1 /*-
2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. 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 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
30 * $FreeBSD: releng/6.3/sys/netinet/ip_input.c 168291 2007-04-03 08:16:05Z julian $
31 */
32
33 #include "opt_bootp.h"
34 #include "opt_ipfw.h"
35 #include "opt_ipstealth.h"
36 #include "opt_ipsec.h"
37 #include "opt_mac.h"
38 #include "opt_carp.h"
39
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/callout.h>
43 #include <sys/mac.h>
44 #include <sys/mbuf.h>
45 #include <sys/malloc.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/syslog.h>
52 #include <sys/sysctl.h>
53
54 #include <net/pfil.h>
55 #include <net/if.h>
56 #include <net/if_types.h>
57 #include <net/if_var.h>
58 #include <net/if_dl.h>
59 #include <net/route.h>
60 #include <net/netisr.h>
61
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/in_var.h>
65 #include <netinet/ip.h>
66 #include <netinet/in_pcb.h>
67 #include <netinet/ip_var.h>
68 #include <netinet/ip_icmp.h>
69 #include <machine/in_cksum.h>
70 #ifdef DEV_CARP
71 #include <netinet/ip_carp.h>
72 #endif
73
74 #include <sys/socketvar.h>
75
76 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
77 #include <netinet/ip_fw.h>
78 #include <netinet/ip_dummynet.h>
79
80 #ifdef IPSEC
81 #include <netinet6/ipsec.h>
82 #include <netkey/key.h>
83 #endif
84
85 #ifdef FAST_IPSEC
86 #include <netipsec/ipsec.h>
87 #include <netipsec/key.h>
88 #endif
89
90 int rsvp_on = 0;
91
92 int ipforwarding = 0;
93 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
94 &ipforwarding, 0, "Enable IP forwarding between interfaces");
95
96 static int ipsendredirects = 1; /* XXX */
97 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
98 &ipsendredirects, 0, "Enable sending IP redirects");
99
100 int ip_defttl = IPDEFTTL;
101 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
102 &ip_defttl, 0, "Maximum TTL on IP packets");
103
104 static int ip_dosourceroute = 0;
105 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
106 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
107
108 static int ip_acceptsourceroute = 0;
109 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
110 CTLFLAG_RW, &ip_acceptsourceroute, 0,
111 "Enable accepting source routed IP packets");
112
113 int ip_doopts = 1; /* 0 = ignore, 1 = process, 2 = reject */
114 SYSCTL_INT(_net_inet_ip, OID_AUTO, process_options, CTLFLAG_RW,
115 &ip_doopts, 0, "Enable IP options processing ([LS]SRR, RR, TS)");
116
117 static int ip_keepfaith = 0;
118 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
119 &ip_keepfaith, 0,
120 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
121
122 static int ip_sendsourcequench = 0;
123 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
124 &ip_sendsourcequench, 0,
125 "Enable the transmission of source quench packets");
126
127 int ip_do_randomid = 0;
128 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
129 &ip_do_randomid, 0,
130 "Assign random ip_id values");
131
132 /*
133 * XXX - Setting ip_checkinterface mostly implements the receive side of
134 * the Strong ES model described in RFC 1122, but since the routing table
135 * and transmit implementation do not implement the Strong ES model,
136 * setting this to 1 results in an odd hybrid.
137 *
138 * XXX - ip_checkinterface currently must be disabled if you use ipnat
139 * to translate the destination address to another local interface.
140 *
141 * XXX - ip_checkinterface must be disabled if you add IP aliases
142 * to the loopback interface instead of the interface where the
143 * packets for those addresses are received.
144 */
145 static int ip_checkinterface = 0;
146 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
147 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
148
149 #ifdef DIAGNOSTIC
150 static int ipprintfs = 0;
151 #endif
152
153 struct pfil_head inet_pfil_hook; /* Packet filter hooks */
154
155 static struct ifqueue ipintrq;
156 static int ipqmaxlen = IFQ_MAXLEN;
157
158 extern struct domain inetdomain;
159 extern struct protosw inetsw[];
160 u_char ip_protox[IPPROTO_MAX];
161 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
162 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
163 u_long in_ifaddrhmask; /* mask for hash table */
164
165 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
166 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
167 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
168 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
169
170 struct ipstat ipstat;
171 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
172 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
173
174 /*
175 * IP datagram reassembly.
176 */
177 #define IPREASS_NHASH_LOG2 6
178 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
179 #define IPREASS_HMASK (IPREASS_NHASH - 1)
180 #define IPREASS_HASH(x,y) \
181 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
182
183 static uma_zone_t ipq_zone;
184 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
185 static struct mtx ipqlock;
186
187 #define IPQ_LOCK() mtx_lock(&ipqlock)
188 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
189 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
190 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
191
192 static void maxnipq_update(void);
193 static void ipq_zone_change(void *);
194
195 static int maxnipq; /* Administrative limit on # reass queues. */
196 static int nipq = 0; /* Total # of reass queues */
197 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, &nipq, 0,
198 "Current number of IPv4 fragment reassembly queue entries");
199
200 static int maxfragsperpacket;
201 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
202 &maxfragsperpacket, 0,
203 "Maximum number of IPv4 fragments allowed per packet");
204
205 struct callout ipport_tick_callout;
206
207 #ifdef IPCTL_DEFMTU
208 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
209 &ip_mtu, 0, "Default MTU");
210 #endif
211
212 #ifdef IPSTEALTH
213 int ipstealth = 0;
214 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
215 &ipstealth, 0, "");
216 #endif
217
218 /*
219 * ipfw_ether and ipfw_bridge hooks.
