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#include "opt_kdtrace.h" 82 83#include <sys/param.h> 84#include <sys/systm.h> 85#include <sys/kernel.h> 86#include <sys/malloc.h> 87#include <sys/mbuf.h> 88#include <sys/protosw.h> 89#include <sys/sdt.h> 90#include <sys/socket.h> 91#include <sys/sysctl.h> 92 93#include <net/pfil.h> 94#include <net/if.h> 95#include <net/if_types.h> 96#include <net/if_var.h> 97#include <net/if_dl.h> 98#include <net/route.h> 99#include <net/vnet.h> 100 101#include <netinet/in.h> 102#include <netinet/in_kdtrace.h> 103#include <netinet/in_systm.h> 104#include <netinet/in_var.h> 105#include <netinet/ip.h> 106#include <netinet/ip_var.h> 107#include <netinet/ip_icmp.h> 108#include <netinet/ip_options.h> 109 110#include <machine/in_cksum.h> 111 112static VNET_DEFINE(int, ipfastforward_active); 113#define V_ipfastforward_active VNET(ipfastforward_active) 114 115SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fastforwarding, CTLFLAG_RW, 116 &VNET_NAME(ipfastforward_active), 0, "Enable fast IP forwarding"); 117 118static struct sockaddr_in * 119ip_findroute(struct route *ro, struct in_addr dest, struct mbuf *m) 120{ 121 struct sockaddr_in *dst; 122 struct rtentry *rt; 123 124 /* 125 * Find route to destination. 126 */ 127 bzero(ro, sizeof(*ro)); 128 dst = (struct sockaddr_in *)&ro->ro_dst; 129 dst->sin_family = AF_INET; 130 dst->sin_len = sizeof(*dst); 131 dst->sin_addr.s_addr = dest.s_addr; 132 in_rtalloc_ign(ro, 0, M_GETFIB(m)); 133 134 /* 135 * Route there and interface still up? 136 */ 137 rt = ro->ro_rt; 138 if (rt && (rt->rt_flags & RTF_UP) && 139 (rt->rt_ifp->if_flags & IFF_UP) && 140 (rt->rt_ifp->if_drv_flags & IFF_DRV_RUNNING)) { 141 if (rt->rt_flags & RTF_GATEWAY) 142 dst = (struct sockaddr_in *)rt->rt_gateway; 143 } else { 144 IPSTAT_INC(ips_noroute); 145 IPSTAT_INC(ips_cantforward); 146 if (rt) 147 RTFREE(rt); 148 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 149 return NULL; 150 } 151 return dst; 152} 153 154/* 155 * Try to forward a packet based on the destination address. 156 * This is a fast path optimized for the plain forwarding case. 157 * If the packet is handled (and consumed) here then we return NULL; 158 * otherwise mbuf is returned and the packet should be delivered 159 * to ip_input for full processing. 160 */ 161struct mbuf * 162ip_fastforward(struct mbuf *m) 163{ 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 uint16_t sum, ip_len, ip_off; 171 int error = 0; 172 int hlen, mtu; 173 struct m_tag *fwd_tag = NULL; 174 175 /* 176 * Are we active and forwarding packets? 177 */ 178 if (!V_ipfastforward_active || !V_ipforwarding) 179 return m; 180 181 M_ASSERTVALID(m); 182 M_ASSERTPKTHDR(m); 183 184 bzero(&ro, sizeof(ro)); 185 186 /* 187 * Step 1: check for packet drop conditions (and sanity checks) 188 */ 189 190 /* 191 * Is entire packet big enough? 192 */ 193 if (m->m_pkthdr.len < sizeof(struct ip)) { 194 IPSTAT_INC(ips_tooshort); 195 goto drop; 196 } 197 198 /* 199 * Is first mbuf large enough for ip header and is header present? 200 */ 201 if (m->m_len < sizeof (struct ip) && 202 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 203 IPSTAT_INC(ips_toosmall); 204 return NULL; /* mbuf already free'd */ 205 } 206 207 ip = mtod(m, struct ip *); 208 209 /* 210 * Is it IPv4? 