1 /*- 2 * Copyright (c) 2003 Andre Oppermann, Internet Business Solutions AG 3 * 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 * 3. The name of the author may not be used to endorse or promote 14 * products derived from this software without specific prior written 15 * permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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 30 /* 31 * ip_fastforward gets its speed from processing the forwarded packet to 32 * completion (if_output on the other side) without any queues or netisr's. 33 * The receiving interface DMAs the packet into memory, the upper half of 34 * driver calls ip_fastforward, we do our routing table lookup and directly 35 * send it off to the outgoing interface, which DMAs the packet to the 36 * network card. The only part of the packet we touch with the CPU is the 37 * IP header (unless there are complex firewall rules touching other parts 38 * of the packet, but that is up to you). We are essentially limited by bus 39 * bandwidth and how fast the network card/driver can set up receives and 40 * transmits. 41 * 42 * We handle basic errors, IP header errors, checksum errors, 43 * destination unreachable, fragmentation and fragmentation needed and 44 * report them via ICMP to the sender. 45 * 46 * Else if something is not pure IPv4 unicast forwarding we fall back to 47 * the normal ip_input processing path. We should only be called from 48 * interfaces connected to the outside world. 49 * 50 * Firewalling is fully supported including divert, ipfw fwd and ipfilter 51 * ipnat and address rewrite. 52 * 53 * IPSEC is not supported if this host is a tunnel broker. IPSEC is 54 * supported for connections to/from local host. 55 * 56 * We try to do the least expensive (in CPU ops) checks and operations 57 * first to catch junk with as little overhead as possible. 58 * 59 * We take full advantage of hardware support for IP checksum and 60 * fragmentation offloading. 61 * 62 * We don't do ICMP redirect in the fast forwarding path. I have had my own 63 * cases where two core routers with Zebra routing suite would send millions 64 * ICMP redirects to connected hosts if the destination router was not the 65 * default gateway. In one case it was filling the routing table of a host 66 * with approximately 300.000 cloned redirect entries until it ran out of 67 * kernel memory. However the networking code proved very robust and it didn't 68 * crash or fail in other ways. 69 */ 70 71 /* 72 * Many thanks to Matt Thomas of NetBSD for basic structure of ip_flow.c which 73 * is being followed here. 74 */ 75 76 #include <sys/cdefs.h> 77 __FBSDID("$FreeBSD$"); 78 79 #include "opt_ipfw.h" 80 #include "opt_ipstealth.h" 81 82 #include <sys/param.h> 83 #include <sys/systm.h> 84 #include <sys/kernel.h> 85 #include <sys/malloc.h> 86 #include <sys/mbuf.h> 87 #include <sys/protosw.h> 88 #include <sys/socket.h> 89 #include <sys/sysctl.h> 90 #include <sys/vimage.h> 91 92 #include <net/pfil.h> 93 #include <net/if.h> 94 #include <net/if_types.h> 95 #include <net/if_var.h> 96 #include <net/if_dl.h> 97 #include <net/route.h> 98 99 #include <netinet/in.h> 100 #include <netinet/in_systm.h> 101 #include <netinet/in_var.h> 102 #include <netinet/ip.h> 103 #include <netinet/ip_var.h> 104 #include <netinet/ip_icmp.h> 105 #include <netinet/ip_options.h> 106 #include <netinet/vinet.h> 107 108 #include <machine/in_cksum.h> 109 110 #ifdef VIMAGE_GLOBALS 111 static int ipfastforward_active; 112 #endif 113 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fastforwarding, 114 CTLFLAG_RW, ipfastforward_active, 0, "Enable fast IP forwarding"); 115 116 static struct sockaddr_in * 117 ip_findroute(struct route *ro, struct in_addr dest, struct mbuf *m) 118 { 119 INIT_VNET_INET(curvnet); 120 struct sockaddr_in *dst; 121 struct rtentry *rt; 122 123 /* 124 * Find route to destination. 125 */ 126 bzero(ro, sizeof(*ro)); 127 dst = (struct sockaddr_in *)&ro->ro_dst; 128 dst->sin_family = AF_INET; 129 dst->sin_len = sizeof(*dst); 130 dst->sin_addr.s_addr = dest.s_addr; 131 in_rtalloc_ign(ro, 0, M_GETFIB(m)); 132 133 /* 134 * Route there and interface still up? 135 */ 136 rt = ro->ro_rt; 137 if (rt && (rt->rt_flags & RTF_UP) && 138 (rt->rt_ifp->if_flags & IFF_UP) && 139 (rt->rt_ifp->if_drv_flags & IFF_DRV_RUNNING)) { 140 if (rt->rt_flags & RTF_GATEWAY) 141 dst = (struct sockaddr_in *)rt->rt_gateway; 142 } else { 143 V_ipstat.ips_noroute++; 144 V_ipstat.ips_cantforward++; 145 if (rt) 146 RTFREE(rt); 147 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 148 return NULL; 149 } 150 return dst; 151 } 152 153 /* 154 * Try to forward a packet based on the destination address. 155 * This is a fast path optimized for the plain forwarding case. 156 * If the packet is handled (and consumed) here then we return 1; 157 * otherwise 0 is returned and the packet should be delivered 158 * to ip_input for full processing. 159 */ 160 struct mbuf * 161 ip_fastforward(struct mbuf *m) 162 { 163 INIT_VNET_INET(curvnet); 164 struct ip *ip; 165 struct mbuf *m0 = NULL; 166 struct route ro; 167 struct sockaddr_in *dst = NULL; 168 struct ifnet *ifp; 169 struct in_addr odest, dest; 170 u_short sum, ip_len; 171 int error = 0; 172 int hlen, mtu; 173 #ifdef IPFIREWALL_FORWARD 174 struct m_tag *fwd_tag; 175 #endif 176 177 /* 178 * Are we active and forwarding packets? 179 */ 180 if (!V_ipfastforward_active || !V_ipforwarding) 181 return m; 182 183 M_ASSERTVALID(m); 184 M_ASSERTPKTHDR(m); 185 186 ro.ro_rt = NULL; 187 188 /* 189 * Step 1: check for packet drop conditions (and sanity checks) 190 */ 191 192 /* 193 * Is entire packet big enough? 194 */ 195 if (m->m_pkthdr.len < sizeof(struct ip)) { 196 V_ipstat.ips_tooshort++; 197 goto drop; 198 } 199 200 /* 201 * Is first mbuf large enough for ip header and is header present? 202 */ 203 if (m->m_len < sizeof (struct ip) && 204 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 205 V_ipstat.ips_toosmall++; 206 return NULL; /* mbuf already free'd */ 207 } 208 209 ip = mtod(m, struct ip *); 210 211 /* 212 * Is it IPv4? 213 */ 214 if (ip->ip_v != IPVERSION) { 215 V_ipstat.ips_badvers++; 216 goto drop; 217 } 218 219 /* 220 * Is IP header length correct and is it in first mbuf? 221 */ 222 hlen = ip->ip_hl << 2; 223 if (hlen < sizeof(struct ip)) { /* minimum header length */ 224 V_ipstat.ips_badlen++; 225 goto drop; 226 } 227 if (hlen > m->m_len) { 228 if ((m = m_pullup(m, hlen)) == NULL) { 229 V_ipstat.ips_badhlen++; 230 return NULL; /* mbuf already free'd */ 231 } 232 ip = mtod(m, struct ip *); 233 } 234 235 /* 236 * Checksum correct? 