EMSGSIZE; 715 } 716 717 /* 718 * Must be able to put at least 8 bytes per fragment. 719 */ 720 if (len < 8) 721 return EMSGSIZE; 722 723 /* 724 * If the interface will not calculate checksums on 725 * fragmented packets, then do it here. 726 */ 727 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 728 in_delayed_cksum(m0); 729 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 730 } 731 #ifdef SCTP 732 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) { 733 sctp_delayed_cksum(m0, hlen); 734 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP; 735 } 736 #endif 737 if (len > PAGE_SIZE) { 738 /* 739 * Fragment large datagrams such that each segment 740 * contains a multiple of PAGE_SIZE amount of data, 741 * plus headers. This enables a receiver to perform 742 * page-flipping zero-copy optimizations. 743 * 744 * XXX When does this help given that sender and receiver 745 * could have different page sizes, and also mtu could 746 * be less than the receiver's page size ? 747 */ 748 int newlen; 749 struct mbuf *m; 750 751 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 752 off += m->m_len; 753 754 /* 755 * firstlen (off - hlen) must be aligned on an 756 * 8-byte boundary 757 */ 758 if (off < hlen) 759 goto smart_frag_failure; 760 off = ((off - hlen) & ~7) + hlen; 761 newlen = (~PAGE_MASK) & mtu; 762 if ((newlen + sizeof (struct ip)) > mtu) { 763 /* we failed, go back the default */ 764 smart_frag_failure: 765 newlen = len; 766 off = hlen + len; 767 } 768 len = newlen; 769 770 } else { 771 off = hlen + len; 772 } 773 774 firstlen = off - hlen; 775 mnext = &m0->m_nextpkt; /* pointer to next packet */ 776 777 /* 778 * Loop through length of segment after first fragment, 779 * make new header and copy data of each part and link onto chain. 780 * Here, m0 is the original packet, m is the fragment being created. 781 * The fragments are linked off the m_nextpkt of the original 782 * packet, which after processing serves as the first fragment. 783 */ 784 for (nfrags = 1; off < ip_len; off += len, nfrags++) { 785 struct ip *mhip; /* ip header on the fragment */ 786 struct mbuf *m; 787 int mhlen = sizeof (struct ip); 788 789 m = m_gethdr(M_NOWAIT, MT_DATA); 790 if (m == NULL) { 791 error = ENOBUFS; 792 IPSTAT_INC(ips_odropped); 793 goto done; 794 } 795 m->m_flags |= (m0->m_flags & M_MCAST); 796 /* 797 * In the first mbuf, leave room for the link header, then 798 * copy the original IP header including options. The payload 799 * goes into an additional mbuf chain returned by m_copym(). 800 */ 801 m->m_data += max_linkhdr; 802 mhip = mtod(m, struct ip *); 803 *mhip = *ip; 804 if (hlen > sizeof (struct ip)) { 805 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 806 mhip->ip_v = IPVERSION; 807 mhip->ip_hl = mhlen >> 2; 808 } 809 m->m_len = mhlen; 810 /* XXX do we need to add ip_off below ? */ 811 mhip->ip_off = ((off - hlen) >> 3) + ip_off; 812 if (off + len >= ip_len) 813 len = ip_len - off; 814 else 815 mhip->ip_off |= IP_MF; 816 mhip->ip_len = htons((u_short)(len + mhlen)); 817 m->m_next = m_copym(m0, off, len, M_NOWAIT); 818 if (m->m_next == NULL) { /* copy failed */ 819 m_free(m); 820 error = ENOBUFS; /* ??? */ 821 IPSTAT_INC(ips_odropped); 822 goto done; 823 } 824 m->m_pkthdr.len = mhlen + len; 825 m->m_pkthdr.rcvif = NULL; 826 #ifdef MAC 827 mac_netinet_fragment(m0, m); 828 #endif 829 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 830 mhip->ip_off = htons(mhip->ip_off); 831 mhip->ip_sum = 0; 832 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { 833 mhip->ip_sum = in_cksum(m, mhlen); 834 m->m_pkthdr.csum_flags &= ~CSUM_IP; 835 } 836 *mnext = m; 837 mnext = &m->m_nextpkt; 838 } 839 IPSTAT_ADD(ips_ofragments, nfrags); 840 841 /* 842 * Update first fragment by trimming what's been copied out 843 * and updating header. 844 */ 845 m_adj(m0, hlen + firstlen - ip_len); 846 m0->m_pkthdr.len = hlen + firstlen; 847 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 848 ip->ip_off = htons(ip_off | IP_MF); 849 ip->ip_sum = 0; 850 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { 851 ip->ip_sum = in_cksum(m0, hlen); 852 m0->m_pkthdr.csum_flags &= ~CSUM_IP; 853 } 854 855 done: 856 *m_frag = m0; 857 return error; 858 } 859 860 void 861 in_delayed_cksum(struct mbuf *m) 862 { 863 struct ip *ip; 864 uint16_t csum, offset, ip_len; 865 866 ip = mtod(m, struct ip *); 867 offset = ip->ip_hl << 2 ; 868 ip_len = ntohs(ip->ip_len); 869 csum = in_cksum_skip(m, ip_len, offset); 870 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 871 csum = 0xffff; 872 offset += m->m_pkthdr.csum_data; /* checksum offset */ 873 874 if (offset + sizeof(u_short) > m->m_len) { 875 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 876 m->m_len, offset, ip->ip_p); 877 /* 878 * XXX 879 * this shouldn't happen, but if it does, the 880 * correct behavior may be to insert the checksum 881 * in the appropriate next mbuf in the chain. 882 */ 883 return; 884 } 885 *(u_short *)(m->m_data + offset) = csum; 886 } 887 888 /* 889 * IP socket option processing. 890 */ 891 int 892 ip_ctloutput(struct socket *so, struct sockopt *sopt) 893 { 894 struct inpcb *inp = sotoinpcb(so); 895 int error, optval; 896 897 error = optval = 0; 898 if (sopt->sopt_level != IPPROTO_IP) { 899 error = EINVAL; 900 901 if (sopt->sopt_level == SOL_SOCKET && 902 sopt->sopt_dir == SOPT_SET) { 903 switch (sopt->sopt_name) { 904 case SO_REUSEADDR: 905 INP_WLOCK(inp); 906 if ((so->so_options & SO_REUSEADDR) != 0) 907 inp->inp_flags2 |= INP_REUSEADDR; 908 else 909 inp->inp_flags2 &= ~INP_REUSEADDR; 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