220 * XXX: Temporary until those are converted to pfil_hooks as well.
221 */
222 ip_fw_chk_t *ip_fw_chk_ptr = NULL;
223 ip_dn_io_t *ip_dn_io_ptr = NULL;
224 int fw_enable = 1;
225 int fw_one_pass = 1;
226
227 /*
228 * XXX this is ugly. IP options source routing magic.
229 */
230 struct ipoptrt {
231 struct in_addr dst; /* final destination */
232 char nop; /* one NOP to align */
233 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
234 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
235 };
236
237 struct ipopt_tag {
238 struct m_tag tag;
239 int ip_nhops;
240 struct ipoptrt ip_srcrt;
241 };
242
243 static void save_rte(struct mbuf *, u_char *, struct in_addr);
244 static int ip_dooptions(struct mbuf *m, int);
245 static void ip_forward(struct mbuf *m, int srcrt);
246 static void ip_freef(struct ipqhead *, struct ipq *);
247
248 /*
249 * IP initialization: fill in IP protocol switch table.
250 * All protocols not implemented in kernel go to raw IP protocol handler.
251 */
252 void
253 ip_init()
254 {
255 register struct protosw *pr;
256 register int i;
257
258 TAILQ_INIT(&in_ifaddrhead);
259 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
260 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
261 if (pr == NULL)
262 panic("ip_init: PF_INET not found");
263
264 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
265 for (i = 0; i < IPPROTO_MAX; i++)
266 ip_protox[i] = pr - inetsw;
267 /*
268 * Cycle through IP protocols and put them into the appropriate place
269 * in ip_protox[].
270 */
271 for (pr = inetdomain.dom_protosw;
272 pr < inetdomain.dom_protoswNPROTOSW; pr++)
273 if (pr->pr_domain->dom_family == PF_INET &&
274 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
275 /* Be careful to only index valid IP protocols. */
276 if (pr->pr_protocol < IPPROTO_MAX)
277 ip_protox[pr->pr_protocol] = pr - inetsw;
278 }
279
280 /* Initialize packet filter hooks. */
281 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
282 inet_pfil_hook.ph_af = AF_INET;
283 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
284 printf("%s: WARNING: unable to register pfil hook, "
285 "error %d\n", __func__, i);
286
287 /* Initialize IP reassembly queue. */
288 IPQ_LOCK_INIT();
289 for (i = 0; i < IPREASS_NHASH; i++)
290 TAILQ_INIT(&ipq[i]);
291 maxnipq = nmbclusters / 32;
292 maxfragsperpacket = 16;
293 ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
294 NULL, UMA_ALIGN_PTR, 0);
295 maxnipq_update();
296
297 /* Start ipport_tick. */
298 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
299 ipport_tick(NULL);
300 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
301 SHUTDOWN_PRI_DEFAULT);
302 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
303 NULL, EVENTHANDLER_PRI_ANY);
304
305 /* Initialize various other remaining things. */
306 ip_id = time_second & 0xffff;
307 ipintrq.ifq_maxlen = ipqmaxlen;
308 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
309 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
310 }
311
312 void ip_fini(xtp)
313 void *xtp;
314 {
315 callout_stop(&ipport_tick_callout);
316 }
317
318 /*
319 * Ip input routine. Checksum and byte swap header. If fragmented
320 * try to reassemble. Process options. Pass to next level.
321 */
322 void
323 ip_input(struct mbuf *m)
324 {
325 struct ip *ip = NULL;
326 struct in_ifaddr *ia = NULL;
327 struct ifaddr *ifa;
328 int checkif, hlen = 0;
329 u_short sum;
330 int dchg = 0; /* dest changed after fw */
331 struct in_addr odst; /* original dst address */
332 #ifdef FAST_IPSEC
333 struct m_tag *mtag;
334 struct tdb_ident *tdbi;
335 struct secpolicy *sp;
336 int s, error;
337 #endif /* FAST_IPSEC */
338
339 M_ASSERTPKTHDR(m);
340
341 if (m->m_flags & M_FASTFWD_OURS) {
342 /*
343 * Firewall or NAT changed destination to local.
344 * We expect ip_len and ip_off to be in host byte order.
345 */
346 m->m_flags &= ~M_FASTFWD_OURS;
347 /* Set up some basics that will be used later. */
348 ip = mtod(m, struct ip *);
349 hlen = ip->ip_hl << 2;
350 goto ours;
351 }
352
353 ipstat.ips_total++;
354
355 if (m->m_pkthdr.len < sizeof(struct ip))
356 goto tooshort;
357
358 if (m->m_len < sizeof (struct ip) &&
359 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
360 ipstat.ips_toosmall++;
361 return;
362 }
363 ip = mtod(m, struct ip *);
364
365 if (ip->ip_v != IPVERSION) {
366 ipstat.ips_badvers++;
367 goto bad;
368 }
369
370 hlen = ip->ip_hl << 2;
371 if (hlen < sizeof(struct ip)) { /* minimum header length */
372 ipstat.ips_badhlen++;
373 goto bad;
374 }
375 if (hlen > m->m_len) {
376 if ((m = m_pullup(m, hlen)) == NULL) {
377 ipstat.ips_badhlen++;
378 return;
379 }
380 ip = mtod(m, struct ip *);
381 }
382
383 /* 127/8 must not appear on wire - RFC1122 */
384 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
385 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
386 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
387 ipstat.ips_badaddr++;
388 goto bad;
389 }
390 }
391
392 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
393 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
394 } else {
395 if (hlen == sizeof(struct ip)) {
396 sum = in_cksum_hdr(ip);
397 } else {
398 sum = in_cksum(m, hlen);
399 }
400 }
401 if (sum) {
402 ipstat.ips_badsum++;
403 goto bad;
404 }
405
406 #ifdef ALTQ
407 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
408 /* packet is dropped by traffic conditioner */
409 return;
410 #endif
411
412 /*
413 * Convert fields to host representation.
