1/*-
2 * Copyright (c) 2003 Andre Oppermann, Internet Business Solutions AG
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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
15 * permission.
16 *
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
27 * SUCH DAMAGE.
28 */
29
30/*
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
40 * transmits.
41 *
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.
45 *
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.
49 *
50 * Firewalling is fully supported including divert, ipfw fwd and ipfilter
51 * ipnat and address rewrite.
52 *
53 * IPSEC is not supported if this host is a tunnel broker. IPSEC is
54 * supported for connections to/from local host.
55 *
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.
58 *
59 * We take full advantage of hardware support for IP checksum and
60 * fragmentation offloading.
61 *
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.
69 */
70
71/*
72 * Many thanks to Matt Thomas of NetBSD for basic structure of ip_flow.c which
73 * is being followed here.
74 */
75
76#include <sys/cdefs.h>
| 1/*-
2 * Copyright (c) 2003 Andre Oppermann, Internet Business Solutions AG
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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
15 * permission.
16 *
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
27 * SUCH DAMAGE.
28 */
29
30/*
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
40 * transmits.
41 *
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.
45 *
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.
49 *
50 * Firewalling is fully supported including divert, ipfw fwd and ipfilter
51 * ipnat and address rewrite.
52 *
53 * IPSEC is not supported if this host is a tunnel broker. IPSEC is
54 * supported for connections to/from local host.
55 *
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.
58 *
59 * We take full advantage of hardware support for IP checksum and
60 * fragmentation offloading.
61 *
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.
69 */
70
71/*
72 * Many thanks to Matt Thomas of NetBSD for basic structure of ip_flow.c which
73 * is being followed here.
74 */
75
76#include <sys/cdefs.h>
|
110#define V_ipfastforward_active VNET(ipfastforward_active)
111
112SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fastforwarding, CTLFLAG_RW,
113 &VNET_NAME(ipfastforward_active), 0, "Enable fast IP forwarding");
114
115static struct sockaddr_in *
116ip_findroute(struct route *ro, struct in_addr dest, struct mbuf *m)
117{
118 struct sockaddr_in *dst;
119 struct rtentry *rt;
120
121 /*
122 * Find route to destination.
123 */
124 bzero(ro, sizeof(*ro));
125 dst = (struct sockaddr_in *)&ro->ro_dst;
126 dst->sin_family = AF_INET;
127 dst->sin_len = sizeof(*dst);
128 dst->sin_addr.s_addr = dest.s_addr;
129 in_rtalloc_ign(ro, 0, M_GETFIB(m));
130
131 /*
132 * Route there and interface still up?
133 */
134 rt = ro->ro_rt;
135 if (rt && (rt->rt_flags & RTF_UP) &&
136 (rt->rt_ifp->if_flags & IFF_UP) &&
137 (rt->rt_ifp->if_drv_flags & IFF_DRV_RUNNING)) {
138 if (rt->rt_flags & RTF_GATEWAY)
139 dst = (struct sockaddr_in *)rt->rt_gateway;
140 } else {
141 IPSTAT_INC(ips_noroute);
142 IPSTAT_INC(ips_cantforward);
143 if (rt)
144 RTFREE(rt);
145 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
146 return NULL;
147 }
148 return dst;
149}
150
151/*
152 * Try to forward a packet based on the destination address.
153 * This is a fast path optimized for the plain forwarding case.
154 * If the packet is handled (and consumed) here then we return NULL;
155 * otherwise mbuf is returned and the packet should be delivered
156 * to ip_input for full processing.
157 */
158struct mbuf *
159ip_fastforward(struct mbuf *m)
160{
161 struct ip *ip;
162 struct mbuf *m0 = NULL;
163 struct route ro;
164 struct sockaddr_in *dst = NULL;
165 struct ifnet *ifp;
166 struct in_addr odest, dest;
167 u_short sum, ip_len;
168 int error = 0;
169 int hlen, mtu;
170#ifdef IPFIREWALL_FORWARD
171 struct m_tag *fwd_tag;
172#endif
173
174 /*
175 * Are we active and forwarding packets?
