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>
113 #define V_ipsendredirects VNET(ipsendredirects)
116 ip_redir_alloc(struct mbuf *m, struct ip *ip, struct in_addr dest,
121 ip_redir_alloc(struct mbuf *m, struct ip *ip, struct in_addr dest,
124 struct sockaddr_in s;
125 struct nhop4_extended nh;
126 struct mbuf *mcopy = m_gethdr(M_NOWAIT, m->m_type);
131 if (fib4_lookup_nh_ext(M_GETFIB(m), dest, 0, 0, &nh) != 0)
134 if (m_dup_pkthdr(mcopy, m, M_NOWAIT) == 0) {
136 * It's probably ok if the pkthdr dup fails (because
137 * the deep copy of the tag chain failed), but for now
138 * be conservative and just discard the copy since
139 * code below may some day want the tags.
144 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
145 mcopy->m_pkthdr.len = mcopy->m_len;
146 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
148 s.sin_len = sizeof(struct sockaddr_in);
149 s.sin_family= AF_INET;
150 s.sin_addr = nh.nh_src;
152 if (((nh.nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
153 struct in_ifaddr *nh_ia = (struct in_ifaddr *)ifaof_ifpforaddr((struct sockaddr *)&s, nh.nh_ifp);
154 u_long src = ntohl(ip->ip_src.s_addr);
156 if (nh_ia != NULL && (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
157 if (nh.nh_flags & NHF_GATEWAY)
158 *addr = nh.nh_addr.s_addr;
160 *addr = ip->ip_dst.s_addr;
170 ip_findroute(struct nhop4_basic *pnh, struct in_addr dest, struct mbuf *m)
173 bzero(pnh, sizeof(*pnh));
174 if (fib4_lookup_nh_basic(M_GETFIB(m), dest, 0, 0, pnh) != 0) {
175 IPSTAT_INC(ips_noroute);
176 IPSTAT_INC(ips_cantforward);
177 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
178 return (EHOSTUNREACH);
181 * Drop blackholed traffic and directed broadcasts.
183 if ((pnh->nh_flags & (NHF_BLACKHOLE | NHF_BROADCAST)) != 0) {
184 IPSTAT_INC(ips_cantforward);
186 return (EHOSTUNREACH);
189 if (pnh->nh_flags & NHF_REJECT) {
190 IPSTAT_INC(ips_cantforward);
191 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
192 return (EHOSTUNREACH);
199 * Try to forward a packet based on the destination address.
200 * This is a fast path optimized for the plain forwarding case.
201 * If the packet is handled (and consumed) here then we return NULL;
202 * otherwise mbuf is returned and the packet should be delivered
203 * to ip_input for full processing.
206 ip_tryforward(struct mbuf *m)
209 struct mbuf *m0 = NULL;
210 struct nhop4_basic nh;
211 struct sockaddr_in dst;
212 struct in_addr dest, odest, rtdest;
213 uint16_t ip_len, ip_off;
215 struct m_tag *fwd_tag = NULL;
216 struct mbuf *mcopy = NULL;
217 struct in_addr redest;
219 * Are we active and forwarding packets?
227 * Is packet dropped by traffic conditioner?
229 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
234 * Only IP packets without options
236 ip = mtod(m, struct ip *);
238 if (ip->ip_hl != (sizeof(struct ip) >> 2)) {
239 if (V_ip_doopts == 1)
241 else if (V_ip_doopts == 2) {
242 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_FILTER_PROHIB,
244 return NULL; /* mbuf already free'd */
246 /* else ignore IP options and continue */
250 * Only unicast IP, not from loopback, no L2 or IP broadcast,
251 * no multicast, no INADDR_ANY
253 * XXX: Probably some of these checks could be direct drop
254 * conditions. However it is not clear whether there are some
255 * hacks or obscure behaviours which make it necessary to
256 * let ip_input handle it. We play safe here and let ip_input
257 * deal with it until it is proven that we can directly drop it.
259 if ((m->m_flags & (M_BCAST|M_MCAST)) ||
260 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) ||
261 ntohl(ip->ip_src.s_addr) == (u_long)INADDR_BROADCAST ||
262 ntohl(ip->ip_dst.s_addr) == (u_long)INADDR_BROADCAST ||
263 IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
264 IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
265 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) ||
266 IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
267 ip->ip_src.s_addr == INADDR_ANY ||
268 ip->ip_dst.s_addr == INADDR_ANY )
272 * Is it for a local address on this host?
274 if (in_localip(ip->ip_dst))
277 IPSTAT_INC(ips_total);
280 * Step 3: incoming packet firewall processing
283 odest.s_addr = dest.s_addr = ip->ip_dst.s_addr;
286 * Run through list of ipfilter hooks for input packets
288 if (!PFIL_HOOKED(&V_inet_pfil_hook))
292 &V_inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, 0, NULL) ||
299 ip = mtod(m, struct ip *); /* m may have changed by pfil hook */
300 dest.s_addr = ip->ip_dst.s_addr;
303 * Destination address changed?