414 */
415 ip->ip_len = ntohs(ip->ip_len);
416 if (ip->ip_len < hlen) {
417 ipstat.ips_badlen++;
418 goto bad;
419 }
420 ip->ip_off = ntohs(ip->ip_off);
421
422 /*
423 * Check that the amount of data in the buffers
424 * is as at least much as the IP header would have us expect.
425 * Trim mbufs if longer than we expect.
426 * Drop packet if shorter than we expect.
427 */
428 if (m->m_pkthdr.len < ip->ip_len) {
429 tooshort:
430 ipstat.ips_tooshort++;
431 goto bad;
432 }
433 if (m->m_pkthdr.len > ip->ip_len) {
434 if (m->m_len == m->m_pkthdr.len) {
435 m->m_len = ip->ip_len;
436 m->m_pkthdr.len = ip->ip_len;
437 } else
438 m_adj(m, ip->ip_len - m->m_pkthdr.len);
439 }
440 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
441 /*
442 * Bypass packet filtering for packets from a tunnel (gif).
443 */
444 if (ipsec_getnhist(m))
445 goto passin;
446 #endif
447 #if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF)
448 /*
449 * Bypass packet filtering for packets from a tunnel (gif).
450 */
451 if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL)
452 goto passin;
453 #endif
454
455 /*
456 * Run through list of hooks for input packets.
457 *
458 * NB: Beware of the destination address changing (e.g.
459 * by NAT rewriting). When this happens, tell
460 * ip_forward to do the right thing.
461 */
462
463 /* Jump over all PFIL processing if hooks are not active. */
464 if (inet_pfil_hook.ph_busy_count == -1)
465 goto passin;
466
467 odst = ip->ip_dst;
468 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
469 PFIL_IN, NULL) != 0)
470 return;
471 if (m == NULL) /* consumed by filter */
472 return;
473
474 ip = mtod(m, struct ip *);
475 dchg = (odst.s_addr != ip->ip_dst.s_addr);
476
477 #ifdef IPFIREWALL_FORWARD
478 if (m->m_flags & M_FASTFWD_OURS) {
479 m->m_flags &= ~M_FASTFWD_OURS;
480 goto ours;
481 }
482 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
483 /*
484 * Directly ship on the packet. This allows to forward packets
485 * that were destined for us to some other directly connected
486 * host.
487 */
488 ip_forward(m, dchg);
489 return;
490 }
491 #endif /* IPFIREWALL_FORWARD */
492
493 passin:
494 /*
495 * Process options and, if not destined for us,
496 * ship it on. ip_dooptions returns 1 when an
497 * error was detected (causing an icmp message
498 * to be sent and the original packet to be freed).
499 */
500 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
501 return;
502
503 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
504 * matter if it is destined to another node, or whether it is
505 * a multicast one, RSVP wants it! and prevents it from being forwarded
506 * anywhere else. Also checks if the rsvp daemon is running before
507 * grabbing the packet.
508 */
509 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
510 goto ours;
511
512 /*
513 * Check our list of addresses, to see if the packet is for us.
514 * If we don't have any addresses, assume any unicast packet
515 * we receive might be for us (and let the upper layers deal
516 * with it).
517 */
518 if (TAILQ_EMPTY(&in_ifaddrhead) &&
519 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
520 goto ours;
521
522 /*
523 * Enable a consistency check between the destination address
524 * and the arrival interface for a unicast packet (the RFC 1122
525 * strong ES model) if IP forwarding is disabled and the packet
526 * is not locally generated and the packet is not subject to
527 * 'ipfw fwd'.
528 *
529 * XXX - Checking also should be disabled if the destination
530 * address is ipnat'ed to a different interface.
531 *
532 * XXX - Checking is incompatible with IP aliases added
533 * to the loopback interface instead of the interface where
534 * the packets are received.
535 *
536 * XXX - This is the case for carp vhost IPs as well so we
537 * insert a workaround. If the packet got here, we already
538 * checked with carp_iamatch() and carp_forus().
539 */
540 checkif = ip_checkinterface && (ipforwarding == 0) &&
541 m->m_pkthdr.rcvif != NULL &&
542 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
543 #ifdef DEV_CARP
544 !m->m_pkthdr.rcvif->if_carp &&
545 #endif
546 (dchg == 0);
547
548 /*
549 * Check for exact addresses in the hash bucket.
550 */
551 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
552 /*
553 * If the address matches, verify that the packet
554 * arrived via the correct interface if checking is
555 * enabled.
556 */
557 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
558 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
559 goto ours;
560 }
561 /*
562 * Check for broadcast addresses.
563 *
564 * Only accept broadcast packets that arrive via the matching
565 * interface. Reception of forwarded directed broadcasts would
566 * be handled via ip_forward() and ether_output() with the loopback
567 * into the stack for SIMPLEX interfaces handled by ether_output().