176 */
177 if (!V_ipfastforward_active || !V_ipforwarding)
178 return m;
179
180 M_ASSERTVALID(m);
181 M_ASSERTPKTHDR(m);
182
183 bzero(&ro, sizeof(ro));
184
185 /*
186 * Step 1: check for packet drop conditions (and sanity checks)
187 */
188
189 /*
190 * Is entire packet big enough?
191 */
192 if (m->m_pkthdr.len < sizeof(struct ip)) {
193 IPSTAT_INC(ips_tooshort);
194 goto drop;
195 }
196
197 /*
198 * Is first mbuf large enough for ip header and is header present?
199 */
200 if (m->m_len < sizeof (struct ip) &&
201 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
202 IPSTAT_INC(ips_toosmall);
203 return NULL; /* mbuf already free'd */
204 }
205
206 ip = mtod(m, struct ip *);
207
208 /*
209 * Is it IPv4?
210 */
211 if (ip->ip_v != IPVERSION) {
212 IPSTAT_INC(ips_badvers);
213 goto drop;
214 }
215
216 /*
217 * Is IP header length correct and is it in first mbuf?
218 */
219 hlen = ip->ip_hl << 2;
220 if (hlen < sizeof(struct ip)) { /* minimum header length */
221 IPSTAT_INC(ips_badlen);
222 goto drop;
223 }
224 if (hlen > m->m_len) {
225 if ((m = m_pullup(m, hlen)) == NULL) {
226 IPSTAT_INC(ips_badhlen);
227 return NULL; /* mbuf already free'd */
228 }
229 ip = mtod(m, struct ip *);
230 }
231
232 /*
233 * Checksum correct?
234 */
235 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED)
236 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
237 else {
238 if (hlen == sizeof(struct ip))
239 sum = in_cksum_hdr(ip);
240 else
241 sum = in_cksum(m, hlen);
242 }
243 if (sum) {
244 IPSTAT_INC(ips_badsum);
245 goto drop;
246 }
247
248 /*
249 * Remember that we have checked the IP header and found it valid.
250 */
251 m->m_pkthdr.csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
252
253 ip_len = ntohs(ip->ip_len);
254
255 /*
256 * Is IP length longer than packet we have got?
257 */
258 if (m->m_pkthdr.len < ip_len) {
259 IPSTAT_INC(ips_tooshort);
260 goto drop;
261 }
262
263 /*
264 * Is packet longer than IP header tells us? If yes, truncate packet.
265 */
266 if (m->m_pkthdr.len > ip_len) {
267 if (m->m_len == m->m_pkthdr.len) {
268 m->m_len = ip_len;
269 m->m_pkthdr.len = ip_len;
270 } else
271 m_adj(m, ip_len - m->m_pkthdr.len);
272 }
273
274 /*
275 * Is packet from or to 127/8?
276 */
277 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
278 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
279 IPSTAT_INC(ips_badaddr);
280 goto drop;
281 }
282
283#ifdef ALTQ
284 /*
285 * Is packet dropped by traffic conditioner?
286 */
287 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
288 goto drop;
289#endif
290
291 /*
292 * Step 2: fallback conditions to normal ip_input path processing
293 */
294
295 /*
296 * Only IP packets without options
297 */
298 if (ip->ip_hl != (sizeof(struct ip) >> 2)) {
299 if (ip_doopts == 1)
300 return m;
301 else if (ip_doopts == 2) {
302 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_FILTER_PROHIB,
303 0, 0);
304 return NULL; /* mbuf already free'd */
305 }
306 /* else ignore IP options and continue */
307 }
308
309 /*
310 * Only unicast IP, not from loopback, no L2 or IP broadcast,
311 * no multicast, no INADDR_ANY
312 *
313 * XXX: Probably some of these checks could be direct drop
314 * conditions. However it is not clear whether there are some
315 * hacks or obscure behaviours which make it neccessary to
316 * let ip_input handle it. We play safe here and let ip_input
317 * deal with it until it is proven that we can directly drop it.
318 */
319 if ((m->m_flags & (M_BCAST|M_MCAST)) ||
320 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) ||
321 ntohl(ip->ip_src.s_addr) == (u_long)INADDR_BROADCAST ||
322 ntohl(ip->ip_dst.s_addr) == (u_long)INADDR_BROADCAST ||
323 IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
324 IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
325 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) ||
326 IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
327 ip->ip_src.s_addr == INADDR_ANY ||
328 ip->ip_dst.s_addr == INADDR_ANY )
329 return m;
330
331 /*
332 * Is it for a local address on this host?
