2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 2003 Andre Oppermann, Internet Business Solutions AG
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. The name of the author may not be used to endorse or promote
16 * products derived from this software without specific prior written
19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * ip_fastforward gets its speed from processing the forwarded packet to
34 * completion (if_output on the other side) without any queues or netisr's.
35 * The receiving interface DMAs the packet into memory, the upper half of
36 * driver calls ip_fastforward, we do our routing table lookup and directly
37 * send it off to the outgoing interface, which DMAs the packet to the
38 * network card. The only part of the packet we touch with the CPU is the
39 * IP header (unless there are complex firewall rules touching other parts
40 * of the packet, but that is up to you). We are essentially limited by bus
41 * bandwidth and how fast the network card/driver can set up receives and
44 * We handle basic errors, IP header errors, checksum errors,
45 * destination unreachable, fragmentation and fragmentation needed and
46 * report them via ICMP to the sender.
48 * Else if something is not pure IPv4 unicast forwarding we fall back to
49 * the normal ip_input processing path. We should only be called from
50 * interfaces connected to the outside world.
52 * Firewalling is fully supported including divert, ipfw fwd and ipfilter
53 * ipnat and address rewrite.
55 * IPSEC is not supported if this host is a tunnel broker. IPSEC is
56 * supported for connections to/from local host.
58 * We try to do the least expensive (in CPU ops) checks and operations
59 * first to catch junk with as little overhead as possible.
61 * We take full advantage of hardware support for IP checksum and
62 * fragmentation offloading.
66 * Many thanks to Matt Thomas of NetBSD for basic structure of ip_flow.c which
67 * is being followed here.
70 #include <sys/cdefs.h>
71 __FBSDID("$FreeBSD$");
73 #include "opt_ipstealth.h"
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/kernel.h>
78 #include <sys/malloc.h>
80 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/sysctl.h>
86 #include <net/if_types.h>
87 #include <net/if_var.h>
88 #include <net/if_dl.h>
90 #include <net/route.h>
91 #include <net/route/nhop.h>
94 #include <netinet/in.h>
95 #include <netinet/in_fib.h>
96 #include <netinet/in_kdtrace.h>
97 #include <netinet/in_systm.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip.h>
100 #include <netinet/ip_var.h>
101 #include <netinet/ip_icmp.h>
102 #include <netinet/ip_options.h>
104 #include <machine/in_cksum.h>
106 #define V_ipsendredirects VNET(ipsendredirects)
109 ip_redir_alloc(struct mbuf *m, struct nhop_object *nh, u_short ip_len,
110 struct in_addr *osrc, struct in_addr *newgw)
112 struct in_ifaddr *nh_ia;
115 KASSERT(nh != NULL, ("%s: m %p nh is NULL\n", __func__, m));
118 * Only send a redirect if:
119 * - Redirects are not disabled (must be checked by caller),
120 * - We have not applied NAT (must be checked by caller as possible),
121 * - Neither a MCAST or BCAST packet (must be checked by caller)
122 * [RFC1009 Appendix A.2].
123 * - The packet does not do IP source routing or having any other
124 * IP options (this case was handled already by ip_input() calling
125 * ip_dooptions() [RFC792, p13],
126 * - The packet is being forwarded out the same physical interface
127 * that it was received from [RFC1812, 5.2.7.2].
131 * - The forwarding route was not created by a redirect
132 * [RFC1812, 5.2.7.2], or
133 * if it was to follow a default route (see below).
134 * - The next-hop is reachable by us [RFC1009 Appendix A.2].
136 if ((nh->nh_flags & (NHF_DEFAULT | NHF_REDIRECT |
137 NHF_BLACKHOLE | NHF_REJECT)) != 0)
140 /* Get the new gateway. */
141 if ((nh->nh_flags & NHF_GATEWAY) == 0 || nh->gw_sa.sa_family != AF_INET)
143 newgw->s_addr = nh->gw4_sa.sin_addr.s_addr;
146 * - The resulting forwarding destination is not "This host on this
147 * network" [RFC1122, Section 3.2.1.3] (default route check above).
149 if (newgw->s_addr == 0)
153 * - We know how to reach the sender and the source address is
154 * directly connected to us [RFC792, p13].
155 * + The new gateway address and the source address are on the same
156 * subnet [RFC1009 Appendix A.2, RFC1122 3.2.2.2, RFC1812, 5.2.7.2].
157 * NB: if you think multiple logical subnets on the same wire should
158 * receive redirects read [RFC1812, APPENDIX C (14->15)].
160 nh_ia = (struct in_ifaddr *)nh->nh_ifa;
161 if ((ntohl(osrc->s_addr) & nh_ia->ia_subnetmask) != nh_ia->ia_subnet)
164 /* Prepare for sending the redirect. */
167 * Make a copy of as much as we need of the packet as the original
168 * one will be forwarded but we need (a portion) for icmp_error().