568 */
569 if (m->m_pkthdr.rcvif != NULL &&
570 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
571 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
572 if (ifa->ifa_addr->sa_family != AF_INET)
573 continue;
574 ia = ifatoia(ifa);
575 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
576 ip->ip_dst.s_addr)
577 goto ours;
578 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
579 goto ours;
580 #ifdef BOOTP_COMPAT
581 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
582 goto ours;
583 #endif
584 }
585 }
586 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
587 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
588 ipstat.ips_cantforward++;
589 m_freem(m);
590 return;
591 }
592 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
593 struct in_multi *inm;
594 if (ip_mrouter) {
595 /*
596 * If we are acting as a multicast router, all
597 * incoming multicast packets are passed to the
598 * kernel-level multicast forwarding function.
599 * The packet is returned (relatively) intact; if
600 * ip_mforward() returns a non-zero value, the packet
601 * must be discarded, else it may be accepted below.
602 */
603 if (ip_mforward &&
604 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
605 ipstat.ips_cantforward++;
606 m_freem(m);
607 return;
608 }
609
610 /*
611 * The process-level routing daemon needs to receive
612 * all multicast IGMP packets, whether or not this
613 * host belongs to their destination groups.
614 */
615 if (ip->ip_p == IPPROTO_IGMP)
616 goto ours;
617 ipstat.ips_forward++;
618 }
619 /*
620 * See if we belong to the destination multicast group on the
621 * arrival interface.
622 */
623 IN_MULTI_LOCK();
624 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
625 IN_MULTI_UNLOCK();
626 if (inm == NULL) {
627 ipstat.ips_notmember++;
628 m_freem(m);
629 return;
630 }
631 goto ours;
632 }
633 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
634 goto ours;
635 if (ip->ip_dst.s_addr == INADDR_ANY)
636 goto ours;
637
638 /*
639 * FAITH(Firewall Aided Internet Translator)
640 */
641 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
642 if (ip_keepfaith) {
643 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
644 goto ours;
645 }
646 m_freem(m);
647 return;
648 }
649
650 /*
651 * Not for us; forward if possible and desirable.
652 */
653 if (ipforwarding == 0) {
654 ipstat.ips_cantforward++;
655 m_freem(m);
656 } else {
657 #ifdef IPSEC
658 /*
659 * Enforce inbound IPsec SPD.
660 */
661 if (ipsec4_in_reject(m, NULL)) {
662 ipsecstat.in_polvio++;
663 goto bad;
664 }
665 #endif /* IPSEC */
666 #ifdef FAST_IPSEC
667 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
668 s = splnet();
669 if (mtag != NULL) {
670 tdbi = (struct tdb_ident *)(mtag + 1);
671 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
672 } else {
673 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
674 IP_FORWARDING, &error);
675 }
676 if (sp == NULL) { /* NB: can happen if error */
677 splx(s);
678 /*XXX error stat???*/
679 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
680 goto bad;
681 }
682
683 /*
684 * Check security policy against packet attributes.
685 */
686 error = ipsec_in_reject(sp, m);
687 KEY_FREESP(&sp);
688 splx(s);
689 if (error) {
690 ipstat.ips_cantforward++;
691 goto bad;
692 }
693 #endif /* FAST_IPSEC */
694 ip_forward(m, dchg);
695 }
696 return;
697
698 ours:
699 #ifdef IPSTEALTH
700 /*
701 * IPSTEALTH: Process non-routing options only
702 * if the packet is destined for us.
703 */
704 if (ipstealth && hlen > sizeof (struct ip) &&
705 ip_dooptions(m, 1))
706 return;
707 #endif /* IPSTEALTH */
708
709 /* Count the packet in the ip address stats */
710 if (ia != NULL) {
711 ia->ia_ifa.if_ipackets++;
712 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
713 }
714
715 /*
716 * Attempt reassembly; if it succeeds, proceed.
717 * ip_reass() will return a different mbuf.
718 */
719 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
720 m = ip_reass(m);
721 if (m == NULL)
722 return;
723 ip = mtod(m, struct ip *);
724 /* Get the header length of the reassembled packet */
725 hlen = ip->ip_hl << 2;
726 }
727
728 /*
729 * Further protocols expect the packet length to be w/o the
730 * IP header.
731 */
732 ip->ip_len -= hlen;
733
734 #ifdef IPSEC
735 /*
736 * enforce IPsec policy checking if we are seeing last header.
737 * note that we do not visit this with protocols with pcb layer
738 * code - like udp/tcp/raw ip.
739 */
740 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
741 ipsec4_in_reject(m, NULL)) {
742 ipsecstat.in_polvio++;
743 goto bad;
744 }
745 #endif
746 #if FAST_IPSEC
747 /*
748 * enforce IPsec policy checking if we are seeing last header.
749 * note that we do not visit this with protocols with pcb layer
750 * code - like udp/tcp/raw ip.
751 */
752 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
753 /*
754 * Check if the packet has already had IPsec processing
755 * done. If so, then just pass it along. This tag gets
756 * set during AH, ESP, etc. input handling, before the
757 * packet is returned to the ip input queue for delivery.
758 */
759 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
760 s = splnet();
761 if (mtag != NULL) {
762 tdbi = (struct tdb_ident *)(mtag + 1);
763 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
764 } else {
765 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
766 IP_FORWARDING, &error);
767 }
768 if (sp != NULL) {
769 /*
770 * Check security policy against packet attributes.