333 */
334 if (in_localip(ip->ip_dst))
335 return m;
336
337 IPSTAT_INC(ips_total);
338
339 /*
340 * Step 3: incoming packet firewall processing
341 */
342
343 /*
344 * Convert to host representation
345 */
346 ip->ip_len = ntohs(ip->ip_len);
347 ip->ip_off = ntohs(ip->ip_off);
348
349 odest.s_addr = dest.s_addr = ip->ip_dst.s_addr;
350
351 /*
352 * Run through list of ipfilter hooks for input packets
353 */
354 if (!PFIL_HOOKED(&V_inet_pfil_hook))
355 goto passin;
356
357 if (pfil_run_hooks(
358 &V_inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL) ||
359 m == NULL)
360 goto drop;
361
362 M_ASSERTVALID(m);
363 M_ASSERTPKTHDR(m);
364
365 ip = mtod(m, struct ip *); /* m may have changed by pfil hook */
366 dest.s_addr = ip->ip_dst.s_addr;
367
368 /*
369 * Destination address changed?
370 */
371 if (odest.s_addr != dest.s_addr) {
372 /*
373 * Is it now for a local address on this host?
374 */
375 if (in_localip(dest))
376 goto forwardlocal;
377 /*
378 * Go on with new destination address
379 */
380 }
381#ifdef IPFIREWALL_FORWARD
382 if (m->m_flags & M_FASTFWD_OURS) {
383 /*
384 * ipfw changed it for a local address on this host.
385 */
386 goto forwardlocal;
387 }
388#endif /* IPFIREWALL_FORWARD */
389
390passin:
391 /*
392 * Step 4: decrement TTL and look up route
393 */
394
395 /*
396 * Check TTL
397 */
398#ifdef IPSTEALTH
399 if (!V_ipstealth) {
400#endif
401 if (ip->ip_ttl <= IPTTLDEC) {
402 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
403 return NULL; /* mbuf already free'd */
404 }
405
406 /*
407 * Decrement the TTL and incrementally change the IP header checksum.
408 * Don't bother doing this with hw checksum offloading, it's faster
409 * doing it right here.
410 */
411 ip->ip_ttl -= IPTTLDEC;
412 if (ip->ip_sum >= (u_int16_t) ~htons(IPTTLDEC << 8))
413 ip->ip_sum -= ~htons(IPTTLDEC << 8);
414 else
415 ip->ip_sum += htons(IPTTLDEC << 8);
416#ifdef IPSTEALTH
417 }
418#endif
419
420 /*
421 * Find route to destination.
422 */
423 if ((dst = ip_findroute(&ro, dest, m)) == NULL)
424 return NULL; /* icmp unreach already sent */
425 ifp = ro.ro_rt->rt_ifp;
426
427 /*
428 * Immediately drop blackholed traffic, and directed broadcasts
429 * for either the all-ones or all-zero subnet addresses on
430 * locally attached networks.
431 */
432 if ((ro.ro_rt->rt_flags & (RTF_BLACKHOLE|RTF_BROADCAST)) != 0)
433 goto drop;
434
435 /*
436 * Step 5: outgoing firewall packet processing
437 */
438
439 /*
440 * Run through list of hooks for output packets.
441 */
442 if (!PFIL_HOOKED(&V_inet_pfil_hook))
443 goto passout;
444
445 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, NULL) || m == NULL) {
446 goto drop;
447 }
448
449 M_ASSERTVALID(m);
450 M_ASSERTPKTHDR(m);
451
452 ip = mtod(m, struct ip *);
453 dest.s_addr = ip->ip_dst.s_addr;
454
455 /*
456 * Destination address changed?
457 */
458#ifndef IPFIREWALL_FORWARD
459 if (odest.s_addr != dest.s_addr) {
460#else
461 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
462 if (odest.s_addr != dest.s_addr || fwd_tag != NULL) {
463#endif /* IPFIREWALL_FORWARD */
464 /*
465 * Is it now for a local address on this host?