237 */ 238 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) 239 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 240 else { 241 if (hlen == sizeof(struct ip)) 242 sum = in_cksum_hdr(ip); 243 else 244 sum = in_cksum(m, hlen); 245 } 246 if (sum) { 247 V_ipstat.ips_badsum++; 248 goto drop; 249 } 250 251 /* 252 * Remember that we have checked the IP header and found it valid. 253 */ 254 m->m_pkthdr.csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 255 256 ip_len = ntohs(ip->ip_len); 257 258 /* 259 * Is IP length longer than packet we have got? 260 */ 261 if (m->m_pkthdr.len < ip_len) { 262 V_ipstat.ips_tooshort++; 263 goto drop; 264 } 265 266 /* 267 * Is packet longer than IP header tells us? If yes, truncate packet. 268 */ 269 if (m->m_pkthdr.len > ip_len) { 270 if (m->m_len == m->m_pkthdr.len) { 271 m->m_len = ip_len; 272 m->m_pkthdr.len = ip_len; 273 } else 274 m_adj(m, ip_len - m->m_pkthdr.len); 275 } 276 277 /* 278 * Is packet from or to 127/8? 279 */ 280 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 281 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 282 V_ipstat.ips_badaddr++; 283 goto drop; 284 } 285 286 #ifdef ALTQ 287 /* 288 * Is packet dropped by traffic conditioner? 289 */ 290 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 291 goto drop; 292 #endif 293 294 /* 295 * Step 2: fallback conditions to normal ip_input path processing 296 */ 297 298 /* 299 * Only IP packets without options 300 */ 301 if (ip->ip_hl != (sizeof(struct ip) >> 2)) { 302 if (ip_doopts == 1) 303 return m; 304 else if (ip_doopts == 2) { 305 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_FILTER_PROHIB, 306 0, 0); 307 return NULL; /* mbuf already free'd */ 308 } 309 /* else ignore IP options and continue */ 310 } 311 312 /* 313 * Only unicast IP, not from loopback, no L2 or IP broadcast, 314 * no multicast, no INADDR_ANY 315 * 316 * XXX: Probably some of these checks could be direct drop 317 * conditions. However it is not clear whether there are some 318 * hacks or obscure behaviours which make it neccessary to 319 * let ip_input handle it. We play safe here and let ip_input 320 * deal with it until it is proven that we can directly drop it. 321 */ 322 if ((m->m_flags & (M_BCAST|M_MCAST)) || 323 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) || 324 ntohl(ip->ip_src.s_addr) == (u_long)INADDR_BROADCAST || 325 ntohl(ip->ip_dst.s_addr) == (u_long)INADDR_BROADCAST || 326 IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || 327 IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || 328 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) || 329 IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) || 330 ip->ip_src.s_addr == INADDR_ANY || 331 ip->ip_dst.s_addr == INADDR_ANY ) 332 return m; 333 334 /* 335 * Is it for a local address on this host? 336 */ 337 if (in_localip(ip->ip_dst)) 338 return m; 339 340 V_ipstat.ips_total++; 341 342 /* 343 * Step 3: incoming packet firewall processing 344 */ 345 346 /* 347 * Convert to host representation 348 */ 349 ip->ip_len = ntohs(ip->ip_len); 350 ip->ip_off = ntohs(ip->ip_off); 351 352 odest.s_addr = dest.s_addr = ip->ip_dst.s_addr; 353 354 /* 355 * Run through list of ipfilter hooks for input packets 356 */ 357 if (!PFIL_HOOKED(&inet_pfil_hook)) 358 goto passin; 359 360 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL) || 361 m == NULL) 362 goto drop; 363 364 M_ASSERTVALID(m); 365 M_ASSERTPKTHDR(m); 366 367 ip = mtod(m, struct ip *); /* m may have changed by pfil hook */ 368 dest.s_addr = ip->ip_dst.s_addr; 369 370 /* 371 * Destination address changed? 