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.
64 * We don't do ICMP redirect in the fast forwarding path. I have had my own
65 * cases where two core routers with Zebra routing suite would send millions
66 * ICMP redirects to connected hosts if the destination router was not the
67 * default gateway. In one case it was filling the routing table of a host
68 * with approximately 300.000 cloned redirect entries until it ran out of
69 * kernel memory. However the networking code proved very robust and it didn't
70 * crash or fail in other ways.
74 * Many thanks to Matt Thomas of NetBSD for basic structure of ip_flow.c which
75 * is being followed here.
78 #include <sys/cdefs.h>
79 __FBSDID("$FreeBSD$");
81 #include "opt_ipstealth.h"
83 #include <sys/param.h>
84 #include <sys/systm.h>
85 #include <sys/kernel.h>
86 #include <sys/malloc.h>
88 #include <sys/protosw.h>
90 #include <sys/socket.h>
91 #include <sys/sysctl.h>
95 #include <net/if_types.h>
96 #include <net/if_var.h>
97 #include <net/if_dl.h>
98 #include <net/route.h>
101 #include <netinet/in.h>
102 #include <netinet/in_fib.h>
103 #include <netinet/in_kdtrace.h>
104 #include <netinet/in_systm.h>
105 #include <netinet/in_var.h>
106 #include <netinet/ip.h>
107 #include <netinet/ip_var.h>
108 #include <netinet/ip_icmp.h>
109 #include <netinet/ip_options.h>
111 #include <machine/in_cksum.h>
114 ip_findroute(struct nhop4_basic *pnh, struct in_addr dest, struct mbuf *m)
117 bzero(pnh, sizeof(*pnh));
118 if (fib4_lookup_nh_basic(M_GETFIB(m), dest, 0, 0, pnh) != 0) {
119 IPSTAT_INC(ips_noroute);
120 IPSTAT_INC(ips_cantforward);
121 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
122 return (EHOSTUNREACH);
125 * Drop blackholed traffic and directed broadcasts.
127 if ((pnh->nh_flags & (NHF_BLACKHOLE | NHF_BROADCAST)) != 0) {
128 IPSTAT_INC(ips_cantforward);
130 return (EHOSTUNREACH);
133 if (pnh->nh_flags & NHF_REJECT) {
134 IPSTAT_INC(ips_cantforward);
135 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
136 return (EHOSTUNREACH);
143 * Try to forward a packet based on the destination address.
144 * This is a fast path optimized for the plain forwarding case.
145 * If the packet is handled (and consumed) here then we return NULL;
146 * otherwise mbuf is returned and the packet should be delivered
147 * to ip_input for full processing.
150 ip_tryforward(struct mbuf *m)
153 struct mbuf *m0 = NULL;
154 struct nhop4_basic nh;
155 struct sockaddr_in dst;
156 struct in_addr dest, odest, rtdest;
157 uint16_t ip_len, ip_off;
159 struct m_tag *fwd_tag = NULL;
162 * Are we active and forwarding packets?
170 * Is packet dropped by traffic conditioner?
172 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
177 * Only IP packets without options
179 ip = mtod(m, struct ip *);
181 if (ip->ip_hl != (sizeof(struct ip) >> 2)) {
182 if (V_ip_doopts == 1)
184 else if (V_ip_doopts == 2) {
185 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_FILTER_PROHIB,
187 return NULL; /* mbuf already free'd */
189 /* else ignore IP options and continue */
193 * Only unicast IP, not from loopback, no L2 or IP broadcast,
194 * no multicast, no INADDR_ANY
196 * XXX: Probably some of these checks could be direct drop
197 * conditions. However it is not clear whether there are some
198 * hacks or obscure behaviours which make it necessary to
199 * let ip_input handle it. We play safe here and let ip_input
200 * deal with it until it is proven that we can directly drop it.
202 if ((m->m_flags & (M_BCAST|M_MCAST)) ||
203 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) ||
204 ntohl(ip->ip_src.s_addr) == (u_long)INADDR_BROADCAST ||
205 ntohl(ip->ip_dst.s_addr) == (u_long)INADDR_BROADCAST ||
206 IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
207 IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
208 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) ||
209 IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
210 ip->ip_src.s_addr == INADDR_ANY ||
211 ip->ip_dst.s_addr == INADDR_ANY )
215 * Is it for a local address on this host?
217 if (in_localip(ip->ip_dst))
220 IPSTAT_INC(ips_total);
223 * Step 3: incoming packet firewall processing
226 odest.s_addr = dest.s_addr = ip->ip_dst.s_addr;
229 * Run through list of ipfilter hooks for input packets
231 if (!PFIL_HOOKED(&V_inet_pfil_hook))
235 &V_inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, 0, NULL) ||
242 ip = mtod(m, struct ip *); /* m may have changed by pfil hook */
243 dest.s_addr = ip->ip_dst.s_addr;
246 * Destination address changed?