466 */
467#ifndef IPFIREWALL_FORWARD
468 if (in_localip(dest)) {
469#else
470 if (m->m_flags & M_FASTFWD_OURS || in_localip(dest)) {
471#endif /* IPFIREWALL_FORWARD */
472forwardlocal:
473 /*
474 * Return packet for processing by ip_input().
475 * Keep host byte order as expected at ip_input's
476 * "ours"-label.
477 */
478 m->m_flags |= M_FASTFWD_OURS;
479 if (ro.ro_rt)
480 RTFREE(ro.ro_rt);
481 return m;
482 }
483 /*
484 * Redo route lookup with new destination address
485 */
486#ifdef IPFIREWALL_FORWARD
487 if (fwd_tag) {
488 dest.s_addr = ((struct sockaddr_in *)
489 (fwd_tag + 1))->sin_addr.s_addr;
490 m_tag_delete(m, fwd_tag);
491 }
492#endif /* IPFIREWALL_FORWARD */
493 RTFREE(ro.ro_rt);
494 if ((dst = ip_findroute(&ro, dest, m)) == NULL)
495 return NULL; /* icmp unreach already sent */
496 ifp = ro.ro_rt->rt_ifp;
497 }
498
499passout:
500 /*
501 * Step 6: send off the packet
502 */
503
504 /*
505 * Check if route is dampned (when ARP is unable to resolve)
506 */
507 if ((ro.ro_rt->rt_flags & RTF_REJECT) &&
508 (ro.ro_rt->rt_rmx.rmx_expire == 0 ||
509 time_uptime < ro.ro_rt->rt_rmx.rmx_expire)) {
510 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
511 goto consumed;
512 }
513
514#ifndef ALTQ
515 /*
516 * Check if there is enough space in the interface queue
517 */
518 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
519 ifp->if_snd.ifq_maxlen) {
520 IPSTAT_INC(ips_odropped);
521 /* would send source quench here but that is depreciated */
522 goto drop;
523 }
524#endif
525
526 /*
527 * Check if media link state of interface is not down
528 */
529 if (ifp->if_link_state == LINK_STATE_DOWN) {
530 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
531 goto consumed;
532 }
533
534 /*
535 * Check if packet fits MTU or if hardware will fragment for us
536 */
537 if (ro.ro_rt->rt_rmx.rmx_mtu)
538 mtu = min(ro.ro_rt->rt_rmx.rmx_mtu, ifp->if_mtu);
539 else
540 mtu = ifp->if_mtu;
541
542 if (ip->ip_len <= mtu ||
543 (ifp->if_hwassist & CSUM_FRAGMENT && (ip->ip_off & IP_DF) == 0)) {
544 /*
545 * Restore packet header fields to original values
546 */
547 ip->ip_len = htons(ip->ip_len);
548 ip->ip_off = htons(ip->ip_off);
549 /*
550 * Send off the packet via outgoing interface
551 */
552 error = (*ifp->if_output)(ifp, m,
553 (struct sockaddr *)dst, &ro);
554 } else {
555 /*
556 * Handle EMSGSIZE with icmp reply needfrag for TCP MTU discovery
557 */
558 if (ip->ip_off & IP_DF) {
559 IPSTAT_INC(ips_cantfrag);
560 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG,
561 0, mtu);
562 goto consumed;
563 } else {
564 /*
565 * We have to fragment the packet
566 */
567 m->m_pkthdr.csum_flags |= CSUM_IP;
568 /*
569 * ip_fragment expects ip_len and ip_off in host byte
570 * order but returns all packets in network byte order
571 */
572 if (ip_fragment(ip, &m, mtu, ifp->if_hwassist,
573 (~ifp->if_hwassist & CSUM_DELAY_IP))) {
574 goto drop;
575 }
576 KASSERT(m != NULL, ("null mbuf and no error"));
577 /*
578 * Send off the fragments via outgoing interface
579 */
580 error = 0;
581 do {
582 m0 = m->m_nextpkt;
583 m->m_nextpkt = NULL;
584
585 error = (*ifp->if_output)(ifp, m,
586 (struct sockaddr *)dst, &ro);
