ip_fastfwd.c revision 133497
1122702Sandre/*
2122702Sandre * Copyright (c) 2003 Andre Oppermann, Internet Business Solutions AG
3122702Sandre * All rights reserved.
4122702Sandre *
5122702Sandre * Redistribution and use in source and binary forms, with or without
6122702Sandre * modification, are permitted provided that the following conditions
7122702Sandre * are met:
8122702Sandre * 1. Redistributions of source code must retain the above copyright
9122702Sandre *    notice, this list of conditions and the following disclaimer.
10122702Sandre * 2. Redistributions in binary form must reproduce the above copyright
11122702Sandre *    notice, this list of conditions and the following disclaimer in the
12122702Sandre *    documentation and/or other materials provided with the distribution.
13122702Sandre * 3. The name of the author may not be used to endorse or promote
14122702Sandre *    products derived from this software without specific prior written
15122702Sandre *    permission.
16122702Sandre *
17122702Sandre * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18122702Sandre * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19122702Sandre * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20122702Sandre * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21122702Sandre * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22122702Sandre * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23122702Sandre * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24122702Sandre * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25122702Sandre * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26122702Sandre * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27122702Sandre * SUCH DAMAGE.
28122702Sandre *
29122702Sandre * $FreeBSD: head/sys/netinet/ip_fastfwd.c 133497 2004-08-11 12:32:10Z andre $
30122702Sandre */
31122702Sandre
32122702Sandre/*
33122702Sandre * ip_fastforward gets its speed from processing the forwarded packet to
34122702Sandre * completion (if_output on the other side) without any queues or netisr's.
35122702Sandre * The receiving interface DMAs the packet into memory, the upper half of
36122702Sandre * driver calls ip_fastforward, we do our routing table lookup and directly
37122702Sandre * send it off to the outgoing interface which DMAs the packet to the
38122702Sandre * network card. The only part of the packet we touch with the CPU is the
39122759Sandre * IP header (unless there are complex firewall rules touching other parts
40122759Sandre * of the packet, but that is up to you). We are essentially limited by bus
41122759Sandre * bandwidth and how fast the network card/driver can set up receives and
42122759Sandre * transmits.
43122702Sandre *
44122702Sandre * We handle basic errors, ip header errors, checksum errors,
45122702Sandre * destination unreachable, fragmentation and fragmentation needed and
46122702Sandre * report them via icmp to the sender.
47122702Sandre *
48122702Sandre * Else if something is not pure IPv4 unicast forwarding we fall back to
49122702Sandre * the normal ip_input processing path. We should only be called from
50122702Sandre * interfaces connected to the outside world.
51122702Sandre *
52122702Sandre * Firewalling is fully supported including divert, ipfw fwd and ipfilter
53122702Sandre * ipnat and address rewrite.
54122702Sandre *
55122702Sandre * IPSEC is not supported if this host is a tunnel broker. IPSEC is
56122702Sandre * supported for connections to/from local host.
57122702Sandre *
58122702Sandre * We try to do the least expensive (in CPU ops) checks and operations
59122702Sandre * first to catch junk with as little overhead as possible.
60122702Sandre *
61122702Sandre * We take full advantage of hardware support for ip checksum and
62122702Sandre * fragmentation offloading.
63122702Sandre *
64122702Sandre * We don't do ICMP redirect in the fast forwarding path. I have had my own
65122702Sandre * cases where two core routers with Zebra routing suite would send millions
66122702Sandre * ICMP redirects to connected hosts if the router to dest was not the default
67122702Sandre * gateway. In one case it was filling the routing table of a host with close
68122702Sandre * 300'000 cloned redirect entries until it ran out of kernel memory. However
69122702Sandre * the networking code proved very robust and it didn't crash or went ill
70122702Sandre * otherwise.
71122702Sandre */
72122702Sandre
73122702Sandre/*
74122702Sandre * Many thanks to Matt Thomas of NetBSD for basic structure of ip_flow.c which
75122702Sandre * is being followed here.