211 */ 212 if (ip->ip_v != IPVERSION) { 213 IPSTAT_INC(ips_badvers); 214 goto drop; 215 } 216 217 /* 218 * Is IP header length correct and is it in first mbuf? 219 */ 220 hlen = ip->ip_hl << 2; 221 if (hlen < sizeof(struct ip)) { /* minimum header length */ 222 IPSTAT_INC(ips_badhlen); 223 goto drop; 224 } 225 if (hlen > m->m_len) { 226 if ((m = m_pullup(m, hlen)) == NULL) { 227 IPSTAT_INC(ips_badhlen); 228 return NULL; /* mbuf already free'd */ 229 } 230 ip = mtod(m, struct ip *); 231 } 232 233 /* 234 * Checksum correct? 235 */ 236 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) 237 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 238 else { 239 if (hlen == sizeof(struct ip)) 240 sum = in_cksum_hdr(ip); 241 else 242 sum = in_cksum(m, hlen); 243 } 244 if (sum) { 245 IPSTAT_INC(ips_badsum); 246 goto drop; 247 } 248 249 /* 250 * Remember that we have checked the IP header and found it valid. 251 */ 252 m->m_pkthdr.csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 253 254 ip_len = ntohs(ip->ip_len); 255 256 /* 257 * Is IP length longer than packet we have got? 258 */ 259 if (m->m_pkthdr.len < ip_len) { 260 IPSTAT_INC(ips_tooshort); 261 goto drop; 262 } 263 264 /* 265 * Is packet longer than IP header tells us? If yes, truncate packet. 266 */ 267 if (m->m_pkthdr.len > ip_len) { 268 if (m->m_len == m->m_pkthdr.len) { 269 m->m_len = ip_len; 270 m->m_pkthdr.len = ip_len; 271 } else 272 m_adj(m, ip_len - m->m_pkthdr.len); 273 } 274 275 /* 276 * Is packet from or to 127/8? 277 */ 278 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 279 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 280 IPSTAT_INC(ips_badaddr); 281 goto drop; 282 } 283 284#ifdef ALTQ 285 /* 286 * Is packet dropped by traffic conditioner? 287 */ 288 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 289 goto drop; 290#endif 291 292 /* 293 * Step 2: fallback conditions to normal ip_input path processing 294 */ 295 296 /* 297 * Only IP packets without options 298 */ 299 if (ip->ip_hl != (sizeof(struct ip) >> 2)) { 300 if (V_ip_doopts == 1) 301 return m; 302 else if (V_ip_doopts == 2) { 303 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_FILTER_PROHIB, 304 0, 0); 305 return NULL; /* mbuf already free'd */ 306 } 307 /* else ignore IP options and continue */ 308 } 309 310 /* 311 * Only unicast IP, not from loopback, no L2 or IP broadcast, 312 * no multicast, no INADDR_ANY 313 * 314 * XXX: Probably some of these checks could be direct drop 315 * conditions. However it is not clear whether there are some 316 * hacks or obscure behaviours which make it neccessary to 317 * let ip_input handle it. We play safe here and let ip_input 318 * deal with it until it is proven that we can directly drop it. 319 */ 320 if ((m->m_flags & (M_BCAST|M_MCAST)) || 321 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) || 322 ntohl(ip->ip_src.s_addr) == (u_long)INADDR_BROADCAST || 323 ntohl(ip->ip_dst.s_addr) == (u_long)INADDR_BROADCAST || 324 IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || 325 IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || 326 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) || 327 IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) || 328 ip->ip_src.s_addr == INADDR_ANY || 329 ip->ip_dst.s_addr == INADDR_ANY ) 330 return m; 331 332 /* 333 * Is it for a local address on this host? 334 */ 335 if (in_localip(ip->ip_dst)) 336 return m; 337 338 IPSTAT_INC(ips_total); 339 340 /* 341 * Step 3: incoming packet firewall processing 342 */ 343 344 odest.s_addr = dest.s_addr = ip->ip_dst.s_addr; 345 346 /* 347 * Run through list of ipfilter hooks for input packets 348 */ 349 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 350 goto passin; 351 352 if (pfil_run_hooks( 353 &V_inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL) || 354 m == NULL) 355 goto drop; 356 357 M_ASSERTVALID(m); 358 M_ASSERTPKTHDR(m); 359 360 ip = mtod(m, struct ip *); /* m may have changed by pfil hook */ 361 dest.s_addr = ip->ip_dst.s_addr; 362 363 /* 364 * Destination address changed? 365 */ 366 if (odest.s_addr != dest.s_addr) { 367 /* 368 * Is it now for a local address on this host? 369 */ 370 if (in_localip(dest)) 371 goto forwardlocal; 372 /* 373 * Go on with new destination address 374 */ 375 } 376 377 if (m->m_flags & M_FASTFWD_OURS) { 378 /* 379 * ipfw changed it for a local address on this host. 380 */ 381 goto forwardlocal; 382 } 383 384passin: 385 /* 386 * Step 4: decrement TTL and look up route 387 */ 388 389 /* 390 * Check TTL 391 */ 392#ifdef IPSTEALTH 393 if (!V_ipstealth) { 394#endif 395 if (ip->ip_ttl <= IPTTLDEC) { 396 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0); 397 return NULL; /* mbuf already free'd */ 398 } 399 400 /* 401 * Decrement the TTL and incrementally change the IP header checksum. 402 * Don't bother doing this with hw checksum offloading, it's faster 403 * doing it right here. 404 */ 405 ip->ip_ttl -= IPTTLDEC; 406 if (ip->ip_sum >= (u_int16_t) ~htons(IPTTLDEC << 8)) 407 ip->ip_sum -= ~htons(IPTTLDEC << 8); 408 else 409 ip->ip_sum += htons(IPTTLDEC << 8); 410#ifdef IPSTEALTH 411 } 412#endif 413 414 /* 415 * Find route to destination. 416 */ 417 if ((dst = ip_findroute(&ro, dest, m)) == NULL) 418 return NULL; /* icmp unreach already sent */ 419 ifp = ro.ro_rt->rt_ifp; 420 421 /* 422 * Immediately drop blackholed traffic, and directed broadcasts 423 * for either the all-ones or all-zero subnet addresses on 424 * locally attached networks. 425 */ 426 if ((ro.ro_rt->rt_flags & (RTF_BLACKHOLE|RTF_BROADCAST)) != 0) 427 goto drop; 428 429 /* 430 * Step 5: outgoing firewall packet processing 431 */ 432 433 /* 434 * Run through list of hooks for output packets. 435 */ 436 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 437 goto passout; 438 439 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, NULL) || m == NULL) { 440 goto drop; 441 } 442 443 M_ASSERTVALID(m); 444 M_ASSERTPKTHDR(m); 445 446 ip = mtod(m, struct ip *); 447 dest.s_addr = ip->ip_dst.s_addr; 448 449 /* 450 * Destination address changed? 451 */ 452 if (m->m_flags & M_IP_NEXTHOP) 453 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 454 if (odest.s_addr != dest.s_addr || fwd_tag != NULL) { 455 /* 456 * Is it now for a local address on this host? 457 */ 458 if (m->m_flags & M_FASTFWD_OURS || in_localip(dest)) { 459forwardlocal: 460 /* 461 * Return packet for processing by ip_input(). 