587 if (error)
588 break;
589 } while ((m = m0) != NULL);
590 if (error) {
591 /* Reclaim remaining fragments */
592 for (m = m0; m; m = m0) {
593 m0 = m->m_nextpkt;
594 m_freem(m);
595 }
596 } else
597 IPSTAT_INC(ips_fragmented);
598 }
599 }
600
601 if (error != 0)
602 IPSTAT_INC(ips_odropped);
603 else {
604 ro.ro_rt->rt_rmx.rmx_pksent++;
605 IPSTAT_INC(ips_forward);
606 IPSTAT_INC(ips_fastforward);
607 }
608consumed:
609 RTFREE(ro.ro_rt);
610 return NULL;
611drop:
612 if (m)
613 m_freem(m);
614 if (ro.ro_rt)
615 RTFREE(ro.ro_rt);
616 return NULL;
617}
| 110#define V_ipfastforward_active VNET(ipfastforward_active)
111
112SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fastforwarding, CTLFLAG_RW,
113 &VNET_NAME(ipfastforward_active), 0, "Enable fast IP forwarding");
114
115static struct sockaddr_in *
116ip_findroute(struct route *ro, struct in_addr dest, struct mbuf *m)
117{
118 struct sockaddr_in *dst;
119 struct rtentry *rt;
120
121 /*
122 * Find route to destination.
123 */
124 bzero(ro, sizeof(*ro));
125 dst = (struct sockaddr_in *)&ro->ro_dst;
126 dst->sin_family = AF_INET;
127 dst->sin_len = sizeof(*dst);
128 dst->sin_addr.s_addr = dest.s_addr;
129 in_rtalloc_ign(ro, 0, M_GETFIB(m));
130
131 /*
132 * Route there and interface still up?
133 */
134 rt = ro->ro_rt;
135 if (rt && (rt->rt_flags & RTF_UP) &&
136 (rt->rt_ifp->if_flags & IFF_UP) &&
137 (rt->rt_ifp->if_drv_flags & IFF_DRV_RUNNING)) {
138 if (rt->rt_flags & RTF_GATEWAY)
139 dst = (struct sockaddr_in *)rt->rt_gateway;
140 } else {
141 IPSTAT_INC(ips_noroute);
142 IPSTAT_INC(ips_cantforward);
143 if (rt)
144 RTFREE(rt);
145 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
146 return NULL;
147 }
148 return dst;
149}
150
151/*
152 * Try to forward a packet based on the destination address.
153 * This is a fast path optimized for the plain forwarding case.
154 * If the packet is handled (and consumed) here then we return NULL;
155 * otherwise mbuf is returned and the packet should be delivered
156 * to ip_input for full processing.
157 */
158struct mbuf *
159ip_fastforward(struct mbuf *m)
160{
161 struct ip *ip;
162 struct mbuf *m0 = NULL;
163 struct route ro;
164 struct sockaddr_in *dst = NULL;
165 struct ifnet *ifp;
166 struct in_addr odest, dest;
167 u_short sum, ip_len;
168 int error = 0;
169 int hlen, mtu;
170#ifdef IPFIREWALL_FORWARD
171 struct m_tag *fwd_tag;
172#endif
173
174 /*
175 * Are we active and forwarding packets?
176 */
177 if (!V_ipfastforward_active || !V_ipforwarding)
178 return m;
179
180 M_ASSERTVALID(m);
181 M_ASSERTPKTHDR(m);
182
183 bzero(&ro, sizeof(ro));
184
185 /*
186 * Step 1: check for packet drop conditions (and sanity checks)
187 */
188
189 /*
190 * Is entire packet big enough?
191 */
192 if (m->m_pkthdr.len < sizeof(struct ip)) {
193 IPSTAT_INC(ips_tooshort);
194 goto drop;
195 }
196
197 /*
198 * Is first mbuf large enough for ip header and is header present?
199 */
200 if (m->m_len < sizeof (struct ip) &&
201 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
202 IPSTAT_INC(ips_toosmall);
203 return NULL; /* mbuf already free'd */
204 }
205
206 ip = mtod(m, struct ip *);
207
208 /*
209 * Is it IPv4?