76122702Sandre */
77122702Sandre
78122702Sandre#include "opt_ipfw.h"
79122702Sandre#include "opt_ipdn.h"
80122702Sandre#include "opt_ipdivert.h"
81122702Sandre#include "opt_ipfilter.h"
82122702Sandre#include "opt_ipstealth.h"
83122702Sandre#include "opt_pfil_hooks.h"
84122702Sandre
85122702Sandre#include <sys/param.h>
86122702Sandre#include <sys/systm.h>
87122702Sandre#include <sys/kernel.h>
88122702Sandre#include <sys/malloc.h>
89122702Sandre#include <sys/mbuf.h>
90122702Sandre#include <sys/protosw.h>
91122702Sandre#include <sys/socket.h>
92122702Sandre#include <sys/sysctl.h>
93122702Sandre
94122702Sandre#include <net/pfil.h>
95122702Sandre#include <net/if.h>
96122702Sandre#include <net/if_types.h>
97122702Sandre#include <net/if_var.h>
98122702Sandre#include <net/if_dl.h>
99122702Sandre#include <net/route.h>
100122702Sandre
101122702Sandre#include <netinet/in.h>
102122702Sandre#include <netinet/in_systm.h>
103122702Sandre#include <netinet/in_var.h>
104122702Sandre#include <netinet/ip.h>
105122702Sandre#include <netinet/ip_var.h>
106122702Sandre#include <netinet/ip_icmp.h>
107122702Sandre
108122702Sandre#include <machine/in_cksum.h>
109122702Sandre
110122702Sandre#include <netinet/ip_fw.h>
111126239Smlaier#include <netinet/ip_divert.h>
112122702Sandre#include <netinet/ip_dummynet.h>
113122702Sandre
114122702Sandrestatic int ipfastforward_active = 0;
115122702SandreSYSCTL_INT(_net_inet_ip, OID_AUTO, fastforwarding, CTLFLAG_RW,
116122702Sandre    &ipfastforward_active, 0, "Enable fast IP forwarding");
117122702Sandre
118128872Sandrestatic struct sockaddr_in *
119133497Sandreip_findroute(struct route *ro, struct in_addr dest, struct mbuf *m)
120128872Sandre{
121128872Sandre	struct sockaddr_in *dst;
122128872Sandre	struct rtentry *rt;
123128872Sandre
124128872Sandre	/*
125128872Sandre	 * Find route to destination.
126128872Sandre	 */
127128872Sandre	bzero(ro, sizeof(*ro));
128128872Sandre	dst = (struct sockaddr_in *)&ro->ro_dst;
129128872Sandre	dst->sin_family = AF_INET;
130128872Sandre	dst->sin_len = sizeof(*dst);
131133497Sandre	dst->sin_addr.s_addr = dest.s_addr;
132128872Sandre	rtalloc_ign(ro, RTF_CLONING);
133128872Sandre
134128872Sandre	/*
135128872Sandre	 * Route there and interface still up?
136128872Sandre	 */
137128872Sandre	rt = ro->ro_rt;
138128872Sandre	if (rt && (rt->rt_flags & RTF_UP) &&
139128872Sandre	    (rt->rt_ifp->if_flags & IFF_UP) &&
140128872Sandre	    (rt->rt_ifp->if_flags & IFF_RUNNING)) {
141128872Sandre		if (rt->rt_flags & RTF_GATEWAY)
142128872Sandre			dst = (struct sockaddr_in *)rt->rt_gateway;
143128872Sandre	} else {
144128872Sandre		ipstat.ips_noroute++;
145128872Sandre		ipstat.ips_cantforward++;
146128872Sandre		if (rt)
147128872Sandre			RTFREE(rt);
148128872Sandre		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, NULL);
149128872Sandre		return NULL;
150128872Sandre	}
151128872Sandre	return dst;
152128872Sandre}
153128872Sandre
154122702Sandre/*
155122702Sandre * Try to forward a packet based on the destination address.
156122702Sandre * This is a fast path optimized for the plain forwarding case.
157122702Sandre * If the packet is handled (and consumed) here then we return 1;
158122702Sandre * otherwise 0 is returned and the packet should be delivered
159122702Sandre * to ip_input for full processing.
160122702Sandre */
161122702Sandreint
162122702Sandreip_fastforward(struct mbuf *m)
163122702Sandre{
164122702Sandre	struct ip *ip;
165122702Sandre	struct mbuf *m0 = NULL;
166122702Sandre#ifdef IPDIVERT
167122702Sandre	struct ip *tip;
168126239Smlaier	struct mbuf *clone = NULL;
169122702Sandre#endif
170122702Sandre	struct route ro;
171122702Sandre	struct sockaddr_in *dst = NULL;
172122702Sandre	struct in_ifaddr *ia = NULL;
173122702Sandre	struct ifaddr *ifa = NULL;
174128872Sandre	struct ifnet *ifp;
175122702Sandre	struct ip_fw_args args;
176133497Sandre	struct in_addr odest, dest;
177128872Sandre	u_short sum, ip_len;
178122702Sandre	int error = 0;
179122702Sandre	int hlen, ipfw, mtu;
180122702Sandre
181122702Sandre	/*
182122702Sandre	 * Are we active and forwarding packets?