305 if (odest.s_addr != dest.s_addr) {
307 * Is it now for a local address on this host?
309 if (in_localip(dest))
312 * Go on with new destination address
316 if (m->m_flags & M_FASTFWD_OURS) {
318 * ipfw changed it for a local address on this host.
325 * Step 4: decrement TTL and look up route
334 if (ip->ip_ttl <= IPTTLDEC) {
335 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
336 return NULL; /* mbuf already free'd */
340 * Decrement the TTL and incrementally change the IP header checksum.
341 * Don't bother doing this with hw checksum offloading, it's faster
342 * doing it right here.
344 ip->ip_ttl -= IPTTLDEC;
345 if (ip->ip_sum >= (u_int16_t) ~htons(IPTTLDEC << 8))
346 ip->ip_sum -= ~htons(IPTTLDEC << 8);
348 ip->ip_sum += htons(IPTTLDEC << 8);
354 * Next hop forced by pfil(9) hook?
356 if ((m->m_flags & M_IP_NEXTHOP) &&
357 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
359 * Now we will find route to forced destination.
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;
368 * Find route to destination.
370 if (ip_findroute(&nh, dest, m) != 0)
371 return (NULL); /* icmp unreach already sent */
374 * Avoid second route lookup by caching destination.
376 rtdest.s_addr = dest.s_addr;
379 * Step 5: outgoing firewall packet processing
381 if (!PFIL_HOOKED(&V_inet_pfil_hook))
384 if (pfil_run_hooks(&V_inet_pfil_hook, &m, nh.nh_ifp, PFIL_OUT, PFIL_FWD,
385 NULL) || m == NULL) {
392 ip = mtod(m, struct ip *);
393 dest.s_addr = ip->ip_dst.s_addr;
396 * Destination address changed?
398 if (m->m_flags & M_IP_NEXTHOP)
399 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
402 if (odest.s_addr != dest.s_addr || fwd_tag != NULL) {
404 * Is it now for a local address on this host?
406 if (m->m_flags & M_FASTFWD_OURS || in_localip(dest)) {
409 * Return packet for processing by ip_input().
411 m->m_flags |= M_FASTFWD_OURS;
415 * Redo route lookup with new destination address
418 dest.s_addr = ((struct sockaddr_in *)
419 (fwd_tag + 1))->sin_addr.s_addr;
420 m_tag_delete(m, fwd_tag);
421 m->m_flags &= ~M_IP_NEXTHOP;
423 if (dest.s_addr != rtdest.s_addr &&
424 ip_findroute(&nh, dest, m) != 0)
425 return (NULL); /* icmp unreach already sent */
430 * Step 6: send off the packet
432 ip_len = ntohs(ip->ip_len);
433 ip_off = ntohs(ip->ip_off);
435 bzero(&dst, sizeof(dst));
436 dst.sin_family = AF_INET;
437 dst.sin_len = sizeof(dst);
438 dst.sin_addr = nh.nh_addr;
441 * Handle redirect case.
444 if (V_ipsendredirects && (nh.nh_ifp == m->m_pkthdr.rcvif))
445 mcopy = ip_redir_alloc(m, ip, dest, &redest.s_addr);
448 * Check if packet fits MTU or if hardware will fragment for us
450 if (ip_len <= nh.nh_mtu) {
452 * Avoid confusing lower layers.
456 * Send off the packet via outgoing interface
458 IP_PROBE(send, NULL, NULL, ip, nh.nh_ifp, ip, NULL);
459 error = (*nh.nh_ifp->if_output)(nh.nh_ifp, m,
460 (struct sockaddr *)&dst, NULL);
463 * Handle EMSGSIZE with icmp reply needfrag for TCP MTU discovery
465 if (ip_off & IP_DF) {
466 IPSTAT_INC(ips_cantfrag);
467 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG,
472 * We have to fragment the packet
474 m->m_pkthdr.csum_flags |= CSUM_IP;
475 if (ip_fragment(ip, &m, nh.nh_mtu,
476 nh.nh_ifp->if_hwassist) != 0)
478 KASSERT(m != NULL, ("null mbuf and no error"));
480 * Send off the fragments via outgoing interface
487 * Avoid confusing lower layers.
491 IP_PROBE(send, NULL, NULL,
492 mtod(m, struct ip *), nh.nh_ifp,
493 mtod(m, struct ip *), NULL);
494 /* XXX: we can use cached route here */
495 error = (*nh.nh_ifp->if_output)(nh.nh_ifp, m,
496 (struct sockaddr *)&dst, NULL);
499 } while ((m = m0) != NULL);
501 /* Reclaim remaining fragments */
502 for (m = m0; m; m = m0) {
507 IPSTAT_INC(ips_fragmented);
512 IPSTAT_INC(ips_odropped);
514 IPSTAT_INC(ips_forward);
515 IPSTAT_INC(ips_fastforward);
518 /* Send required redirect */
520 icmp_error(mcopy, ICMP_REDIRECT, ICMP_REDIRECT_HOST, redest.s_addr, 0);
521 mcopy = NULL; /* Freed by caller */