1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 2003 Andre Oppermann, Internet Business Solutions AG
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
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
17 *    permission.
18 *
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
29 * SUCH DAMAGE.
30 */
31
32/*
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
42 * transmits.
43 *
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.
47 *
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.
51 *
52 * Firewalling is fully supported including divert, ipfw fwd and ipfilter
53 * ipnat and address rewrite.
54 *
55 * IPSEC is not supported if this host is a tunnel broker. IPSEC is
56 * supported for connections to/from local host.
57 *
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.
60 *
61 * We take full advantage of hardware support for IP checksum and
62 * fragmentation offloading.
63 *
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.
71 */
72
73/*
74 * Many thanks to Matt Thomas of NetBSD for basic structure of ip_flow.c which
75 * is being followed here.
76 */
77
78#include <sys/cdefs.h>
79__FBSDID("$FreeBSD$");
80
81#include "opt_ipstealth.h"
82
83#include <sys/param.h>
84#include <sys/systm.h>
85#include <sys/kernel.h>
86#include <sys/malloc.h>
87#include <sys/mbuf.h>
88#include <sys/protosw.h>
89#include <sys/sdt.h>
90#include <sys/socket.h>
91#include <sys/sysctl.h>
92
93#include <net/if.h>
94#include <net/if_types.h>
95#include <net/if_var.h>
96#include <net/if_dl.h>
97#include <net/pfil.h>
98#include <net/route.h>
99#include <net/route/nhop.h>
100#include <net/vnet.h>
101
102#include <netinet/in.h>
103#include <netinet/in_fib.h>
104#include <netinet/in_kdtrace.h>
105#include <netinet/in_systm.h>
106#include <netinet/in_var.h>
107#include <netinet/ip.h>
108#include <netinet/ip_var.h>
109#include <netinet/ip_icmp.h>
110#include <netinet/ip_options.h>
111
112#include <machine/in_cksum.h>
113
114#define	V_ipsendredirects	VNET(ipsendredirects)
115
116static struct mbuf *
117ip_redir_alloc(struct mbuf *m, struct nhop_object *nh,
118    struct ip *ip, in_addr_t *addr)
119{
120	struct mbuf *mcopy = m_gethdr(M_NOWAIT, m->m_type);
121
122	if (mcopy == NULL)
123		return (NULL);
124
125	if (m_dup_pkthdr(mcopy, m, M_NOWAIT) == 0) {
126		/*
127		 * It's probably ok if the pkthdr dup fails (because
128		 * the deep copy of the tag chain failed), but for now
129		 * be conservative and just discard the copy since
130		 * code below may some day want the tags.
131		 */
132		m_free(mcopy);
133		return (NULL);
134	}
135	mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
136	mcopy->m_pkthdr.len = mcopy->m_len;
137	m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
138
139	if (nh != NULL &&
140	    ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
141		struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
142		u_long src = ntohl(ip->ip_src.s_addr);
143
144		if (nh_ia != NULL &&
145		    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
146			if (nh->nh_flags & NHF_GATEWAY)
147				*addr = nh->gw4_sa.sin_addr.s_addr;
148			else
149				*addr = ip->ip_dst.s_addr;
150		}
151	}
152	return (mcopy);
153}
154
155
156static int
157ip_findroute(struct nhop_object **pnh, struct in_addr dest, struct mbuf *m)
158{
159	struct nhop_object *nh;
160
161	nh = fib4_lookup(M_GETFIB(m), dest, 0, NHR_NONE,
162	    m->m_pkthdr.flowid);
163	if (nh == NULL) {
164		IPSTAT_INC(ips_noroute);
165		IPSTAT_INC(ips_cantforward);
166		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
167		return (EHOSTUNREACH);
168	}
169	/*
170	 * Drop blackholed traffic and directed broadcasts.
171	 */
172	if ((nh->nh_flags & (NHF_BLACKHOLE | NHF_BROADCAST)) != 0) {
173		IPSTAT_INC(ips_cantforward);
174		m_freem(m);
175		return (EHOSTUNREACH);
176	}
177
178	if (nh->nh_flags & NHF_REJECT) {
179		IPSTAT_INC(ips_cantforward);
180		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
181		return (EHOSTUNREACH);
182	}
183
184	*pnh = nh;
185
186	return (0);
187}
188
189/*
190 * Try to forward a packet based on the destination address.
191 * This is a fast path optimized for the plain forwarding case.
192 * If the packet is handled (and consumed) here then we return NULL;
193 * otherwise mbuf is returned and the packet should be delivered
194 * to ip_input for full processing.
195 */
196struct mbuf *
197ip_tryforward(struct mbuf *m)
198{
199	struct ip *ip;
200	struct mbuf *m0 = NULL;
201	struct nhop_object *nh = NULL;
202	struct sockaddr_in dst;
203	struct in_addr dest, odest, rtdest;
204	uint16_t ip_len, ip_off;
205	int error = 0;
206	struct m_tag *fwd_tag = NULL;
207	struct mbuf *mcopy = NULL;
208	struct in_addr redest;
209	/*
210	 * Are we active and forwarding packets?