183122702Sandre	 */
184122702Sandre	if (!ipfastforward_active || !ipforwarding)
185122702Sandre		return 0;
186122702Sandre
187122702Sandre	M_ASSERTVALID(m);
188122702Sandre	M_ASSERTPKTHDR(m);
189122702Sandre
190128872Sandre	ro.ro_rt = NULL;
191128872Sandre
192122702Sandre	/*
193122702Sandre	 * Step 1: check for packet drop conditions (and sanity checks)
194122702Sandre	 */
195122702Sandre
196122702Sandre	/*
197122702Sandre	 * Is entire packet big enough?
198122702Sandre	 */
199122702Sandre	if (m->m_pkthdr.len < sizeof(struct ip)) {
200122702Sandre		ipstat.ips_tooshort++;
201122702Sandre		goto drop;
202122702Sandre	}
203122702Sandre
204122702Sandre	/*
205122702Sandre	 * Is first mbuf large enough for ip header and is header present?
206122702Sandre	 */
207122702Sandre	if (m->m_len < sizeof (struct ip) &&
208122702Sandre	   (m = m_pullup(m, sizeof (struct ip))) == 0) {
209122702Sandre		ipstat.ips_toosmall++;
210122702Sandre		goto drop;
211122702Sandre	}
212122702Sandre
213122702Sandre	ip = mtod(m, struct ip *);
214122702Sandre
215122702Sandre	/*
216122702Sandre	 * Is it IPv4?
217122702Sandre	 */
218122702Sandre	if (ip->ip_v != IPVERSION) {
219122702Sandre		ipstat.ips_badvers++;
220122702Sandre		goto drop;
221122702Sandre	}
222122702Sandre
223122702Sandre	/*
224122702Sandre	 * Is IP header length correct and is it in first mbuf?
225122702Sandre	 */
226122702Sandre	hlen = ip->ip_hl << 2;
227122702Sandre	if (hlen < sizeof(struct ip)) {	/* minimum header length */
228122702Sandre		ipstat.ips_badlen++;
229122702Sandre		goto drop;
230122702Sandre	}
231122702Sandre	if (hlen > m->m_len) {
232122702Sandre		if ((m = m_pullup(m, hlen)) == 0) {
233122702Sandre			ipstat.ips_badhlen++;
234122702Sandre			goto drop;
235122702Sandre		}
236122702Sandre		ip = mtod(m, struct ip *);
237122702Sandre	}
238122702Sandre
239122702Sandre	/*
240122702Sandre	 * Checksum correct?
241122702Sandre	 */
242122702Sandre	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED)
243122702Sandre		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
244122702Sandre	else {
245122702Sandre		if (hlen == sizeof(struct ip))
246122702Sandre			sum = in_cksum_hdr(ip);
247122702Sandre		else
248122702Sandre			sum = in_cksum(m, hlen);
249122702Sandre	}
250122702Sandre	if (sum) {
251122702Sandre		ipstat.ips_badsum++;
252122702Sandre		goto drop;
253122702Sandre	}
254122702Sandre	m->m_pkthdr.csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
255122702Sandre
256128872Sandre	ip_len = ntohs(ip->ip_len);
257122702Sandre
258122702Sandre	/*
259122702Sandre	 * Is IP length longer than packet we have got?
260122702Sandre	 */
261128872Sandre	if (m->m_pkthdr.len < ip_len) {
262122702Sandre		ipstat.ips_tooshort++;
263122702Sandre		goto drop;
264122702Sandre	}
265122702Sandre
266122702Sandre	/*
267122702Sandre	 * Is packet longer than IP header tells us? If yes, truncate packet.
268122702Sandre	 */
269128872Sandre	if (m->m_pkthdr.len > ip_len) {
270122702Sandre		if (m->m_len == m->m_pkthdr.len) {
271128872Sandre			m->m_len = ip_len;
272128872Sandre			m->m_pkthdr.len = ip_len;
273122702Sandre		} else
274128872Sandre			m_adj(m, ip_len - m->m_pkthdr.len);
275122702Sandre	}
276122702Sandre
277122702Sandre	/*
278122702Sandre	 * Is packet from or to 127/8?
279122702Sandre	 */
280122702Sandre	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
281122702Sandre	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
282122702Sandre		ipstat.ips_badaddr++;
283122702Sandre		goto drop;
284122702Sandre	}
285122702Sandre
286133480Sandre#ifdef ALTQ
287122702Sandre	/*
288133480Sandre	 * Is packet dropped by traffic conditioner?
