2 * Copyright (c) 2003 Andre Oppermann, Internet Business Solutions AG
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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
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
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
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.
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.
50 * Firewalling is fully supported including divert, ipfw fwd and ipfilter
51 * ipnat and address rewrite.
53 * IPSEC is not supported if this host is a tunnel broker. IPSEC is
54 * supported for connections to/from local host.
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.
59 * We take full advantage of hardware support for IP checksum and
60 * fragmentation offloading.
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.
72 * Many thanks to Matt Thomas of NetBSD for basic structure of ip_flow.c which
73 * is being followed here.
76 #include <sys/cdefs.h>
77 __FBSDID("$FreeBSD$");
79 #include "opt_ipstealth.h"
81 #include <sys/param.h>
82 #include <sys/systm.h>
83 #include <sys/kernel.h>
84 #include <sys/malloc.h>
86 #include <sys/protosw.h>
88 #include <sys/socket.h>
89 #include <sys/sysctl.h>
93 #include <net/if_types.h>
94 #include <net/if_var.h>
95 #include <net/if_dl.h>
96 #include <net/route.h>
99 #include <netinet/in.h>
100 #include <netinet/in_fib.h>
101 #include <netinet/in_kdtrace.h>
102 #include <netinet/in_systm.h>
103 #include <netinet/in_var.h>
104 #include <netinet/ip.h>
105 #include <netinet/ip_var.h>
106 #include <netinet/ip_icmp.h>
107 #include <netinet/ip_options.h>
109 #include <machine/in_cksum.h>
112 ip_findroute(struct nhop4_basic *pnh, struct in_addr dest, struct mbuf *m)
115 bzero(pnh, sizeof(*pnh));
116 if (fib4_lookup_nh_basic(M_GETFIB(m), dest, 0, 0, pnh) != 0) {
117 IPSTAT_INC(ips_noroute);
118 IPSTAT_INC(ips_cantforward);
119 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
120 return (EHOSTUNREACH);
123 * Drop blackholed traffic and directed broadcasts.
125 if ((pnh->nh_flags & (NHF_BLACKHOLE | NHF_BROADCAST)) != 0) {
126 IPSTAT_INC(ips_cantforward);
128 return (EHOSTUNREACH);
131 if (pnh->nh_flags & NHF_REJECT) {
132 IPSTAT_INC(ips_cantforward);
133 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
134 return (EHOSTUNREACH);
141 * Try to forward a packet based on the destination address.
142 * This is a fast path optimized for the plain forwarding case.
143 * If the packet is handled (and consumed) here then we return NULL;
144 * otherwise mbuf is returned and the packet should be delivered
145 * to ip_input for full processing.
148 ip_tryforward(struct mbuf *m)
151 struct mbuf *m0 = NULL;
152 struct nhop4_basic nh;
153 struct sockaddr_in dst;
154 struct in_addr odest, dest;
155 uint16_t ip_len, ip_off;
157 struct m_tag *fwd_tag = NULL;
160 * Are we active and forwarding packets?
168 * Is packet dropped by traffic conditioner?
170 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
175 * Only IP packets without options
177 ip = mtod(m, struct ip *);
179 if (ip->ip_hl != (sizeof(struct ip) >> 2)) {
180 if (V_ip_doopts == 1)
182 else if (V_ip_doopts == 2) {
183 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_FILTER_PROHIB,
185 return NULL; /* mbuf already free'd */
187 /* else ignore IP options and continue */
191 * Only unicast IP, not from loopback, no L2 or IP broadcast,
192 * no multicast, no INADDR_ANY
194 * XXX: Probably some of these checks could be direct drop
195 * conditions. However it is not clear whether there are some
196 * hacks or obscure behaviours which make it necessary to
197 * let ip_input handle it. We play safe here and let ip_input
198 * deal with it until it is proven that we can directly drop it.
200 if ((m->m_flags & (M_BCAST|M_MCAST)) ||
201 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) ||
202 ntohl(ip->ip_src.s_addr) == (u_long)INADDR_BROADCAST ||
203 ntohl(ip->ip_dst.s_addr) == (u_long)INADDR_BROADCAST ||
204 IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
205 IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
206 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) ||
207 IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
208 ip->ip_src.s_addr == INADDR_ANY ||
209 ip->ip_dst.s_addr == INADDR_ANY )
213 * Is it for a local address on this host?