372 */ 373 if (odest.s_addr != dest.s_addr) { 374 /* 375 * Is it now for a local address on this host? 376 */ 377 if (in_localip(dest)) 378 goto forwardlocal; 379 /* 380 * Go on with new destination address 381 */ 382 } 383 #ifdef IPFIREWALL_FORWARD 384 if (m->m_flags & M_FASTFWD_OURS) { 385 /* 386 * ipfw changed it for a local address on this host. 387 */ 388 goto forwardlocal; 389 } 390 #endif /* IPFIREWALL_FORWARD */ 391 392 passin: 393 /* 394 * Step 4: decrement TTL and look up route 395 */ 396 397 /* 398 * Check TTL 399 */ 400 #ifdef IPSTEALTH 401 if (!V_ipstealth) { 402 #endif 403 if (ip->ip_ttl <= IPTTLDEC) { 404 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0); 405 return NULL; /* mbuf already free'd */ 406 } 407 408 /* 409 * Decrement the TTL and incrementally change the IP header checksum. 410 * Don't bother doing this with hw checksum offloading, it's faster 411 * doing it right here. 412 */ 413 ip->ip_ttl -= IPTTLDEC; 414 if (ip->ip_sum >= (u_int16_t) ~htons(IPTTLDEC << 8)) 415 ip->ip_sum -= ~htons(IPTTLDEC << 8); 416 else 417 ip->ip_sum += htons(IPTTLDEC << 8); 418 #ifdef IPSTEALTH 419 } 420 #endif 421 422 /* 423 * Find route to destination. 424 */ 425 if ((dst = ip_findroute(&ro, dest, m)) == NULL) 426 return NULL; /* icmp unreach already sent */ 427 ifp = ro.ro_rt->rt_ifp; 428 429 /* 430 * Immediately drop blackholed traffic, and directed broadcasts 431 * for either the all-ones or all-zero subnet addresses on 432 * locally attached networks. 433 */ 434 if ((ro.ro_rt->rt_flags & (RTF_BLACKHOLE|RTF_BROADCAST)) != 0) 435 goto drop; 436 437 /* 438 * Step 5: outgoing firewall packet processing 439 */ 440 441 /* 442 * Run through list of hooks for output packets. 443 */ 444 if (!PFIL_HOOKED(&inet_pfil_hook)) 445 goto passout; 446 447 if (pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, NULL) || m == NULL) { 448 goto drop; 449 } 450 451 M_ASSERTVALID(m); 452 M_ASSERTPKTHDR(m); 453 454 ip = mtod(m, struct ip *); 455 dest.s_addr = ip->ip_dst.s_addr; 456 457 /* 458 * Destination address changed? 459 */ 460 #ifndef IPFIREWALL_FORWARD 461 if (odest.s_addr != dest.s_addr) { 462 #else 463 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 464 if (odest.s_addr != dest.s_addr || fwd_tag != NULL) { 465 #endif /* IPFIREWALL_FORWARD */ 466 /* 467 * Is it now for a local address on this host? 468 */ 469 #ifndef IPFIREWALL_FORWARD 470 if (in_localip(dest)) { 471 #else 472 if (m->m_flags & M_FASTFWD_OURS || in_localip(dest)) { 473 #endif /* IPFIREWALL_FORWARD */ 474 forwardlocal: 475 /* 476 * Return packet for processing by ip_input(). 477 * Keep host byte order as expected at ip_input's 478 * "ours"-label. 