170 mcopy = m_gethdr(M_NOWAIT, m->m_type);
174 if (m_dup_pkthdr(mcopy, m, M_NOWAIT) == 0) {
176 * It's probably ok if the pkthdr dup fails (because
177 * the deep copy of the tag chain failed), but for now
178 * be conservative and just discard the copy since
179 * code below may some day want the tags.
184 mcopy->m_len = min(ip_len, M_TRAILINGSPACE(mcopy));
185 mcopy->m_pkthdr.len = mcopy->m_len;
186 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
193 ip_findroute(struct nhop_object **pnh, struct in_addr dest, struct mbuf *m)
195 struct nhop_object *nh;
197 nh = fib4_lookup(M_GETFIB(m), dest, 0, NHR_NONE,
200 IPSTAT_INC(ips_noroute);
201 IPSTAT_INC(ips_cantforward);
202 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
203 return (EHOSTUNREACH);
206 * Drop blackholed traffic and directed broadcasts.
208 if ((nh->nh_flags & (NHF_BLACKHOLE | NHF_BROADCAST)) != 0) {
209 IPSTAT_INC(ips_cantforward);
211 return (EHOSTUNREACH);
214 if (nh->nh_flags & NHF_REJECT) {
215 IPSTAT_INC(ips_cantforward);
216 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
217 return (EHOSTUNREACH);
226 * Try to forward a packet based on the destination address.
227 * This is a fast path optimized for the plain forwarding case.
228 * If the packet is handled (and consumed) here then we return NULL;
229 * otherwise mbuf is returned and the packet should be delivered
230 * to ip_input for full processing.
233 ip_tryforward(struct mbuf *m)
236 struct mbuf *m0 = NULL;
237 struct nhop_object *nh = NULL;
239 struct sockaddr_in *dst;
240 const struct sockaddr *gw;
241 struct in_addr dest, odest, rtdest, osrc;
242 uint16_t ip_len, ip_off;
244 struct m_tag *fwd_tag = NULL;
245 struct mbuf *mcopy = NULL;
246 struct in_addr redest;
248 * Are we active and forwarding packets?
255 * Only IP packets without options
257 ip = mtod(m, struct ip *);
259 if (ip->ip_hl != (sizeof(struct ip) >> 2)) {
260 if (V_ip_doopts == 1)
262 else if (V_ip_doopts == 2) {
263 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_FILTER_PROHIB,
265 return NULL; /* mbuf already free'd */
267 /* else ignore IP options and continue */
271 * Only unicast IP, not from loopback, no L2 or IP broadcast,
272 * no multicast, no INADDR_ANY
274 * XXX: Probably some of these checks could be direct drop
275 * conditions. However it is not clear whether there are some
276 * hacks or obscure behaviours which make it necessary to
277 * let ip_input handle it. We play safe here and let ip_input
278 * deal with it until it is proven that we can directly drop it.
280 if ((m->m_flags & (M_BCAST|M_MCAST)) ||
281 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) ||
282 ntohl(ip->ip_src.s_addr) == (u_long)INADDR_BROADCAST ||
283 ntohl(ip->ip_dst.s_addr) == (u_long)INADDR_BROADCAST ||
284 IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
285 IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
286 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) ||
287 IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
288 ip->ip_src.s_addr == INADDR_ANY ||
289 ip->ip_dst.s_addr == INADDR_ANY )
293 * Is it for a local address on this host?
295 if (in_localip(ip->ip_dst))
298 IPSTAT_INC(ips_total);
301 * Step 3: incoming packet firewall processing
304 odest.s_addr = dest.s_addr = ip->ip_dst.s_addr;
305 osrc.s_addr = ip->ip_src.s_addr;
308 * Run through list of ipfilter hooks for input packets
310 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
313 if (pfil_mbuf_in(V_inet_pfil_head, &m, m->m_pkthdr.rcvif,
320 ip = mtod(m, struct ip *); /* m may have changed by pfil hook */
321 dest.s_addr = ip->ip_dst.s_addr;
324 * Destination address changed?
326 if (odest.s_addr != dest.s_addr) {
328 * Is it now for a local address on this host?
330 if (in_localip(dest))
333 * Go on with new destination address
337 if (m->m_flags & M_FASTFWD_OURS) {
339 * ipfw changed it for a local address on this host.
346 * Step 4: decrement TTL and look up route
355 if (ip->ip_ttl <= IPTTLDEC) {
356 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
357 return NULL; /* mbuf already free'd */
361 * Decrement the TTL and incrementally change the IP header checksum.
362 * Don't bother doing this with hw checksum offloading, it's faster
363 * doing it right here.
