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
65 /*
66 * Many thanks to Matt Thomas of NetBSD for basic structure of ip_flow.c which
67 * is being followed here.
68 */
69
70 #include <sys/cdefs.h>
71 #include "opt_ipstealth.h"
72 #include "opt_sctp.h"
73
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/kernel.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/protosw.h>
80 #include <sys/sdt.h>
81 #include <sys/socket.h>
82 #include <sys/sysctl.h>
83
84 #include <net/if.h>
85 #include <net/if_types.h>
86 #include <net/if_var.h>
87 #include <net/if_dl.h>
88 #include <net/if_private.h>
89 #include <net/pfil.h>
90 #include <net/route.h>
91 #include <net/route/nhop.h>
92 #include <net/vnet.h>
93
94 #include <netinet/in.h>
95 #include <netinet/in_fib.h>
96 #include <netinet/in_kdtrace.h>
97 #include <netinet/in_systm.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip.h>
100 #include <netinet/ip_var.h>
101 #include <netinet/ip_icmp.h>
102 #include <netinet/ip_options.h>
103
104 #include <machine/in_cksum.h>
105
106 #if defined(SCTP) || defined(SCTP_SUPPORT)
107 #include <netinet/sctp_crc32.h>
108 #endif
109
110 #define V_ipsendredirects VNET(ipsendredirects)
111
112 static struct mbuf *
ip_redir_alloc(struct mbuf * m,struct nhop_object * nh,u_short ip_len,struct in_addr * osrc,struct in_addr * newgw)113 ip_redir_alloc(struct mbuf *m, struct nhop_object *nh, u_short ip_len,
114 struct in_addr *osrc, struct in_addr *newgw)
115 {
116 struct in_ifaddr *nh_ia;
117 struct mbuf *mcopy;
118
119 KASSERT(nh != NULL, ("%s: m %p nh is NULL\n", __func__, m));
120
121 /*
122 * Only send a redirect if:
123 * - Redirects are not disabled (must be checked by caller),
124 * - We have not applied NAT (must be checked by caller as possible),
125 * - Neither a MCAST or BCAST packet (must be checked by caller)
126 * [RFC1009 Appendix A.2].
127 * - The packet does not do IP source routing or having any other
128 * IP options (this case was handled already by ip_input() calling
129 * ip_dooptions() [RFC792, p13],
130 * - The packet is being forwarded out the same physical interface
131 * that it was received from [RFC1812, 5.2.7.2].
132 */
133
134 /*
135 * - The forwarding route was not created by a redirect
136 * [RFC1812, 5.2.7.2], or
137 * if it was to follow a default route (see below).
138 * - The next-hop is reachable by us [RFC1009 Appendix A.2].
139 */
140 if ((nh->nh_flags & (NHF_DEFAULT | NHF_REDIRECT |
141 NHF_BLACKHOLE | NHF_REJECT)) != 0)
142 return (NULL);
143
144 /* Get the new gateway. */
145 if ((nh->nh_flags & NHF_GATEWAY) == 0 || nh->gw_sa.sa_family != AF_INET)
146 return (NULL);
147 newgw->s_addr = nh->gw4_sa.sin_addr.s_addr;
148
149 /*
150 * - The resulting forwarding destination is not "This host on this
151 * network" [RFC1122, Section 3.2.1.3] (default route check above).
152 */
153 if (newgw->s_addr == 0)
154 return (NULL);
155
156 /*
157 * - We know how to reach the sender and the source address is
158 * directly connected to us [RFC792, p13].
159 * + The new gateway address and the source address are on the same
160 * subnet [RFC1009 Appendix A.2, RFC1122 3.2.2.2, RFC1812, 5.2.7.2].
161 * NB: if you think multiple logical subnets on the same wire should
162 * receive redirects read [RFC1812, APPENDIX C (14->15)].
163 */
164 nh_ia = (struct in_ifaddr *)nh->nh_ifa;
165 if ((ntohl(osrc->s_addr) & nh_ia->ia_subnetmask) != nh_ia->ia_subnet)
166 return (NULL);
167
168 /* Prepare for sending the redirect. */
169
170 /*
171 * Make a copy of as much as we need of the packet as the original
172 * one will be forwarded but we need (a portion) for icmp_error().