462 */ 463 m->m_flags |= M_FASTFWD_OURS; 464 if (ro.ro_rt) 465 RTFREE(ro.ro_rt); 466 return m; 467 } 468 /* 469 * Redo route lookup with new destination address 470 */ 471 if (fwd_tag) { 472 dest.s_addr = ((struct sockaddr_in *) 473 (fwd_tag + 1))->sin_addr.s_addr; 474 m_tag_delete(m, fwd_tag); 475 m->m_flags &= ~M_IP_NEXTHOP; 476 } 477 RTFREE(ro.ro_rt); 478 if ((dst = ip_findroute(&ro, dest, m)) == NULL) 479 return NULL; /* icmp unreach already sent */ 480 ifp = ro.ro_rt->rt_ifp; 481 } 482 483passout: 484 /* 485 * Step 6: send off the packet 486 */ 487 ip_len = ntohs(ip->ip_len); 488 ip_off = ntohs(ip->ip_off); 489 490 /* 491 * Check if route is dampned (when ARP is unable to resolve) 492 */ 493 if ((ro.ro_rt->rt_flags & RTF_REJECT) && 494 (ro.ro_rt->rt_expire == 0 || time_uptime < ro.ro_rt->rt_expire)) { 495 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 496 goto consumed; 497 } 498 499#ifndef ALTQ 500 /* 501 * Check if there is enough space in the interface queue 502 */ 503 if ((ifp->if_snd.ifq_len + ip_len / ifp->if_mtu + 1) >= 504 ifp->if_snd.ifq_maxlen) { 505 IPSTAT_INC(ips_odropped); 506 /* would send source quench here but that is depreciated */ 507 goto drop; 508 } 509#endif 510 511 /* 512 * Check if media link state of interface is not down 513 */ 514 if (ifp->if_link_state == LINK_STATE_DOWN) { 515 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 516 goto consumed; 517 } 518 519 /* 520 * Check if packet fits MTU or if hardware will fragment for us 521 */ 522 if (ro.ro_rt->rt_mtu) 523 mtu = min(ro.ro_rt->rt_mtu, ifp->if_mtu); 524 else 525 mtu = ifp->if_mtu; 526 527 if (ip_len <= mtu || 528 (ifp->if_hwassist & CSUM_FRAGMENT && (ip_off & IP_DF) == 0)) { 529 /* 530 * Avoid confusing lower layers. 531 */ 532 m_clrprotoflags(m); 533 /* 534 * Send off the packet via outgoing interface 535 */ 536 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); 537 error = (*ifp->if_output)(ifp, m, 538 (struct sockaddr *)dst, &ro); 539 } else { 540 /* 541 * Handle EMSGSIZE with icmp reply needfrag for TCP MTU discovery 542 */ 543 if (ip_off & IP_DF) { 544 IPSTAT_INC(ips_cantfrag); 545 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 546 0, mtu); 547 goto consumed; 548 } else { 549 /* 550 * We have to fragment the packet 551 */ 552 m->m_pkthdr.csum_flags |= CSUM_IP; 553 if (ip_fragment(ip, &m, mtu, ifp->if_hwassist)) 554 goto drop; 555 KASSERT(m != NULL, ("null mbuf and no error")); 556 /* 557 * Send off the fragments via outgoing interface 558 */ 559 error = 0; 560 do { 561 m0 = m->m_nextpkt; 562 m->m_nextpkt = NULL; 563 /* 564 * Avoid confusing lower layers. 565 */ 566 m_clrprotoflags(m); 567 568 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); 569 error = (*ifp->if_output)(ifp, m, 570 (struct sockaddr *)dst, &ro); 571 if (error) 572 break; 573 } while ((m = m0) != NULL); 574 if (error) { 575 /* Reclaim remaining fragments */ 576 for (m = m0; m; m = m0) { 577 m0 = m->m_nextpkt; 578 m_freem(m); 579 } 580 } else 581 IPSTAT_INC(ips_fragmented); 582 } 583 } 584 585 if (error != 0) 586 IPSTAT_INC(ips_odropped); 587 else { 588 counter_u64_add(ro.ro_rt->rt_pksent, 1); 589 IPSTAT_INC(ips_forward); 590 IPSTAT_INC(ips_fastforward); 591 } 592consumed: 593 RTFREE(ro.ro_rt); 594 return NULL; 595drop: 596 if (m) 597 m_freem(m); 598 if (ro.ro_rt) 599 RTFREE(ro.ro_rt); 600 return NULL; 601} 602