771 */
772 error = ipsec_in_reject(sp, m);
773 KEY_FREESP(&sp);
774 } else {
775 /* XXX error stat??? */
776 error = EINVAL;
777 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
778 goto bad;
779 }
780 splx(s);
781 if (error)
782 goto bad;
783 }
784 #endif /* FAST_IPSEC */
785
786 /*
787 * Switch out to protocol's input routine.
788 */
789 ipstat.ips_delivered++;
790
791 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
792 return;
793 bad:
794 m_freem(m);
795 }
796
797 /*
798 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
799 * max has slightly different semantics than the sysctl, for historical
800 * reasons.
801 */
802 static void
803 maxnipq_update(void)
804 {
805
806 /*
807 * -1 for unlimited allocation.
808 */
809 if (maxnipq < 0)
810 uma_zone_set_max(ipq_zone, 0);
811 /*
812 * Positive number for specific bound.
813 */
814 if (maxnipq > 0)
815 uma_zone_set_max(ipq_zone, maxnipq);
816 /*
817 * Zero specifies no further fragment queue allocation -- set the
818 * bound very low, but rely on implementation elsewhere to actually
819 * prevent allocation and reclaim current queues.
820 */
821 if (maxnipq == 0)
822 uma_zone_set_max(ipq_zone, 1);
823 }
824
825 static void
826 ipq_zone_change(void *tag)
827 {
828
829 if (maxnipq > 0 && maxnipq < (nmbclusters / 32)) {
830 maxnipq = nmbclusters / 32;
831 maxnipq_update();
832 }
833 }
834
835 static int
836 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
837 {
838 int error, i;
839
840 i = maxnipq;
841 error = sysctl_handle_int(oidp, &i, 0, req);
842 if (error || !req->newptr)
843 return (error);
844
845 /*
846 * XXXRW: Might be a good idea to sanity check the argument and place
847 * an extreme upper bound.
848 */
849 if (i < -1)
850 return (EINVAL);
851 maxnipq = i;
852 maxnipq_update();
853 return (0);
854 }
855
856 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
857 NULL, 0, sysctl_maxnipq, "I",
858 "Maximum number of IPv4 fragment reassembly queue entries");
859
860 /*
861 * Take incoming datagram fragment and try to reassemble it into
862 * whole datagram. If the argument is the first fragment or one
863 * in between the function will return NULL and store the mbuf
864 * in the fragment chain. If the argument is the last fragment
865 * the packet will be reassembled and the pointer to the new
866 * mbuf returned for further processing. Only m_tags attached
867 * to the first packet/fragment are preserved.
868 * The IP header is *NOT* adjusted out of iplen.
869 */
870
871 struct mbuf *
872 ip_reass(struct mbuf *m)
873 {
874 struct ip *ip;
875 struct mbuf *p, *q, *nq, *t;
876 struct ipq *fp = NULL;
877 struct ipqhead *head;
878 int i, hlen, next;
879 u_int8_t ecn, ecn0;
880 u_short hash;
881
882 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
883 if (maxnipq == 0 || maxfragsperpacket == 0) {
884 ipstat.ips_fragments++;
885 ipstat.ips_fragdropped++;
886 m_freem(m);
887 return (NULL);
888 }
889
890 ip = mtod(m, struct ip *);
891 hlen = ip->ip_hl << 2;
892
893 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
894 head = &ipq[hash];
895 IPQ_LOCK();
896
897 /*
898 * Look for queue of fragments
899 * of this datagram.
900 */
901 TAILQ_FOREACH(fp, head, ipq_list)
902 if (ip->ip_id == fp->ipq_id &&
903 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
904 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
905 #ifdef MAC
906 mac_fragment_match(m, fp) &&
907 #endif
908 ip->ip_p == fp->ipq_p)
909 goto found;
910
911 fp = NULL;
912
913 /*
914 * Attempt to trim the number of allocated fragment queues if it
915 * exceeds the administrative limit.
916 */
917 if ((nipq > maxnipq) && (maxnipq > 0)) {
918 /*
919 * drop something from the tail of the current queue
920 * before proceeding further
921 */
922 struct ipq *q = TAILQ_LAST(head, ipqhead);
923 if (q == NULL) { /* gak */
924 for (i = 0; i < IPREASS_NHASH; i++) {
925 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
926 if (r) {
927 ipstat.ips_fragtimeout += r->ipq_nfrags;
928 ip_freef(&ipq[i], r);
929 break;
930 }
931 }
932 } else {
933 ipstat.ips_fragtimeout += q->ipq_nfrags;
934 ip_freef(head, q);
935 }
936 }
937
938 found:
939 /*
940 * Adjust ip_len to not reflect header,
941 * convert offset of this to bytes.
942 */
943 ip->ip_len -= hlen;
944 if (ip->ip_off & IP_MF) {
945 /*
946 * Make sure that fragments have a data length
947 * that's a non-zero multiple of 8 bytes.
948 */
949 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
950 ipstat.ips_toosmall++; /* XXX */
951 goto dropfrag;
952 }
953 m->m_flags |= M_FRAG;
954 } else
955 m->m_flags &= ~M_FRAG;
956 ip->ip_off <<= 3;
957
958
959 /*
960 * Attempt reassembly; if it succeeds, proceed.
961 * ip_reass() will return a different mbuf.
962 */
963 ipstat.ips_fragments++;
964 m->m_pkthdr.header = ip;
965
966 /* Previous ip_reass() started here. */
967 /*
968 * Presence of header sizes in mbufs
969 * would confuse code below.