173 */
174 mcopy = m_gethdr(M_NOWAIT, m->m_type);
175 if (mcopy == NULL)
176 return (NULL);
177
178 if (m_dup_pkthdr(mcopy, m, M_NOWAIT) == 0) {
179 /*
180 * It's probably ok if the pkthdr dup fails (because
181 * the deep copy of the tag chain failed), but for now
182 * be conservative and just discard the copy since
183 * code below may some day want the tags.
184 */
185 m_free(mcopy);
186 return (NULL);
187 }
188 mcopy->m_len = min(ip_len, M_TRAILINGSPACE(mcopy));
189 mcopy->m_pkthdr.len = mcopy->m_len;
190 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
191
192 return (mcopy);
193 }
194
195
196 static int
ip_findroute(struct nhop_object ** pnh,struct in_addr dest,struct mbuf * m)197 ip_findroute(struct nhop_object **pnh, struct in_addr dest, struct mbuf *m)
198 {
199 struct nhop_object *nh;
200
201 nh = fib4_lookup(M_GETFIB(m), dest, 0, NHR_NONE,
202 m->m_pkthdr.flowid);
203 if (nh == NULL) {
204 IPSTAT_INC(ips_noroute);
205 IPSTAT_INC(ips_cantforward);
206 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
207 return (EHOSTUNREACH);
208 }
209 /*
210 * Drop blackholed traffic and directed broadcasts.
211 */
212 if ((nh->nh_flags & (NHF_BLACKHOLE | NHF_BROADCAST)) != 0) {
213 IPSTAT_INC(ips_cantforward);
214 m_freem(m);
215 return (EHOSTUNREACH);
216 }
217
218 if (nh->nh_flags & NHF_REJECT) {
219 IPSTAT_INC(ips_cantforward);
220 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
221 return (EHOSTUNREACH);
222 }
223
224 *pnh = nh;
225
226 return (0);
227 }
228
229 /*
230 * Try to forward a packet based on the destination address.
231 * This is a fast path optimized for the plain forwarding case.
232 * If the packet is handled (and consumed) here then we return NULL;
233 * otherwise mbuf is returned and the packet should be delivered
234 * to ip_input for full processing.
235 */
236 struct mbuf *
ip_tryforward(struct mbuf * m)237 ip_tryforward(struct mbuf *m)
238 {
239 struct ip *ip;
240 struct mbuf *m0 = NULL;
241 struct nhop_object *nh = NULL;
242 struct route ro;
243 struct sockaddr_in *dst;
244 const struct sockaddr *gw;
245 struct in_addr dest, odest, rtdest, osrc;
246 uint16_t ip_len, ip_off;
247 int error = 0;
248 struct m_tag *fwd_tag = NULL;
249 struct mbuf *mcopy = NULL;
250 struct in_addr redest;
251 /*
252 * Are we active and forwarding packets?
253 */
254
255 M_ASSERTVALID(m);
256 M_ASSERTPKTHDR(m);
257
258 /*
259 * Only IP packets without options
260 */
261 ip = mtod(m, struct ip *);
262
263 if (ip->ip_hl != (sizeof(struct ip) >> 2)) {
264 if (V_ip_doopts == 1)
265 return m;
266 else if (V_ip_doopts == 2) {
267 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_FILTER_PROHIB,
268 0, 0);
269 return NULL; /* mbuf already free'd */
270 }
271 /* else ignore IP options and continue */
272 }
273
274 /*
275 * Only unicast IP, not from loopback, no L2 or IP broadcast,
276 * no multicast, no INADDR_ANY
277 *
278 * XXX: Probably some of these checks could be direct drop
279 * conditions. However it is not clear whether there are some
280 * hacks or obscure behaviours which make it necessary to
281 * let ip_input handle it. We play safe here and let ip_input
282 * deal with it until it is proven that we can directly drop it.
283 */
284 if ((m->m_flags & (M_BCAST|M_MCAST)) ||
285 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) ||
286 in_broadcast(ip->ip_src) ||
287 in_broadcast(ip->ip_dst) ||
288 IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
289 IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
290 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) ||
291 IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) )
292 return m;
293
294 /*
295 * Is it for a local address on this host?