289133480Sandre	 */
290133480Sandre	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
291133480Sandre		return 1;
292133480Sandre#endif
293133480Sandre
294133480Sandre	/*
295122702Sandre	 * Step 2: fallback conditions to normal ip_input path processing
296122702Sandre	 */
297122702Sandre
298122702Sandre	/*
299122702Sandre	 * Only IP packets without options
300122702Sandre	 */
301129017Sandre	if (ip->ip_hl != (sizeof(struct ip) >> 2)) {
302129017Sandre		if (ip_doopts == 1)
303129017Sandre			return 0;
304129017Sandre		else if (ip_doopts == 2) {
305129017Sandre			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_FILTER_PROHIB,
306129017Sandre				0, NULL);
307129017Sandre			return 1;
308129017Sandre		}
309129017Sandre		/* else ignore IP options and continue */
310129017Sandre	}
311122702Sandre
312122702Sandre	/*
313122702Sandre	 * Only unicast IP, not from loopback, no L2 or IP broadcast,
314122702Sandre	 * no multicast, no INADDR_ANY
315122702Sandre	 *
316122702Sandre	 * XXX: Probably some of these checks could be direct drop
317122702Sandre	 * conditions.  However it is not clear whether there are some
318122702Sandre	 * hacks or obscure behaviours which make it neccessary to
319122702Sandre	 * let ip_input handle it.  We play safe here and let ip_input
320122702Sandre	 * deal with it until it is proven that we can directly drop it.
321122702Sandre	 */
322122702Sandre	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) ||
323122702Sandre	    ntohl(ip->ip_src.s_addr) == (u_long)INADDR_BROADCAST ||
324122702Sandre	    ntohl(ip->ip_dst.s_addr) == (u_long)INADDR_BROADCAST ||
325122702Sandre	    IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
326122702Sandre	    IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
327122702Sandre	    ip->ip_dst.s_addr == INADDR_ANY )
328128872Sandre		return 0;
329122702Sandre
330122702Sandre	/*
331122702Sandre	 * Is it for a local address on this host?
332122702Sandre	 */
333133497Sandre	if (in_localip(ip->ip_dst))
334133497Sandre		return 0;
335122702Sandre
336122702Sandre	/*
337122702Sandre	 * Or is it for a local IP broadcast address on this host?
338122702Sandre	 */
339133482Sandre	if ((m->m_flags & M_BCAST) &&
340133482Sandre	    (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST)) {
341122702Sandre	        TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
342122702Sandre			if (ifa->ifa_addr->sa_family != AF_INET)
343122702Sandre				continue;
344122702Sandre			ia = ifatoia(ifa);
345122702Sandre			if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
346128872Sandre				return 0;
347122702Sandre			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
348122702Sandre			    ip->ip_dst.s_addr)
349128872Sandre				return 0;
350122702Sandre		}
351122702Sandre	}
352122702Sandre	ipstat.ips_total++;
353122702Sandre
354122702Sandre	/*
355122702Sandre	 * Step 3: incoming packet firewall processing
356122702Sandre	 */
357122702Sandre
358128872Sandre	/*
359128872Sandre	 * Convert to host representation
360128872Sandre	 */
361128872Sandre	ip->ip_len = ntohs(ip->ip_len);
362128872Sandre	ip->ip_off = ntohs(ip->ip_off);
363128872Sandre
364133497Sandre	odest.s_addr = dest.s_addr = ip->ip_dst.s_addr;
365122702Sandre#ifdef PFIL_HOOKS
366122702Sandre	/*
367122702Sandre	 * Run through list of ipfilter hooks for input packets
368122702Sandre	 */
369122702Sandre	if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN) ||
370122702Sandre	    m == NULL)
371122702Sandre		return 1;
372122702Sandre
373122702Sandre	M_ASSERTVALID(m);
374122702Sandre	M_ASSERTPKTHDR(m);
375122702Sandre
376122702Sandre	ip = mtod(m, struct ip *);	/* m may have changed by pfil hook */
377133497Sandre	dest.s_addr = ip->ip_dst.s_addr;
378122702Sandre#endif
379122702Sandre
380122702Sandre	/*
381122702Sandre	 * Run through ipfw for input packets
382122702Sandre	 */
383122702Sandre	if (fw_enable && IPFW_LOADED) {