970 */
971 m->m_data += hlen;
972 m->m_len -= hlen;
973
974 /*
975 * If first fragment to arrive, create a reassembly queue.
976 */
977 if (fp == NULL) {
978 fp = uma_zalloc(ipq_zone, M_NOWAIT);
979 if (fp == NULL)
980 goto dropfrag;
981 #ifdef MAC
982 if (mac_init_ipq(fp, M_NOWAIT) != 0) {
983 uma_zfree(ipq_zone, fp);
984 fp = NULL;
985 goto dropfrag;
986 }
987 mac_create_ipq(m, fp);
988 #endif
989 TAILQ_INSERT_HEAD(head, fp, ipq_list);
990 nipq++;
991 fp->ipq_nfrags = 1;
992 fp->ipq_ttl = IPFRAGTTL;
993 fp->ipq_p = ip->ip_p;
994 fp->ipq_id = ip->ip_id;
995 fp->ipq_src = ip->ip_src;
996 fp->ipq_dst = ip->ip_dst;
997 fp->ipq_frags = m;
998 m->m_nextpkt = NULL;
999 goto done;
1000 } else {
1001 fp->ipq_nfrags++;
1002 #ifdef MAC
1003 mac_update_ipq(m, fp);
1004 #endif
1005 }
1006
1007 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1008
1009 /*
1010 * Handle ECN by comparing this segment with the first one;
1011 * if CE is set, do not lose CE.
1012 * drop if CE and not-ECT are mixed for the same packet.
1013 */
1014 ecn = ip->ip_tos & IPTOS_ECN_MASK;
1015 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
1016 if (ecn == IPTOS_ECN_CE) {
1017 if (ecn0 == IPTOS_ECN_NOTECT)
1018 goto dropfrag;
1019 if (ecn0 != IPTOS_ECN_CE)
1020 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
1021 }
1022 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
1023 goto dropfrag;
1024
1025 /*
1026 * Find a segment which begins after this one does.
1027 */
1028 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1029 if (GETIP(q)->ip_off > ip->ip_off)
1030 break;
1031
1032 /*
1033 * If there is a preceding segment, it may provide some of
1034 * our data already. If so, drop the data from the incoming
1035 * segment. If it provides all of our data, drop us, otherwise
1036 * stick new segment in the proper place.
1037 *
1038 * If some of the data is dropped from the the preceding
1039 * segment, then it's checksum is invalidated.
1040 */
1041 if (p) {
1042 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1043 if (i > 0) {
1044 if (i >= ip->ip_len)
1045 goto dropfrag;
1046 m_adj(m, i);
1047 m->m_pkthdr.csum_flags = 0;
1048 ip->ip_off += i;
1049 ip->ip_len -= i;
1050 }
1051 m->m_nextpkt = p->m_nextpkt;
1052 p->m_nextpkt = m;
1053 } else {
1054 m->m_nextpkt = fp->ipq_frags;
1055 fp->ipq_frags = m;
1056 }
1057
1058 /*
1059 * While we overlap succeeding segments trim them or,
1060 * if they are completely covered, dequeue them.
1061 */
1062 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1063 q = nq) {
1064 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1065 if (i < GETIP(q)->ip_len) {
1066 GETIP(q)->ip_len -= i;
1067 GETIP(q)->ip_off += i;
1068 m_adj(q, i);
1069 q->m_pkthdr.csum_flags = 0;
1070 break;
1071 }
1072 nq = q->m_nextpkt;
1073 m->m_nextpkt = nq;
1074 ipstat.ips_fragdropped++;
1075 fp->ipq_nfrags--;
1076 m_freem(q);
1077 }
1078
1079 /*
1080 * Check for complete reassembly and perform frag per packet
1081 * limiting.
1082 *
1083 * Frag limiting is performed here so that the nth frag has
1084 * a chance to complete the packet before we drop the packet.
1085 * As a result, n+1 frags are actually allowed per packet, but
1086 * only n will ever be stored. (n = maxfragsperpacket.)
1087 *
1088 */
1089 next = 0;
1090 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1091 if (GETIP(q)->ip_off != next) {
1092 if (fp->ipq_nfrags > maxfragsperpacket) {
1093 ipstat.ips_fragdropped += fp->ipq_nfrags;
1094 ip_freef(head, fp);
1095 }
1096 goto done;
1097 }
1098 next += GETIP(q)->ip_len;
1099 }
1100 /* Make sure the last packet didn't have the IP_MF flag */
1101 if (p->m_flags & M_FRAG) {
1102 if (fp->ipq_nfrags > maxfragsperpacket) {
1103 ipstat.ips_fragdropped += fp->ipq_nfrags;
1104 ip_freef(head, fp);
1105 }
1106 goto done;
1107 }
1108
1109 /*
1110 * Reassembly is complete. Make sure the packet is a sane size.
1111 */
1112 q = fp->ipq_frags;
1113 ip = GETIP(q);
1114 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1115 ipstat.ips_toolong++;
1116 ipstat.ips_fragdropped += fp->ipq_nfrags;
1117 ip_freef(head, fp);
1118 goto done;
1119 }
1120
1121 /*
1122 * Concatenate fragments.
1123 */
1124 m = q;
1125 t = m->m_next;
1126 m->m_next = NULL;
1127 m_cat(m, t);
1128 nq = q->m_nextpkt;
1129 q->m_nextpkt = NULL;
1130 for (q = nq; q != NULL; q = nq) {
1131 nq = q->m_nextpkt;
1132 q->m_nextpkt = NULL;
1133 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1134 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1135 m_cat(m, q);
1136 }
1137 /*
1138 * In order to do checksumming faster we do 'end-around carry' here
1139 * (and not in for{} loop), though it implies we are not going to
1140 * reassemble more than 64k fragments.