296 */
297 if (in_localip(ip->ip_dst))
298 return m;
299
300 IPSTAT_INC(ips_total);
301
302 /*
303 * Step 3: incoming packet firewall processing
304 */
305
306 odest.s_addr = dest.s_addr = ip->ip_dst.s_addr;
307 osrc.s_addr = ip->ip_src.s_addr;
308
309 /*
310 * Run through list of ipfilter hooks for input packets
311 */
312 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
313 goto passin;
314
315 if (pfil_mbuf_in(V_inet_pfil_head, &m, m->m_pkthdr.rcvif,
316 NULL) != PFIL_PASS)
317 goto drop;
318
319 M_ASSERTVALID(m);
320 M_ASSERTPKTHDR(m);
321
322 ip = mtod(m, struct ip *); /* m may have changed by pfil hook */
323 dest.s_addr = ip->ip_dst.s_addr;
324
325 /*
326 * Destination address changed?
327 */
328 if (odest.s_addr != dest.s_addr) {
329 /*
330 * Is it now for a local address on this host?
331 */
332 if (in_localip(dest))
333 goto forwardlocal;
334 /*
335 * Go on with new destination address
336 */
337 }
338
339 if (m->m_flags & M_FASTFWD_OURS) {
340 /*
341 * ipfw changed it for a local address on this host.
342 */
343 goto forwardlocal;
344 }
345
346 passin:
347 /*
348 * Step 4: decrement TTL and look up route
349 */
350
351 /*
352 * Check TTL
353 */
354 #ifdef IPSTEALTH
355 if (!V_ipstealth) {
356 #endif
357 if (ip->ip_ttl <= IPTTLDEC) {
358 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
359 return NULL; /* mbuf already free'd */
360 }
361
362 /*
363 * Decrement the TTL and incrementally change the IP header checksum.
364 * Don't bother doing this with hw checksum offloading, it's faster
365 * doing it right here.
366 */
367 ip->ip_ttl -= IPTTLDEC;
368 if (ip->ip_sum >= (u_int16_t) ~htons(IPTTLDEC << 8))
369 ip->ip_sum -= ~htons(IPTTLDEC << 8);
370 else
371 ip->ip_sum += htons(IPTTLDEC << 8);
372 #ifdef IPSTEALTH
373 }
374 #endif
375
376 /*
377 * Next hop forced by pfil(9) hook?
378 */
379 if ((m->m_flags & M_IP_NEXTHOP) &&
380 ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
381 /*
382 * Now we will find route to forced destination.
383 */
384 dest.s_addr = ((struct sockaddr_in *)
385 (fwd_tag + 1))->sin_addr.s_addr;
386 m_tag_delete(m, fwd_tag);
387 m->m_flags &= ~M_IP_NEXTHOP;
388 }
389
390 /*
391 * Find route to destination.
392 */
393 if (ip_findroute(&nh, dest, m) != 0)
394 return (NULL); /* icmp unreach already sent */
395
396 /*
397 * Avoid second route lookup by caching destination.
398 */
399 rtdest.s_addr = dest.s_addr;
400
401 /*
402 * Step 5: outgoing firewall packet processing
403 */
404 if (!PFIL_HOOKED_OUT(V_inet_pfil_head))
405 goto passout;
406
407 if (pfil_mbuf_fwd(V_inet_pfil_head, &m, nh->nh_ifp,
408 NULL) != PFIL_PASS)
409 goto drop;
410
411 M_ASSERTVALID(m);
412 M_ASSERTPKTHDR(m);
413
414 ip = mtod(m, struct ip *);
415 dest.s_addr = ip->ip_dst.s_addr;
416
417 /*
418 * Destination address changed?
419 */
420 if (m->m_flags & M_IP_NEXTHOP)
421 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
422 else
423 fwd_tag = NULL;
424 if (odest.s_addr != dest.s_addr || fwd_tag != NULL) {
425 /*
426 * Is it now for a local address on this host?
427 */
428 if (m->m_flags & M_FASTFWD_OURS || in_localip(dest)) {
429 forwardlocal:
430 /*
431 * Return packet for processing by ip_input().