210 */
211 if (ip->ip_v != IPVERSION) {
212 IPSTAT_INC(ips_badvers);
213 goto drop;
214 }
215
216 /*
217 * Is IP header length correct and is it in first mbuf?
218 */
219 hlen = ip->ip_hl << 2;
220 if (hlen < sizeof(struct ip)) { /* minimum header length */
221 IPSTAT_INC(ips_badlen);
222 goto drop;
223 }
224 if (hlen > m->m_len) {
225 if ((m = m_pullup(m, hlen)) == NULL) {
226 IPSTAT_INC(ips_badhlen);
227 return NULL; /* mbuf already free'd */
228 }
229 ip = mtod(m, struct ip *);
230 }
231
232 /*
233 * Checksum correct?
234 */
235 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED)
236 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
237 else {
238 if (hlen == sizeof(struct ip))
239 sum = in_cksum_hdr(ip);
240 else
241 sum = in_cksum(m, hlen);
242 }
243 if (sum) {
244 IPSTAT_INC(ips_badsum);
245 goto drop;
246 }
247
248 /*
249 * Remember that we have checked the IP header and found it valid.
250 */
251 m->m_pkthdr.csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
252
253 ip_len = ntohs(ip->ip_len);
254
255 /*
256 * Is IP length longer than packet we have got?
257 */
258 if (m->m_pkthdr.len < ip_len) {
259 IPSTAT_INC(ips_tooshort);
260 goto drop;
261 }
262
263 /*
264 * Is packet longer than IP header tells us? If yes, truncate packet.
265 */
266 if (m->m_pkthdr.len > ip_len) {
267 if (m->m_len == m->m_pkthdr.len) {
268 m->m_len = ip_len;
269 m->m_pkthdr.len = ip_len;
270 } else
271 m_adj(m, ip_len - m->m_pkthdr.len);
272 }
273
274 /*
275 * Is packet from or to 127/8?
276 */
277 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
278 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
279 IPSTAT_INC(ips_badaddr);
280 goto drop;
281 }
282
283#ifdef ALTQ
284 /*
285 * Is packet dropped by traffic conditioner?
286 */
287 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
288 goto drop;
289#endif
290
291 /*
292 * Step 2: fallback conditions to normal ip_input path processing
293 */
294
295 /*
296 * Only IP packets without options
297 */
298 if (ip->ip_hl != (sizeof(struct ip) >> 2)) {
299 if (ip_doopts == 1)
300 return m;
301 else if (ip_doopts == 2) {
302 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_FILTER_PROHIB,
303 0, 0);
304 return NULL; /* mbuf already free'd */
305 }
306 /* else ignore IP options and continue */
307 }
308
309 /*
310 * Only unicast IP, not from loopback, no L2 or IP broadcast,
311 * no multicast, no INADDR_ANY
312 *
313 * XXX: Probably some of these checks could be direct drop
314 * conditions. However it is not clear whether there are some
315 * hacks or obscure behaviours which make it neccessary to
316 * let ip_input handle it. We play safe here and let ip_input
317 * deal with it until it is proven that we can directly drop it.
318 */
319 if ((m->m_flags & (M_BCAST|M_MCAST)) ||
320 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) ||
321 ntohl(ip->ip_src.s_addr) == (u_long)INADDR_BROADCAST ||
322 ntohl(ip->ip_dst.s_addr) == (u_long)INADDR_BROADCAST ||
323 IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
324 IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
325 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) ||
326 IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
327 ip->ip_src.s_addr == INADDR_ANY ||
328 ip->ip_dst.s_addr == INADDR_ANY )
329 return m;
330
331 /*
332 * Is it for a local address on this host?
333 */
334 if (in_localip(ip->ip_dst))
335 return m;
336
337 IPSTAT_INC(ips_total);
338
339 /*
340 * Step 3: incoming packet firewall processing
341 */
342
343 /*
344 * Convert to host representation
345 */
346 ip->ip_len = ntohs(ip->ip_len);
347 ip->ip_off = ntohs(ip->ip_off);
348
349 odest.s_addr = dest.s_addr = ip->ip_dst.s_addr;
350
351 /*
352 * Run through list of ipfilter hooks for input packets
353 */
354 if (!PFIL_HOOKED(&V_inet_pfil_hook))
355 goto passin;
356
357 if (pfil_run_hooks(
358 &V_inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL) ||
359 m == NULL)
360 goto drop;
361
362 M_ASSERTVALID(m);
363 M_ASSERTPKTHDR(m);
364
365 ip = mtod(m, struct ip *); /* m may have changed by pfil hook */
366 dest.s_addr = ip->ip_dst.s_addr;
367
368 /*
369 * Destination address changed?