1141 */
1142 m->m_pkthdr.csum_data =
1143 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1144 #ifdef MAC
1145 mac_create_datagram_from_ipq(fp, m);
1146 mac_destroy_ipq(fp);
1147 #endif
1148
1149 /*
1150 * Create header for new ip packet by modifying header of first
1151 * packet; dequeue and discard fragment reassembly header.
1152 * Make header visible.
1153 */
1154 ip->ip_len = (ip->ip_hl << 2) + next;
1155 ip->ip_src = fp->ipq_src;
1156 ip->ip_dst = fp->ipq_dst;
1157 TAILQ_REMOVE(head, fp, ipq_list);
1158 nipq--;
1159 uma_zfree(ipq_zone, fp);
1160 m->m_len += (ip->ip_hl << 2);
1161 m->m_data -= (ip->ip_hl << 2);
1162 /* some debugging cruft by sklower, below, will go away soon */
1163 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1164 m_fixhdr(m);
1165 ipstat.ips_reassembled++;
1166 IPQ_UNLOCK();
1167 return (m);
1168
1169 dropfrag:
1170 ipstat.ips_fragdropped++;
1171 if (fp != NULL)
1172 fp->ipq_nfrags--;
1173 m_freem(m);
1174 done:
1175 IPQ_UNLOCK();
1176 return (NULL);
1177
1178 #undef GETIP
1179 }
1180
1181 /*
1182 * Free a fragment reassembly header and all
1183 * associated datagrams.
1184 */
1185 static void
1186 ip_freef(fhp, fp)
1187 struct ipqhead *fhp;
1188 struct ipq *fp;
1189 {
1190 register struct mbuf *q;
1191
1192 IPQ_LOCK_ASSERT();
1193
1194 while (fp->ipq_frags) {
1195 q = fp->ipq_frags;
1196 fp->ipq_frags = q->m_nextpkt;
1197 m_freem(q);
1198 }
1199 TAILQ_REMOVE(fhp, fp, ipq_list);
1200 uma_zfree(ipq_zone, fp);
1201 nipq--;
1202 }
1203
1204 /*
1205 * IP timer processing;
1206 * if a timer expires on a reassembly
1207 * queue, discard it.
1208 */
1209 void
1210 ip_slowtimo()
1211 {
1212 register struct ipq *fp;
1213 int i;
1214
1215 IPQ_LOCK();
1216 for (i = 0; i < IPREASS_NHASH; i++) {
1217 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1218 struct ipq *fpp;
1219
1220 fpp = fp;
1221 fp = TAILQ_NEXT(fp, ipq_list);
1222 if(--fpp->ipq_ttl == 0) {
1223 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1224 ip_freef(&ipq[i], fpp);
1225 }
1226 }
1227 }
1228 /*
1229 * If we are over the maximum number of fragments
1230 * (due to the limit being lowered), drain off
1231 * enough to get down to the new limit.
1232 */
1233 if (maxnipq >= 0 && nipq > maxnipq) {
1234 for (i = 0; i < IPREASS_NHASH; i++) {
1235 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1236 ipstat.ips_fragdropped +=
1237 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1238 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1239 }
1240 }
1241 }
1242 IPQ_UNLOCK();
1243 }
1244
1245 /*
1246 * Drain off all datagram fragments.
1247 */
1248 void
1249 ip_drain()
1250 {
1251 int i;
1252
1253 IPQ_LOCK();
1254 for (i = 0; i < IPREASS_NHASH; i++) {
1255 while(!TAILQ_EMPTY(&ipq[i])) {
1256 ipstat.ips_fragdropped +=
1257 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1258 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1259 }
1260 }
1261 IPQ_UNLOCK();
1262 in_rtqdrain();
1263 }
1264
1265 /*
1266 * The protocol to be inserted into ip_protox[] must be already registered
1267 * in inetsw[], either statically or through pf_proto_register().
1268 */
1269 int
1270 ipproto_register(u_char ipproto)
1271 {
1272 struct protosw *pr;
1273
1274 /* Sanity checks. */
1275 if (ipproto == 0)
1276 return (EPROTONOSUPPORT);
1277
1278 /*
1279 * The protocol slot must not be occupied by another protocol
1280 * already. An index pointing to IPPROTO_RAW is unused.
1281 */
1282 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1283 if (pr == NULL)
1284 return (EPFNOSUPPORT);
1285 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1286 return (EEXIST);
1287
1288 /* Find the protocol position in inetsw[] and set the index. */
1289 for (pr = inetdomain.dom_protosw;
1290 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1291 if (pr->pr_domain->dom_family == PF_INET &&
1292 pr->pr_protocol && pr->pr_protocol == ipproto) {
1293 /* Be careful to only index valid IP protocols. */
1294 if (pr->pr_protocol < IPPROTO_MAX) {
1295 ip_protox[pr->pr_protocol] = pr - inetsw;
1296 return (0);
1297 } else
1298 return (EINVAL);
1299 }
1300 }
1301 return (EPROTONOSUPPORT);
1302 }
1303
1304 int
1305 ipproto_unregister(u_char ipproto)
1306 {
1307 struct protosw *pr;
1308
1309 /* Sanity checks. */
1310 if (ipproto == 0)
1311 return (EPROTONOSUPPORT);
1312
1313 /* Check if the protocol was indeed registered. */
1314 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1315 if (pr == NULL)
1316 return (EPFNOSUPPORT);
1317 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1318 return (ENOENT);
1319
1320 /* Reset the protocol slot to IPPROTO_RAW. */
1321 ip_protox[ipproto] = pr - inetsw;
1322 return (0);
1323 }
1324
1325
1326 /*
1327 * Do option processing on a datagram,
1328 * possibly discarding it if bad options are encountered,
1329 * or forwarding it if source-routed.