384122702Sandre		bzero(&args, sizeof(args));
385122702Sandre		args.m = m;
386122702Sandre
387122702Sandre		ipfw = ip_fw_chk_ptr(&args);
388122702Sandre		m = args.m;
389122702Sandre
390122702Sandre		M_ASSERTVALID(m);
391122702Sandre		M_ASSERTPKTHDR(m);
392122702Sandre
393122702Sandre		/*
394122702Sandre		 * Packet denied, drop it
395122702Sandre		 */
396122702Sandre		if ((ipfw & IP_FW_PORT_DENY_FLAG) || m == NULL)
397122702Sandre			goto drop;
398122702Sandre		/*
399122702Sandre		 * Send packet to the appropriate pipe
400122702Sandre		 */
401122702Sandre		if (DUMMYNET_LOADED && (ipfw & IP_FW_PORT_DYNT_FLAG) != 0) {
402122702Sandre			ip_dn_io_ptr(m, ipfw & 0xffff, DN_TO_IP_IN, &args);
403122702Sandre			return 1;
404122702Sandre		}
405122702Sandre#ifdef IPDIVERT
406122702Sandre		/*
407122702Sandre		 * Divert packet
408122702Sandre		 */
409122702Sandre		if (ipfw != 0 && (ipfw & IP_FW_PORT_DYNT_FLAG) == 0) {
410122702Sandre			/*
411122702Sandre			 * See if this is a fragment
412122702Sandre			 */
413126239Smlaier			if (ip->ip_off & (IP_MF | IP_OFFMASK))
414122702Sandre				goto droptoours;
415122702Sandre			/*
416122702Sandre			 * Tee packet
417122702Sandre			 */
418122702Sandre			if ((ipfw & IP_FW_PORT_TEE_FLAG) != 0)
419126239Smlaier				clone = divert_clone(m);
420122702Sandre			else
421126239Smlaier				clone = m;
422126239Smlaier			if (clone == NULL)
423122702Sandre				goto passin;
424122702Sandre
425122702Sandre			/*
426122702Sandre			 * Delayed checksums are not compatible
427122702Sandre			 */
428126239Smlaier			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
429126239Smlaier				in_delayed_cksum(m);
430126239Smlaier				m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
431122702Sandre			}
432122702Sandre			/*
433122702Sandre			 * Restore packet header fields to original values
434122702Sandre			 */
435126239Smlaier			tip = mtod(m, struct ip *);
436122702Sandre			tip->ip_len = htons(tip->ip_len);
437122702Sandre			tip->ip_off = htons(tip->ip_off);
438122702Sandre			/*
439122702Sandre			 * Deliver packet to divert input routine
440122702Sandre			 */
441126239Smlaier			divert_packet(m, 0);
442122702Sandre			/*
443122702Sandre			 * If this was not tee, we are done
444122702Sandre			 */
445126239Smlaier			m = clone;
446122702Sandre			if ((ipfw & IP_FW_PORT_TEE_FLAG) == 0)
447122702Sandre				return 1;
448122702Sandre			/* Continue if it was tee */
449122702Sandre			goto passin;
450122702Sandre		}
451122702Sandre#endif
452122702Sandre		if (ipfw == 0 && args.next_hop != NULL) {
453133497Sandre			dest.s_addr = args.next_hop->sin_addr.s_addr;
454122702Sandre			goto passin;
455122702Sandre		}
456122702Sandre		/*
457122702Sandre		 * Let through or not?
458122702Sandre		 */
459122702Sandre		if (ipfw != 0)
460122702Sandre			goto drop;
461122702Sandre	}
462122702Sandrepassin:
463122702Sandre	ip = mtod(m, struct ip *);	/* if m changed during fw processing */
464122702Sandre
465122702Sandre	/*
466122702Sandre	 * Destination address changed?
467122702Sandre	 */
468133497Sandre	if (odest.s_addr != dest.s_addr) {
469122702Sandre		/*
470122702Sandre		 * Is it now for a local address on this host?
471122702Sandre		 */
472133497Sandre		if (in_localip(dest))
473133497Sandre			goto forwardlocal;
474122702Sandre		/*
475122702Sandre		 * Go on with new destination address
476122702Sandre		 */
477122702Sandre	}
478122702Sandre
479122702Sandre	/*
480122702Sandre	 * Step 4: decrement TTL and look up route
481122702Sandre	 */
482122702Sandre
483122702Sandre	/*
484122702Sandre	 * Check TTL
485122702Sandre	 */
486122702Sandre#ifdef IPSTEALTH
487122702Sandre	if (!ipstealth) {
488122702Sandre#endif
489122702Sandre	if (ip->ip_ttl <= IPTTLDEC) {
490123740Speter		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, NULL);
491122702Sandre		return 1;
492122702Sandre	}
493122702Sandre
494122702Sandre	/*
495122702Sandre	 * Decrement the TTL and incrementally change the checksum.