370 */
371 if (odest.s_addr != dest.s_addr) {
372 /*
373 * Is it now for a local address on this host?
374 */
375 if (in_localip(dest))
376 goto forwardlocal;
377 /*
378 * Go on with new destination address
379 */
380 }
381#ifdef IPFIREWALL_FORWARD
382 if (m->m_flags & M_FASTFWD_OURS) {
383 /*
384 * ipfw changed it for a local address on this host.
385 */
386 goto forwardlocal;
387 }
388#endif /* IPFIREWALL_FORWARD */
389
390passin:
391 /*
392 * Step 4: decrement TTL and look up route
393 */
394
395 /*
396 * Check TTL
397 */
398#ifdef IPSTEALTH
399 if (!V_ipstealth) {
400#endif
401 if (ip->ip_ttl <= IPTTLDEC) {
402 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
403 return NULL; /* mbuf already free'd */
404 }
405
406 /*
407 * Decrement the TTL and incrementally change the IP header checksum.
408 * Don't bother doing this with hw checksum offloading, it's faster
409 * doing it right here.
410 */
411 ip->ip_ttl -= IPTTLDEC;
412 if (ip->ip_sum >= (u_int16_t) ~htons(IPTTLDEC << 8))
413 ip->ip_sum -= ~htons(IPTTLDEC << 8);
414 else
415 ip->ip_sum += htons(IPTTLDEC << 8);
416#ifdef IPSTEALTH
417 }
418#endif
419
420 /*
421 * Find route to destination.
422 */
423 if ((dst = ip_findroute(&ro, dest, m)) == NULL)
424 return NULL; /* icmp unreach already sent */
425 ifp = ro.ro_rt->rt_ifp;
426
427 /*
428 * Immediately drop blackholed traffic, and directed broadcasts
429 * for either the all-ones or all-zero subnet addresses on
430 * locally attached networks.
431 */
432 if ((ro.ro_rt->rt_flags & (RTF_BLACKHOLE|RTF_BROADCAST)) != 0)
433 goto drop;
434
435 /*
436 * Step 5: outgoing firewall packet processing
437 */
438
439 /*
440 * Run through list of hooks for output packets.
441 */
442 if (!PFIL_HOOKED(&V_inet_pfil_hook))
443 goto passout;
444
445 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, NULL) || m == NULL) {
446 goto drop;
447 }
448
449 M_ASSERTVALID(m);
450 M_ASSERTPKTHDR(m);
451
452 ip = mtod(m, struct ip *);
453 dest.s_addr = ip->ip_dst.s_addr;
454
455 /*
456 * Destination address changed?
457 */
458#ifndef IPFIREWALL_FORWARD
459 if (odest.s_addr != dest.s_addr) {
460#else
461 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
462 if (odest.s_addr != dest.s_addr || fwd_tag != NULL) {
463#endif /* IPFIREWALL_FORWARD */
464 /*
465 * Is it now for a local address on this host?
466 */
467#ifndef IPFIREWALL_FORWARD
468 if (in_localip(dest)) {
469#else
470 if (m->m_flags & M_FASTFWD_OURS || in_localip(dest)) {
471#endif /* IPFIREWALL_FORWARD */
472forwardlocal:
473 /*
474 * Return packet for processing by ip_input().
475 * Keep host byte order as expected at ip_input's
476 * "ours"-label.