432 */
433 m->m_flags |= M_FASTFWD_OURS;
434 return (m);
435 }
436 /*
437 * Redo route lookup with new destination address
438 */
439 if (fwd_tag) {
440 dest.s_addr = ((struct sockaddr_in *)
441 (fwd_tag + 1))->sin_addr.s_addr;
442 m_tag_delete(m, fwd_tag);
443 m->m_flags &= ~M_IP_NEXTHOP;
444 }
445 if (dest.s_addr != rtdest.s_addr &&
446 ip_findroute(&nh, dest, m) != 0)
447 return (NULL); /* icmp unreach already sent */
448 }
449
450 passout:
451 /*
452 * Step 6: send off the packet
453 */
454 ip_len = ntohs(ip->ip_len);
455 ip_off = ntohs(ip->ip_off);
456
457 bzero(&ro, sizeof(ro));
458 dst = (struct sockaddr_in *)&ro.ro_dst;
459 dst->sin_family = AF_INET;
460 dst->sin_len = sizeof(*dst);
461 dst->sin_addr = dest;
462 if (nh->nh_flags & NHF_GATEWAY) {
463 gw = &nh->gw_sa;
464 ro.ro_flags |= RT_HAS_GW;
465 } else
466 gw = (const struct sockaddr *)dst;
467
468 /*
469 * If TCP/UDP header still needs a valid checksum and interface will not
470 * calculate it for us, do it here.
471 */
472 if (__predict_false(m->m_pkthdr.csum_flags & CSUM_DELAY_DATA &
473 ~nh->nh_ifp->if_hwassist)) {
474 in_delayed_cksum(m);
475 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
476 }
477 #if defined(SCTP) || defined(SCTP_SUPPORT)
478 if (__predict_false(m->m_pkthdr.csum_flags & CSUM_IP_SCTP &
479 ~nh->nh_ifp->if_hwassist)) {
480 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
481 m->m_pkthdr.csum_flags &= ~CSUM_IP_SCTP;
482 }
483 #endif
484
485 /* Handle redirect case. */
486 redest.s_addr = 0;
487 if (V_ipsendredirects && osrc.s_addr == ip->ip_src.s_addr &&
488 nh->nh_ifp == m->m_pkthdr.rcvif)
489 mcopy = ip_redir_alloc(m, nh, ip_len, &osrc, &redest);
490
491 /*
492 * Check if packet fits MTU or if hardware will fragment for us
493 */
494 if (ip_len <= nh->nh_mtu) {
495 /*
496 * Avoid confusing lower layers.
497 */
498 m_clrprotoflags(m);
499 /*
500 * Send off the packet via outgoing interface
501 */
502 IP_PROBE(send, NULL, NULL, ip, nh->nh_ifp, ip, NULL);
503 error = (*nh->nh_ifp->if_output)(nh->nh_ifp, m, gw, &ro);
504 } else {
505 /*
506 * Handle EMSGSIZE with icmp reply needfrag for TCP MTU discovery
507 */
508 if (ip_off & IP_DF) {
509 IPSTAT_INC(ips_cantfrag);
510 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG,
511 0, nh->nh_mtu);
512 goto consumed;
513 } else {
514 /*
515 * We have to fragment the packet
516 */
517 m->m_pkthdr.csum_flags |= CSUM_IP;
518 if (ip_fragment(ip, &m, nh->nh_mtu,
519 nh->nh_ifp->if_hwassist) != 0)
520 goto drop;
521 KASSERT(m != NULL, ("null mbuf and no error"));
522 /*
523 * Send off the fragments via outgoing interface
524 */
525 error = 0;
526 do {
527 m0 = m->m_nextpkt;
528 m->m_nextpkt = NULL;
529 /*
530 * Avoid confusing lower layers.
531 */
532 m_clrprotoflags(m);
533
534 IP_PROBE(send, NULL, NULL,
535 mtod(m, struct ip *), nh->nh_ifp,
536 mtod(m, struct ip *), NULL);
537 error = (*nh->nh_ifp->if_output)(nh->nh_ifp, m,
538 gw, &ro);
539 if (error)
540 break;
541 } while ((m = m0) != NULL);
542 if (error) {
543 /* Reclaim remaining fragments */
544 for (m = m0; m; m = m0) {
545 m0 = m->m_nextpkt;
546 m_freem(m);
547 }
548 } else
549 IPSTAT_INC(ips_fragmented);
550 }
551 }
552
553 if (error != 0)
554 IPSTAT_INC(ips_odropped);
555 else {
556 IPSTAT_INC(ips_forward);
557 IPSTAT_INC(ips_fastforward);
558 }
559
560 /* Send required redirect */
561 if (mcopy != NULL) {
562 icmp_error(mcopy, ICMP_REDIRECT, ICMP_REDIRECT_HOST, redest.s_addr, 0);
563 mcopy = NULL; /* Was consumed by callee. */
564 }
565
566 consumed:
567 if (mcopy != NULL)
568 m_freem(mcopy);
569 return NULL;
570 drop:
571 if (m)
572 m_freem(m);
573 return NULL;
574 }
575