496122702Sandre	 * Don't bother doing this with hw checksum offloading.
497122702Sandre	 */
498122702Sandre	ip->ip_ttl -= IPTTLDEC;
499122702Sandre	if (ip->ip_sum >= (u_int16_t) ~htons(IPTTLDEC << 8))
500122702Sandre		ip->ip_sum -= ~htons(IPTTLDEC << 8);
501122702Sandre	else
502122702Sandre		ip->ip_sum += htons(IPTTLDEC << 8);
503122702Sandre#ifdef IPSTEALTH
504122702Sandre	}
505122702Sandre#endif
506122702Sandre
507122702Sandre	/*
508122702Sandre	 * Find route to destination.
509122702Sandre	 */
510128872Sandre	if ((dst = ip_findroute(&ro, dest, m)) == NULL)
511128872Sandre		return 1;	/* icmp unreach already sent */
512128872Sandre	ifp = ro.ro_rt->rt_ifp;
513122702Sandre
514122702Sandre	/*
515122702Sandre	 * Step 5: outgoing firewall packet processing
516122702Sandre	 */
517122702Sandre
518122702Sandre#ifdef PFIL_HOOKS
519122702Sandre	/*
520122702Sandre	 * Run through list of hooks for output packets.
521122702Sandre	 */
522122702Sandre	if (pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT) || m == NULL) {
523128872Sandre		goto consumed;
524122702Sandre	}
525122702Sandre
526122702Sandre	M_ASSERTVALID(m);
527122702Sandre	M_ASSERTPKTHDR(m);
528122702Sandre
529122702Sandre	ip = mtod(m, struct ip *);
530133497Sandre	dest.s_addr = ip->ip_dst.s_addr;
531122702Sandre#endif
532122702Sandre	if (fw_enable && IPFW_LOADED && !args.next_hop) {
533122759Sandre		bzero(&args, sizeof(args));
534122702Sandre		args.m = m;
535122702Sandre		args.oif = ifp;
536122702Sandre
537122702Sandre		ipfw = ip_fw_chk_ptr(&args);
538122702Sandre		m = args.m;
539122702Sandre
540122702Sandre		M_ASSERTVALID(m);
541122702Sandre		M_ASSERTPKTHDR(m);
542122702Sandre
543128872Sandre		if ((ipfw & IP_FW_PORT_DENY_FLAG) || m == NULL)
544122702Sandre			goto drop;
545128872Sandre
546122702Sandre		if (DUMMYNET_LOADED && (ipfw & IP_FW_PORT_DYNT_FLAG) != 0) {
547122702Sandre			/*
548122702Sandre			 * XXX note: if the ifp or rt entry are deleted
549122702Sandre			 * while a pkt is in dummynet, we are in trouble!
550122702Sandre			 */
551122702Sandre			args.ro = &ro;		/* dummynet does not save it */
552122702Sandre			args.dst = dst;
553122702Sandre
554122702Sandre			ip_dn_io_ptr(m, ipfw & 0xffff, DN_TO_IP_OUT, &args);
555128872Sandre			goto consumed;
556122702Sandre		}
557122702Sandre#ifdef IPDIVERT
558122702Sandre		if (ipfw != 0 && (ipfw & IP_FW_PORT_DYNT_FLAG) == 0) {
559122702Sandre			/*
560122702Sandre			 * See if this is a fragment
561122702Sandre			 */
562126239Smlaier			if (ip->ip_off & (IP_MF | IP_OFFMASK))
563122702Sandre				goto droptoours;
564122702Sandre			/*
565122702Sandre			 * Tee packet
566122702Sandre			 */
567122702Sandre			if ((ipfw & IP_FW_PORT_TEE_FLAG) != 0)
568126239Smlaier				clone = divert_clone(m);
569122702Sandre			else
570126239Smlaier				clone = m;
571126239Smlaier			if (clone == NULL)
572122702Sandre				goto passout;
573122702Sandre
574122702Sandre			/*
575122702Sandre			 * Delayed checksums are not compatible with divert
576122702Sandre			 */
577126239Smlaier			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
578126239Smlaier				in_delayed_cksum(m);
579126239Smlaier				m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
580122702Sandre			}
581122702Sandre			/*
582122702Sandre			 * Restore packet header fields to original values
583122702Sandre			 */
584126239Smlaier			tip = mtod(m, struct ip *);
585122702Sandre			tip->ip_len = htons(tip->ip_len);
586122702Sandre			tip->ip_off = htons(tip->ip_off);
587122702Sandre			/*
588122702Sandre			 * Deliver packet to divert input routine
589122702Sandre			 */
590126239Smlaier			divert_packet(m, 0);
591122702Sandre			/*
592122702Sandre			 * If this was not tee, we are done
593122702Sandre			 */
594126239Smlaier			m = clone;
595122702Sandre			if ((ipfw & IP_FW_PORT_TEE_FLAG) == 0) {
596128872Sandre				goto consumed;
597122702Sandre			}
598122702Sandre			/* Continue if it was tee */
599122702Sandre			goto passout;
600122702Sandre		}
601122702Sandre#endif
602122702Sandre		if (ipfw == 0 && args.next_hop != NULL) {
603133497Sandre			dest.s_addr = args.next_hop->sin_addr.s_addr;
604122702Sandre			goto passout;
605122702Sandre		}
606122702Sandre		/*
607122702Sandre		 * Let through or not?