215 if (in_localip(ip->ip_dst))
218 IPSTAT_INC(ips_total);
221 * Step 3: incoming packet firewall processing
224 odest.s_addr = dest.s_addr = ip->ip_dst.s_addr;
227 * Run through list of ipfilter hooks for input packets
229 if (!PFIL_HOOKED(&V_inet_pfil_hook))
233 &V_inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL) ||
240 ip = mtod(m, struct ip *); /* m may have changed by pfil hook */
241 dest.s_addr = ip->ip_dst.s_addr;
244 * Destination address changed?
246 if (odest.s_addr != dest.s_addr) {
248 * Is it now for a local address on this host?
250 if (in_localip(dest))
253 * Go on with new destination address
257 if (m->m_flags & M_FASTFWD_OURS) {
259 * ipfw changed it for a local address on this host.
266 * Step 4: decrement TTL and look up route
275 if (ip->ip_ttl <= IPTTLDEC) {
276 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
277 return NULL; /* mbuf already free'd */
281 * Decrement the TTL and incrementally change the IP header checksum.
282 * Don't bother doing this with hw checksum offloading, it's faster
283 * doing it right here.
285 ip->ip_ttl -= IPTTLDEC;
286 if (ip->ip_sum >= (u_int16_t) ~htons(IPTTLDEC << 8))
287 ip->ip_sum -= ~htons(IPTTLDEC << 8);
289 ip->ip_sum += htons(IPTTLDEC << 8);
295 * Find route to destination.
297 if (ip_findroute(&nh, dest, m) != 0)
298 return (NULL); /* icmp unreach already sent */
301 * Step 5: outgoing firewall packet processing
303 if (!PFIL_HOOKED(&V_inet_pfil_hook))
306 if (pfil_run_hooks(&V_inet_pfil_hook, &m, nh.nh_ifp, PFIL_OUT, NULL) ||
314 ip = mtod(m, struct ip *);
315 dest.s_addr = ip->ip_dst.s_addr;
318 * Destination address changed?
320 if (m->m_flags & M_IP_NEXTHOP)
321 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
322 if (odest.s_addr != dest.s_addr || fwd_tag != NULL) {
324 * Is it now for a local address on this host?
326 if (m->m_flags & M_FASTFWD_OURS || in_localip(dest)) {
329 * Return packet for processing by ip_input().
331 m->m_flags |= M_FASTFWD_OURS;
335 * Redo route lookup with new destination address
338 dest.s_addr = ((struct sockaddr_in *)
339 (fwd_tag + 1))->sin_addr.s_addr;
340 m_tag_delete(m, fwd_tag);
341 m->m_flags &= ~M_IP_NEXTHOP;
343 if (ip_findroute(&nh, dest, m) != 0)
344 return (NULL); /* icmp unreach already sent */
349 * Step 6: send off the packet
351 ip_len = ntohs(ip->ip_len);
352 ip_off = ntohs(ip->ip_off);
354 bzero(&dst, sizeof(dst));
355 dst.sin_family = AF_INET;
356 dst.sin_len = sizeof(dst);
357 dst.sin_addr = nh.nh_addr;
360 * Check if packet fits MTU or if hardware will fragment for us
362 if (ip_len <= nh.nh_mtu) {
364 * Avoid confusing lower layers.
368 * Send off the packet via outgoing interface
370 IP_PROBE(send, NULL, NULL, ip, nh.nh_ifp, ip, NULL);
371 error = (*nh.nh_ifp->if_output)(nh.nh_ifp, m,
372 (struct sockaddr *)&dst, NULL);
375 * Handle EMSGSIZE with icmp reply needfrag for TCP MTU discovery
377 if (ip_off & IP_DF) {
378 IPSTAT_INC(ips_cantfrag);
379 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG,
384 * We have to fragment the packet
386 m->m_pkthdr.csum_flags |= CSUM_IP;
387 if (ip_fragment(ip, &m, nh.nh_mtu,
388 nh.nh_ifp->if_hwassist) != 0)
390 KASSERT(m != NULL, ("null mbuf and no error"));
392 * Send off the fragments via outgoing interface
399 * Avoid confusing lower layers.
403 IP_PROBE(send, NULL, NULL,
404 mtod(m, struct ip *), nh.nh_ifp,
405 mtod(m, struct ip *), NULL);
406 /* XXX: we can use cached route here */
407 error = (*nh.nh_ifp->if_output)(nh.nh_ifp, m,
408 (struct sockaddr *)&dst, NULL);
411 } while ((m = m0) != NULL);
413 /* Reclaim remaining fragments */
414 for (m = m0; m; m = m0) {
419 IPSTAT_INC(ips_fragmented);
424 IPSTAT_INC(ips_odropped);
426 IPSTAT_INC(ips_forward);
427 IPSTAT_INC(ips_fastforward);