479 */ 480 m->m_flags |= M_FASTFWD_OURS; 481 if (ro.ro_rt) 482 RTFREE(ro.ro_rt); 483 return m; 484 } 485 /* 486 * Redo route lookup with new destination address 487 */ 488 #ifdef IPFIREWALL_FORWARD 489 if (fwd_tag) { 490 dest.s_addr = ((struct sockaddr_in *) 491 (fwd_tag + 1))->sin_addr.s_addr; 492 m_tag_delete(m, fwd_tag); 493 } 494 #endif /* IPFIREWALL_FORWARD */ 495 RTFREE(ro.ro_rt); 496 if ((dst = ip_findroute(&ro, dest, m)) == NULL) 497 return NULL; /* icmp unreach already sent */ 498 ifp = ro.ro_rt->rt_ifp; 499 } 500 501 passout: 502 /* 503 * Step 6: send off the packet 504 */ 505 506 /* 507 * Check if route is dampned (when ARP is unable to resolve) 508 */ 509 if ((ro.ro_rt->rt_flags & RTF_REJECT) && 510 (ro.ro_rt->rt_rmx.rmx_expire == 0 || 511 time_uptime < ro.ro_rt->rt_rmx.rmx_expire)) { 512 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 513 goto consumed; 514 } 515 516 #ifndef ALTQ 517 /* 518 * Check if there is enough space in the interface queue 519 */ 520 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 521 ifp->if_snd.ifq_maxlen) { 522 V_ipstat.ips_odropped++; 523 /* would send source quench here but that is depreciated */ 524 goto drop; 525 } 526 #endif 527 528 /* 529 * Check if media link state of interface is not down 530 */ 531 if (ifp->if_link_state == LINK_STATE_DOWN) { 532 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 533 goto consumed; 534 } 535 536 /* 537 * Check if packet fits MTU or if hardware will fragment for us 538 */ 539 if (ro.ro_rt->rt_rmx.rmx_mtu) 540 mtu = min(ro.ro_rt->rt_rmx.rmx_mtu, ifp->if_mtu); 541 else 542 mtu = ifp->if_mtu; 543 544 if (ip->ip_len <= mtu || 545 (ifp->if_hwassist & CSUM_FRAGMENT && (ip->ip_off & IP_DF) == 0)) { 546 /* 547 * Restore packet header fields to original values 548 */ 549 ip->ip_len = htons(ip->ip_len); 550 ip->ip_off = htons(ip->ip_off); 551 /* 552 * Send off the packet via outgoing interface 553 */ 554 error = (*ifp->if_output)(ifp, m, 555 (struct sockaddr *)dst, ro.ro_rt); 556 } else { 557 /* 558 * Handle EMSGSIZE with icmp reply needfrag for TCP MTU discovery 559 */ 560 if (ip->ip_off & IP_DF) { 561 V_ipstat.ips_cantfrag++; 562 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 563 0, mtu); 564 goto consumed; 565 } else { 566 /* 567 * We have to fragment the packet 568 */ 569 m->m_pkthdr.csum_flags |= CSUM_IP; 570 /* 571 * ip_fragment expects ip_len and ip_off in host byte 572 * order but returns all packets in network byte order 573 */ 574 if (ip_fragment(ip, &m, mtu, ifp->if_hwassist, 575 (~ifp->if_hwassist & CSUM_DELAY_IP))) { 576 goto drop; 577 } 578 KASSERT(m != NULL, ("null mbuf and no error")); 579 /* 580 * Send off the fragments via outgoing interface 581 */ 582 error = 0; 583 do { 584 m0 = m->m_nextpkt; 585 m->m_nextpkt = NULL; 586 587 error = (*ifp->if_output)(ifp, m, 588 (struct sockaddr *)dst, ro.ro_rt); 589 if (error) 590 break; 591 } while ((m = m0) != NULL); 592 if (error) { 593 /* Reclaim remaining fragments */ 594 for (m = m0; m; m = m0) { 595 m0 = m->m_nextpkt; 596 m_freem(m); 597 } 598 } else 599 V_ipstat.ips_fragmented++; 600 } 601 } 602 603 if (error != 0) 604 V_ipstat.ips_odropped++; 605 else { 606 ro.ro_rt->rt_rmx.rmx_pksent++; 607 V_ipstat.ips_forward++; 608 V_ipstat.ips_fastforward++; 609 } 610 consumed: 611 RTFREE(ro.ro_rt); 612 return NULL; 613 drop: 614 if (m) 615 m_freem(m); 616 if (ro.ro_rt) 617 RTFREE(ro.ro_rt); 618 return NULL; 619 } 620