248 if (odest.s_addr != dest.s_addr) {
250 * Is it now for a local address on this host?
252 if (in_localip(dest))
255 * Go on with new destination address
259 if (m->m_flags & M_FASTFWD_OURS) {
261 * ipfw changed it for a local address on this host.
268 * Step 4: decrement TTL and look up route
277 if (ip->ip_ttl <= IPTTLDEC) {
278 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
279 return NULL; /* mbuf already free'd */
283 * Decrement the TTL and incrementally change the IP header checksum.
284 * Don't bother doing this with hw checksum offloading, it's faster
285 * doing it right here.
287 ip->ip_ttl -= IPTTLDEC;
288 if (ip->ip_sum >= (u_int16_t) ~htons(IPTTLDEC << 8))
289 ip->ip_sum -= ~htons(IPTTLDEC << 8);
291 ip->ip_sum += htons(IPTTLDEC << 8);
297 * Next hop forced by pfil(9) hook?
299 if ((m->m_flags & M_IP_NEXTHOP) &&
300 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
302 * Now we will find route to forced destination.
304 dest.s_addr = ((struct sockaddr_in *)
305 (fwd_tag + 1))->sin_addr.s_addr;
306 m_tag_delete(m, fwd_tag);
307 m->m_flags &= ~M_IP_NEXTHOP;
311 * Find route to destination.
313 if (ip_findroute(&nh, dest, m) != 0)
314 return (NULL); /* icmp unreach already sent */
317 * Avoid second route lookup by caching destination.
319 rtdest.s_addr = dest.s_addr;
322 * Step 5: outgoing firewall packet processing
324 if (!PFIL_HOOKED(&V_inet_pfil_hook))
327 if (pfil_run_hooks(&V_inet_pfil_hook, &m, nh.nh_ifp, PFIL_OUT, PFIL_FWD,
328 NULL) || m == NULL) {
335 ip = mtod(m, struct ip *);
336 dest.s_addr = ip->ip_dst.s_addr;
339 * Destination address changed?
341 if (m->m_flags & M_IP_NEXTHOP)
342 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
345 if (odest.s_addr != dest.s_addr || fwd_tag != NULL) {
347 * Is it now for a local address on this host?
349 if (m->m_flags & M_FASTFWD_OURS || in_localip(dest)) {
352 * Return packet for processing by ip_input().
354 m->m_flags |= M_FASTFWD_OURS;
358 * Redo route lookup with new destination address
361 dest.s_addr = ((struct sockaddr_in *)
362 (fwd_tag + 1))->sin_addr.s_addr;
363 m_tag_delete(m, fwd_tag);
364 m->m_flags &= ~M_IP_NEXTHOP;
366 if (dest.s_addr != rtdest.s_addr &&
367 ip_findroute(&nh, dest, m) != 0)
368 return (NULL); /* icmp unreach already sent */
373 * Step 6: send off the packet
375 ip_len = ntohs(ip->ip_len);
376 ip_off = ntohs(ip->ip_off);
378 bzero(&dst, sizeof(dst));
379 dst.sin_family = AF_INET;
380 dst.sin_len = sizeof(dst);
381 dst.sin_addr = nh.nh_addr;
384 * Check if packet fits MTU or if hardware will fragment for us
386 if (ip_len <= nh.nh_mtu) {
388 * Avoid confusing lower layers.
392 * Send off the packet via outgoing interface
394 IP_PROBE(send, NULL, NULL, ip, nh.nh_ifp, ip, NULL);
395 error = (*nh.nh_ifp->if_output)(nh.nh_ifp, m,
396 (struct sockaddr *)&dst, NULL);
399 * Handle EMSGSIZE with icmp reply needfrag for TCP MTU discovery
401 if (ip_off & IP_DF) {
402 IPSTAT_INC(ips_cantfrag);
403 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG,
408 * We have to fragment the packet
410 m->m_pkthdr.csum_flags |= CSUM_IP;
411 if (ip_fragment(ip, &m, nh.nh_mtu,
412 nh.nh_ifp->if_hwassist) != 0)
414 KASSERT(m != NULL, ("null mbuf and no error"));
416 * Send off the fragments via outgoing interface
423 * Avoid confusing lower layers.
427 IP_PROBE(send, NULL, NULL,
428 mtod(m, struct ip *), nh.nh_ifp,
429 mtod(m, struct ip *), NULL);
430 /* XXX: we can use cached route here */
431 error = (*nh.nh_ifp->if_output)(nh.nh_ifp, m,
432 (struct sockaddr *)&dst, NULL);
435 } while ((m = m0) != NULL);
437 /* Reclaim remaining fragments */
438 for (m = m0; m; m = m0) {
443 IPSTAT_INC(ips_fragmented);
448 IPSTAT_INC(ips_odropped);
450 IPSTAT_INC(ips_forward);
451 IPSTAT_INC(ips_fastforward);