365 ip->ip_ttl -= IPTTLDEC;
366 if (ip->ip_sum >= (u_int16_t) ~htons(IPTTLDEC << 8))
367 ip->ip_sum -= ~htons(IPTTLDEC << 8);
369 ip->ip_sum += htons(IPTTLDEC << 8);
375 * Next hop forced by pfil(9) hook?
377 if ((m->m_flags & M_IP_NEXTHOP) &&
378 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
380 * Now we will find route to forced destination.
382 dest.s_addr = ((struct sockaddr_in *)
383 (fwd_tag + 1))->sin_addr.s_addr;
384 m_tag_delete(m, fwd_tag);
385 m->m_flags &= ~M_IP_NEXTHOP;
389 * Find route to destination.
391 if (ip_findroute(&nh, dest, m) != 0)
392 return (NULL); /* icmp unreach already sent */
395 * Avoid second route lookup by caching destination.
397 rtdest.s_addr = dest.s_addr;
400 * Step 5: outgoing firewall packet processing
402 if (!PFIL_HOOKED_OUT(V_inet_pfil_head))
405 if (pfil_mbuf_out(V_inet_pfil_head, &m, nh->nh_ifp,
412 ip = mtod(m, struct ip *);
413 dest.s_addr = ip->ip_dst.s_addr;
416 * Destination address changed?
418 if (m->m_flags & M_IP_NEXTHOP)
419 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
422 if (odest.s_addr != dest.s_addr || fwd_tag != NULL) {
424 * Is it now for a local address on this host?
426 if (m->m_flags & M_FASTFWD_OURS || in_localip(dest)) {
429 * Return packet for processing by ip_input().
431 m->m_flags |= M_FASTFWD_OURS;
435 * Redo route lookup with new destination address
438 dest.s_addr = ((struct sockaddr_in *)
439 (fwd_tag + 1))->sin_addr.s_addr;
440 m_tag_delete(m, fwd_tag);
441 m->m_flags &= ~M_IP_NEXTHOP;
443 if (dest.s_addr != rtdest.s_addr &&
444 ip_findroute(&nh, dest, m) != 0)
445 return (NULL); /* icmp unreach already sent */
450 * Step 6: send off the packet
452 ip_len = ntohs(ip->ip_len);
453 ip_off = ntohs(ip->ip_off);
455 bzero(&ro, sizeof(ro));
456 dst = (struct sockaddr_in *)&ro.ro_dst;
457 dst->sin_family = AF_INET;
458 dst->sin_len = sizeof(*dst);
459 dst->sin_addr = dest;
460 if (nh->nh_flags & NHF_GATEWAY) {
462 ro.ro_flags |= RT_HAS_GW;
464 gw = (const struct sockaddr *)dst;
466 /* Handle redirect case. */
468 if (V_ipsendredirects && osrc.s_addr == ip->ip_src.s_addr &&
469 nh->nh_ifp == m->m_pkthdr.rcvif)
470 mcopy = ip_redir_alloc(m, nh, ip_len, &osrc, &redest);
473 * Check if packet fits MTU or if hardware will fragment for us
475 if (ip_len <= nh->nh_mtu) {
477 * Avoid confusing lower layers.
481 * Send off the packet via outgoing interface
483 IP_PROBE(send, NULL, NULL, ip, nh->nh_ifp, ip, NULL);
484 error = (*nh->nh_ifp->if_output)(nh->nh_ifp, m, gw, &ro);
487 * Handle EMSGSIZE with icmp reply needfrag for TCP MTU discovery
489 if (ip_off & IP_DF) {
490 IPSTAT_INC(ips_cantfrag);
491 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG,
496 * We have to fragment the packet
498 m->m_pkthdr.csum_flags |= CSUM_IP;
499 if (ip_fragment(ip, &m, nh->nh_mtu,
500 nh->nh_ifp->if_hwassist) != 0)
502 KASSERT(m != NULL, ("null mbuf and no error"));
504 * Send off the fragments via outgoing interface
511 * Avoid confusing lower layers.
515 IP_PROBE(send, NULL, NULL,
516 mtod(m, struct ip *), nh->nh_ifp,
517 mtod(m, struct ip *), NULL);
518 error = (*nh->nh_ifp->if_output)(nh->nh_ifp, m,
522 } while ((m = m0) != NULL);
524 /* Reclaim remaining fragments */
525 for (m = m0; m; m = m0) {
530 IPSTAT_INC(ips_fragmented);
535 IPSTAT_INC(ips_odropped);
537 IPSTAT_INC(ips_forward);
538 IPSTAT_INC(ips_fastforward);
541 /* Send required redirect */
543 icmp_error(mcopy, ICMP_REDIRECT, ICMP_REDIRECT_HOST, redest.s_addr, 0);
544 mcopy = NULL; /* Was consumed by callee. */