211	 */
212
213	M_ASSERTVALID(m);
214	M_ASSERTPKTHDR(m);
215
216#ifdef ALTQ
217	/*
218	 * Is packet dropped by traffic conditioner?
219	 */
220	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
221		goto drop;
222#endif
223
224	/*
225	 * Only IP packets without options
226	 */
227	ip = mtod(m, struct ip *);
228
229	if (ip->ip_hl != (sizeof(struct ip) >> 2)) {
230		if (V_ip_doopts == 1)
231			return m;
232		else if (V_ip_doopts == 2) {
233			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_FILTER_PROHIB,
234				0, 0);
235			return NULL;	/* mbuf already free'd */
236		}
237		/* else ignore IP options and continue */
238	}
239
240	/*
241	 * Only unicast IP, not from loopback, no L2 or IP broadcast,
242	 * no multicast, no INADDR_ANY
243	 *
244	 * XXX: Probably some of these checks could be direct drop
245	 * conditions.  However it is not clear whether there are some
246	 * hacks or obscure behaviours which make it necessary to
247	 * let ip_input handle it.  We play safe here and let ip_input
248	 * deal with it until it is proven that we can directly drop it.
249	 */
250	if ((m->m_flags & (M_BCAST|M_MCAST)) ||
251	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) ||
252	    ntohl(ip->ip_src.s_addr) == (u_long)INADDR_BROADCAST ||
253	    ntohl(ip->ip_dst.s_addr) == (u_long)INADDR_BROADCAST ||
254	    IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
255	    IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
256	    IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) ||
257	    IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
258	    ip->ip_src.s_addr == INADDR_ANY ||
259	    ip->ip_dst.s_addr == INADDR_ANY )
260		return m;
261
262	/*
263	 * Is it for a local address on this host?
264	 */
265	if (in_localip(ip->ip_dst))
266		return m;
267
268	IPSTAT_INC(ips_total);
269
270	/*
271	 * Step 3: incoming packet firewall processing
272	 */
273
274	odest.s_addr = dest.s_addr = ip->ip_dst.s_addr;
275
276	/*
277	 * Run through list of ipfilter hooks for input packets
278	 */
279	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
280		goto passin;
281
282	if (pfil_run_hooks(V_inet_pfil_head, &m, m->m_pkthdr.rcvif, PFIL_IN,
283	    NULL) != PFIL_PASS)
284		goto drop;
285
286	M_ASSERTVALID(m);
287	M_ASSERTPKTHDR(m);
288
289	ip = mtod(m, struct ip *);	/* m may have changed by pfil hook */
290	dest.s_addr = ip->ip_dst.s_addr;
291
292	/*
293	 * Destination address changed?
294	 */
295	if (odest.s_addr != dest.s_addr) {
296		/*
297		 * Is it now for a local address on this host?
298		 */
299		if (in_localip(dest))
300			goto forwardlocal;
301		/*
302		 * Go on with new destination address
303		 */
304	}
305
306	if (m->m_flags & M_FASTFWD_OURS) {
307		/*
308		 * ipfw changed it for a local address on this host.
309		 */
310		goto forwardlocal;
311	}
312
313passin:
314	/*
315	 * Step 4: decrement TTL and look up route
316	 */
317
318	/*
319	 * Check TTL
320	 */
321#ifdef IPSTEALTH
322	if (!V_ipstealth) {
323#endif
324	if (ip->ip_ttl <= IPTTLDEC) {
325		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
326		return NULL;	/* mbuf already free'd */
327	}
328
329	/*
330	 * Decrement the TTL and incrementally change the IP header checksum.
331	 * Don't bother doing this with hw checksum offloading, it's faster
332	 * doing it right here.
333	 */
334	ip->ip_ttl -= IPTTLDEC;
335	if (ip->ip_sum >= (u_int16_t) ~htons(IPTTLDEC << 8))
336		ip->ip_sum -= ~htons(IPTTLDEC << 8);
337	else
338		ip->ip_sum += htons(IPTTLDEC << 8);
339#ifdef IPSTEALTH
340	}
341#endif
342
343	/*
344	 * Next hop forced by pfil(9) hook?
345	 */
346	if ((m->m_flags & M_IP_NEXTHOP) &&
347	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
348		/*
349		 * Now we will find route to forced destination.
350		 */
351		dest.s_addr = ((struct sockaddr_in *)
352			    (fwd_tag + 1))->sin_addr.s_addr;
353		m_tag_delete(m, fwd_tag);
354		m->m_flags &= ~M_IP_NEXTHOP;
355	}
356
357	/*
358	 * Find route to destination.
359	 */
360	if (ip_findroute(&nh, dest, m) != 0)
361		return (NULL);	/* icmp unreach already sent */
362
363	/*
364	 * Avoid second route lookup by caching destination.