608122702Sandre		 */
609122702Sandre		if (ipfw != 0)
610122702Sandre			goto drop;
611122702Sandre	}
612122702Sandrepassout:
613122702Sandre	ip = mtod(m, struct ip *);
614122702Sandre
615122702Sandre	/*
616122702Sandre	 * Destination address changed?
617122702Sandre	 */
618133497Sandre	if (odest.s_addr != dest.s_addr) {
619122702Sandre		/*
620122702Sandre		 * Is it now for a local address on this host?
621122702Sandre		 */
622133497Sandre		if (in_localip(dest)) {
623122702Sandreforwardlocal:
624133497Sandre			if (args.next_hop) {
625133497Sandre				struct m_tag *mtag = m_tag_get(
626133497Sandre				    PACKET_TAG_IPFORWARD,
627133497Sandre				    sizeof(struct sockaddr_in *),
628133497Sandre				    M_NOWAIT);
629133497Sandre				if (mtag == NULL) {
630133497Sandre					goto drop;
631122702Sandre				}
632133497Sandre				*(struct sockaddr_in **)(mtag+1) =
633133497Sandre				    args.next_hop;
634133497Sandre				m_tag_prepend(m, mtag);
635133497Sandre			}
636122702Sandre#ifdef IPDIVERT
637122702Sandredroptoours:	/* Used for DIVERT */
638122702Sandre#endif
639133497Sandre			/* for ip_input */
640133497Sandre			m->m_flags |= M_FASTFWD_OURS;
641122702Sandre
642133497Sandre			/* ip still points to the real packet */
643133497Sandre			ip->ip_len = htons(ip->ip_len);
644133497Sandre			ip->ip_off = htons(ip->ip_off);
645122702Sandre
646133497Sandre			/*
647133497Sandre			 * Return packet for processing by ip_input
648133497Sandre			 */
649133497Sandre			if (ro.ro_rt)
650133497Sandre				RTFREE(ro.ro_rt);
651133497Sandre			return 0;
652122702Sandre		}
653122702Sandre		/*
654122702Sandre		 * Redo route lookup with new destination address
655122702Sandre		 */
656122702Sandre		RTFREE(ro.ro_rt);
657128872Sandre		if ((dst = ip_findroute(&ro, dest, m)) == NULL)
658128872Sandre			return 1;	/* icmp unreach already sent */
659128872Sandre		ifp = ro.ro_rt->rt_ifp;
660122702Sandre	}
661122702Sandre
662122702Sandre	/*
663122702Sandre	 * Step 6: send off the packet
664122702Sandre	 */
665122702Sandre
666122702Sandre	/*
667128872Sandre	 * Check if route is dampned (when ARP is unable to resolve)
668128872Sandre	 */
669128872Sandre	if ((ro.ro_rt->rt_flags & RTF_REJECT) &&
670128872Sandre	    ro.ro_rt->rt_rmx.rmx_expire >= time_second) {
671128872Sandre		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, NULL);
672128872Sandre		goto consumed;
673128872Sandre	}
674128872Sandre
675133480Sandre#ifndef ALTQ
676128872Sandre	/*
677128872Sandre	 * Check if there is enough space in the interface queue
678128872Sandre	 */
679128872Sandre	if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
680128872Sandre	    ifp->if_snd.ifq_maxlen) {
681128872Sandre		ipstat.ips_odropped++;
682128872Sandre		/* would send source quench here but that is depreciated */
683128872Sandre		goto drop;
684128872Sandre	}
685133480Sandre#endif
686128872Sandre
687128872Sandre	/*
688128872Sandre	 * Check if media link state of interface is not down
689128872Sandre	 */
690128872Sandre	if (ifp->if_link_state == LINK_STATE_DOWN) {
691128872Sandre		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, NULL);
692128872Sandre		goto consumed;
693128872Sandre	}
694128872Sandre