477 */
478 m->m_flags |= M_FASTFWD_OURS;
479 if (ro.ro_rt)
480 RTFREE(ro.ro_rt);
481 return m;
482 }
483 /*
484 * Redo route lookup with new destination address
485 */
486#ifdef IPFIREWALL_FORWARD
487 if (fwd_tag) {
488 dest.s_addr = ((struct sockaddr_in *)
489 (fwd_tag + 1))->sin_addr.s_addr;
490 m_tag_delete(m, fwd_tag);
491 }
492#endif /* IPFIREWALL_FORWARD */
493 RTFREE(ro.ro_rt);
494 if ((dst = ip_findroute(&ro, dest, m)) == NULL)
495 return NULL; /* icmp unreach already sent */
496 ifp = ro.ro_rt->rt_ifp;
497 }
498
499passout:
500 /*
501 * Step 6: send off the packet
502 */
503
504 /*
505 * Check if route is dampned (when ARP is unable to resolve)
506 */
507 if ((ro.ro_rt->rt_flags & RTF_REJECT) &&
508 (ro.ro_rt->rt_rmx.rmx_expire == 0 ||
509 time_uptime < ro.ro_rt->rt_rmx.rmx_expire)) {
510 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
511 goto consumed;
512 }
513
514#ifndef ALTQ
515 /*
516 * Check if there is enough space in the interface queue
517 */
518 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
519 ifp->if_snd.ifq_maxlen) {
520 IPSTAT_INC(ips_odropped);
521 /* would send source quench here but that is depreciated */
522 goto drop;
523 }
524#endif
525
526 /*
527 * Check if media link state of interface is not down
528 */
529 if (ifp->if_link_state == LINK_STATE_DOWN) {
530 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
531 goto consumed;
532 }
533
534 /*
535 * Check if packet fits MTU or if hardware will fragment for us
536 */
537 if (ro.ro_rt->rt_rmx.rmx_mtu)
538 mtu = min(ro.ro_rt->rt_rmx.rmx_mtu, ifp->if_mtu);
539 else
540 mtu = ifp->if_mtu;
541
542 if (ip->ip_len <= mtu ||
543 (ifp->if_hwassist & CSUM_FRAGMENT && (ip->ip_off & IP_DF) == 0)) {
544 /*
545 * Restore packet header fields to original values
546 */
547 ip->ip_len = htons(ip->ip_len);
548 ip->ip_off = htons(ip->ip_off);
549 /*
550 * Send off the packet via outgoing interface
551 */
552 error = (*ifp->if_output)(ifp, m,
553 (struct sockaddr *)dst, &ro);
554 } else {
555 /*
556 * Handle EMSGSIZE with icmp reply needfrag for TCP MTU discovery
557 */
558 if (ip->ip_off & IP_DF) {
559 IPSTAT_INC(ips_cantfrag);
560 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG,
561 0, mtu);
562 goto consumed;
563 } else {
564 /*
565 * We have to fragment the packet
566 */
567 m->m_pkthdr.csum_flags |= CSUM_IP;
568 /*
569 * ip_fragment expects ip_len and ip_off in host byte
570 * order but returns all packets in network byte order
571 */
572 if (ip_fragment(ip, &m, mtu, ifp->if_hwassist,
573 (~ifp->if_hwassist & CSUM_DELAY_IP))) {
574 goto drop;
575 }
576 KASSERT(m != NULL, ("null mbuf and no error"));
577 /*
578 * Send off the fragments via outgoing interface
579 */
580 error = 0;
581 do {
582 m0 = m->m_nextpkt;
583 m->m_nextpkt = NULL;
584
585 error = (*ifp->if_output)(ifp, m,
586 (struct sockaddr *)dst, &ro);
587 if (error)
588 break;
589 } while ((m = m0) != NULL);
590 if (error) {
591 /* Reclaim remaining fragments */
592 for (m = m0; m; m = m0) {
593 m0 = m->m_nextpkt;
594 m_freem(m);
595 }
596 } else
597 IPSTAT_INC(ips_fragmented);
598 }
599 }
600
601 if (error != 0)
602 IPSTAT_INC(ips_odropped);
603 else {
604 ro.ro_rt->rt_rmx.rmx_pksent++;
605 IPSTAT_INC(ips_forward);
606 IPSTAT_INC(ips_fastforward);
607 }
608consumed:
609 RTFREE(ro.ro_rt);
610 return NULL;
611drop:
612 if (m)
613 m_freem(m);
614 if (ro.ro_rt)
615 RTFREE(ro.ro_rt);
616 return NULL;
617}
|