365	 */
366	rtdest.s_addr = dest.s_addr;
367
368	/*
369	 * Step 5: outgoing firewall packet processing
370	 */
371	if (!PFIL_HOOKED_OUT(V_inet_pfil_head))
372		goto passout;
373
374	if (pfil_run_hooks(V_inet_pfil_head, &m, nh->nh_ifp,
375	    PFIL_OUT | PFIL_FWD, NULL) != PFIL_PASS)
376		goto drop;
377
378	M_ASSERTVALID(m);
379	M_ASSERTPKTHDR(m);
380
381	ip = mtod(m, struct ip *);
382	dest.s_addr = ip->ip_dst.s_addr;
383
384	/*
385	 * Destination address changed?
386	 */
387	if (m->m_flags & M_IP_NEXTHOP)
388		fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
389	else
390		fwd_tag = NULL;
391	if (odest.s_addr != dest.s_addr || fwd_tag != NULL) {
392		/*
393		 * Is it now for a local address on this host?
394		 */
395		if (m->m_flags & M_FASTFWD_OURS || in_localip(dest)) {
396forwardlocal:
397			/*
398			 * Return packet for processing by ip_input().
399			 */
400			m->m_flags |= M_FASTFWD_OURS;
401			return (m);
402		}
403		/*
404		 * Redo route lookup with new destination address
405		 */
406		if (fwd_tag) {
407			dest.s_addr = ((struct sockaddr_in *)
408				    (fwd_tag + 1))->sin_addr.s_addr;
409			m_tag_delete(m, fwd_tag);
410			m->m_flags &= ~M_IP_NEXTHOP;
411		}
412		if (dest.s_addr != rtdest.s_addr &&
413		    ip_findroute(&nh, dest, m) != 0)
414			return (NULL);	/* icmp unreach already sent */
415	}
416
417passout:
418	/*
419	 * Step 6: send off the packet
420	 */
421	ip_len = ntohs(ip->ip_len);
422	ip_off = ntohs(ip->ip_off);
423
424	bzero(&dst, sizeof(dst));
425	dst.sin_family = AF_INET;
426	dst.sin_len = sizeof(dst);
427	if (nh->nh_flags & NHF_GATEWAY)
428		dst.sin_addr = nh->gw4_sa.sin_addr;
429	else
430		dst.sin_addr = dest;
431
432	/*
433	 * Handle redirect case.
434	 */
435	redest.s_addr = 0;
436	if (V_ipsendredirects && (nh->nh_ifp == m->m_pkthdr.rcvif))
437		mcopy = ip_redir_alloc(m, nh, ip, &redest.s_addr);
438
439	/*
440	 * Check if packet fits MTU or if hardware will fragment for us
441	 */
442	if (ip_len <= nh->nh_mtu) {
443		/*
444		 * Avoid confusing lower layers.
445		 */
446		m_clrprotoflags(m);
447		/*
448		 * Send off the packet via outgoing interface
449		 */
450		IP_PROBE(send, NULL, NULL, ip, nh->nh_ifp, ip, NULL);
451		error = (*nh->nh_ifp->if_output)(nh->nh_ifp, m,
452		    (struct sockaddr *)&dst, NULL);
453	} else {
454		/*
455		 * Handle EMSGSIZE with icmp reply needfrag for TCP MTU discovery
456		 */
457		if (ip_off & IP_DF) {
458			IPSTAT_INC(ips_cantfrag);
459			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG,
460				0, nh->nh_mtu);
461			goto consumed;
462		} else {
463			/*
464			 * We have to fragment the packet
465			 */
466			m->m_pkthdr.csum_flags |= CSUM_IP;
467			if (ip_fragment(ip, &m, nh->nh_mtu,
468			    nh->nh_ifp->if_hwassist) != 0)
469				goto drop;
470			KASSERT(m != NULL, ("null mbuf and no error"));
471			/*
472			 * Send off the fragments via outgoing interface
473			 */
474			error = 0;
475			do {
476				m0 = m->m_nextpkt;
477				m->m_nextpkt = NULL;
478				/*
479				 * Avoid confusing lower layers.
480				 */
481				m_clrprotoflags(m);
482
483				IP_PROBE(send, NULL, NULL,
484				    mtod(m, struct ip *), nh->nh_ifp,
485				    mtod(m, struct ip *), NULL);
486				error = (*nh->nh_ifp->if_output)(nh->nh_ifp, m,
487				    (struct sockaddr *)&dst, NULL);
488				if (error)
489					break;
490			} while ((m = m0) != NULL);
491			if (error) {
492				/* Reclaim remaining fragments */
493				for (m = m0; m; m = m0) {
494					m0 = m->m_nextpkt;
495					m_freem(m);
496				}
497			} else
498				IPSTAT_INC(ips_fragmented);
499		}
500	}
501
502	if (error != 0)
503		IPSTAT_INC(ips_odropped);
504	else {
505		IPSTAT_INC(ips_forward);
506		IPSTAT_INC(ips_fastforward);
507	}
508
509	/* Send required redirect */
510	if (mcopy != NULL) {
511		icmp_error(mcopy, ICMP_REDIRECT, ICMP_REDIRECT_HOST, redest.s_addr, 0);
512		mcopy = NULL; /* Freed by caller */
513	}
514
515consumed:
516	if (mcopy != NULL)
517		m_freem(mcopy);
518	return NULL;
519drop:
520	if (m)
521		m_freem(m);
522	return NULL;
523}
524