695128872Sandre	/*
696122702Sandre	 * Check if packet fits MTU or if hardware will fragement for us
697122702Sandre	 */
698122702Sandre	if (ro.ro_rt->rt_rmx.rmx_mtu)
699122702Sandre		mtu = min(ro.ro_rt->rt_rmx.rmx_mtu, ifp->if_mtu);
700122702Sandre	else
701122702Sandre		mtu = ifp->if_mtu;
702122702Sandre
703122702Sandre	if (ip->ip_len <= mtu ||
704122702Sandre	    (ifp->if_hwassist & CSUM_FRAGMENT && (ip->ip_off & IP_DF) == 0)) {
705122702Sandre		/*
706122702Sandre		 * Restore packet header fields to original values
707122702Sandre		 */
708122702Sandre		ip->ip_len = htons(ip->ip_len);
709122702Sandre		ip->ip_off = htons(ip->ip_off);
710122702Sandre		/*
711122702Sandre		 * Send off the packet via outgoing interface
712122702Sandre		 */
713122702Sandre		error = (*ifp->if_output)(ifp, m,
714122702Sandre				(struct sockaddr *)dst, ro.ro_rt);
715122702Sandre	} else {
716122702Sandre		/*
717128872Sandre		 * Handle EMSGSIZE with icmp reply needfrag for TCP MTU discovery
718122702Sandre		 */
719122702Sandre		if (ip->ip_off & IP_DF) {
720128872Sandre			ipstat.ips_cantfrag++;
721122702Sandre			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG,
722123740Speter				0, ifp);
723128872Sandre			goto consumed;
724122702Sandre		} else {
725122702Sandre			/*
726122702Sandre			 * We have to fragement the packet
727122702Sandre			 */
728122702Sandre			m->m_pkthdr.csum_flags |= CSUM_IP;
729128872Sandre			/*
730128872Sandre			 * ip_fragment expects ip_len and ip_off in host byte
731128872Sandre			 * order but returns all packets in network byte order
732128872Sandre			 */
733122702Sandre			if (ip_fragment(ip, &m, mtu, ifp->if_hwassist,
734122702Sandre					(~ifp->if_hwassist & CSUM_DELAY_IP))) {
735122702Sandre				goto drop;
736122702Sandre			}
737122702Sandre			KASSERT(m != NULL, ("null mbuf and no error"));
738122702Sandre			/*
739122702Sandre			 * Send off the fragments via outgoing interface
740122702Sandre			 */
741122702Sandre			error = 0;
742122702Sandre			do {
743122702Sandre				m0 = m->m_nextpkt;
744122702Sandre				m->m_nextpkt = NULL;
745122702Sandre
746122702Sandre				error = (*ifp->if_output)(ifp, m,
747122702Sandre					(struct sockaddr *)dst, ro.ro_rt);
748122702Sandre				if (error)
749122702Sandre					break;
750122702Sandre			} while ((m = m0) != NULL);
751122702Sandre			if (error) {
752122702Sandre				/* Reclaim remaining fragments */
753122702Sandre				for (; m; m = m0) {
754122702Sandre					m0 = m->m_nextpkt;
755122702Sandre					m->m_nextpkt = NULL;
756122702Sandre					m_freem(m);
757122702Sandre				}
758122702Sandre			} else
759122702Sandre				ipstat.ips_fragmented++;
760122702Sandre		}
761122702Sandre	}
762122702Sandre
763122702Sandre	if (error != 0)
764122702Sandre		ipstat.ips_odropped++;
765122702Sandre	else {
766122702Sandre		ro.ro_rt->rt_rmx.rmx_pksent++;
767122702Sandre		ipstat.ips_forward++;
768122702Sandre		ipstat.ips_fastforward++;
769122702Sandre	}
770128872Sandreconsumed:
771122702Sandre	RTFREE(ro.ro_rt);
772122702Sandre	return 1;
773122702Sandredrop:
774122702Sandre	if (m)
775122702Sandre		m_freem(m);
776128872Sandre	if (ro.ro_rt)
777128872Sandre		RTFREE(ro.ro_rt);
778122702Sandre	return 1;
779122702Sandre}
780