1330 * The pass argument is used when operating in the IPSTEALTH
1331 * mode to tell what options to process:
1332 * [LS]SRR (pass 0) or the others (pass 1).
1333 * The reason for as many as two passes is that when doing IPSTEALTH,
1334 * non-routing options should be processed only if the packet is for us.
1335 * Returns 1 if packet has been forwarded/freed,
1336 * 0 if the packet should be processed further.
1337 */
1338 static int
1339 ip_dooptions(struct mbuf *m, int pass)
1340 {
1341 struct ip *ip = mtod(m, struct ip *);
1342 u_char *cp;
1343 struct in_ifaddr *ia;
1344 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1345 struct in_addr *sin, dst;
1346 n_time ntime;
1347 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1348
1349 /* ignore or reject packets with IP options */
1350 if (ip_doopts == 0)
1351 return 0;
1352 else if (ip_doopts == 2) {
1353 type = ICMP_UNREACH;
1354 code = ICMP_UNREACH_FILTER_PROHIB;
1355 goto bad;
1356 }
1357
1358 dst = ip->ip_dst;
1359 cp = (u_char *)(ip + 1);
1360 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1361 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1362 opt = cp[IPOPT_OPTVAL];
1363 if (opt == IPOPT_EOL)
1364 break;
1365 if (opt == IPOPT_NOP)
1366 optlen = 1;
1367 else {
1368 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1369 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1370 goto bad;
1371 }
1372 optlen = cp[IPOPT_OLEN];
1373 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1374 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1375 goto bad;
1376 }
1377 }
1378 switch (opt) {
1379
1380 default:
1381 break;
1382
1383 /*
1384 * Source routing with record.
1385 * Find interface with current destination address.
1386 * If none on this machine then drop if strictly routed,
1387 * or do nothing if loosely routed.
1388 * Record interface address and bring up next address
1389 * component. If strictly routed make sure next
1390 * address is on directly accessible net.
1391 */
1392 case IPOPT_LSRR:
1393 case IPOPT_SSRR:
1394 #ifdef IPSTEALTH
1395 if (ipstealth && pass > 0)
1396 break;
1397 #endif
1398 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1399 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1400 goto bad;
1401 }
1402 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1403 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1404 goto bad;
1405 }
1406 ipaddr.sin_addr = ip->ip_dst;
1407 ia = (struct in_ifaddr *)
1408 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1409 if (ia == NULL) {
1410 if (opt == IPOPT_SSRR) {
1411 type = ICMP_UNREACH;
1412 code = ICMP_UNREACH_SRCFAIL;
1413 goto bad;
1414 }
1415 if (!ip_dosourceroute)
1416 goto nosourcerouting;
1417 /*
1418 * Loose routing, and not at next destination
1419 * yet; nothing to do except forward.
1420 */
1421 break;
1422 }
1423 off--; /* 0 origin */
1424 if (off > optlen - (int)sizeof(struct in_addr)) {
1425 /*
1426 * End of source route. Should be for us.
1427 */
1428 if (!ip_acceptsourceroute)
1429 goto nosourcerouting;
1430 save_rte(m, cp, ip->ip_src);
1431 break;
1432 }
1433 #ifdef IPSTEALTH
1434 if (ipstealth)
1435 goto dropit;
1436 #endif
1437 if (!ip_dosourceroute) {
1438 if (ipforwarding) {
1439 char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1440 /*
1441 * Acting as a router, so generate ICMP
1442 */
1443 nosourcerouting:
1444 strcpy(buf, inet_ntoa(ip->ip_dst));
1445 log(LOG_WARNING,
1446 "attempted source route from %s to %s\n",
1447 inet_ntoa(ip->ip_src), buf);
1448 type = ICMP_UNREACH;
1449 code = ICMP_UNREACH_SRCFAIL;
1450 goto bad;
1451 } else {
1452 /*
1453 * Not acting as a router, so silently drop.
1454 */
1455 #ifdef IPSTEALTH
1456 dropit:
1457 #endif
1458 ipstat.ips_cantforward++;
1459 m_freem(m);
1460 return (1);
1461 }
1462 }
1463
1464 /*
1465 * locate outgoing interface
1466 */
1467 (void)memcpy(&ipaddr.sin_addr, cp + off,
1468 sizeof(ipaddr.sin_addr));
1469
1470 if (opt == IPOPT_SSRR) {
1471 #define INA struct in_ifaddr *
1472 #define SA struct sockaddr *
1473 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == NULL)
1474 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1475 } else
1476 ia = ip_rtaddr(ipaddr.sin_addr);
1477 if (ia == NULL) {
1478 type = ICMP_UNREACH;
1479 code = ICMP_UNREACH_SRCFAIL;
1480 goto bad;
1481 }
1482 ip->ip_dst = ipaddr.sin_addr;
1483 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1484 sizeof(struct in_addr));
1485 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1486 /*
1487 * Let ip_intr's mcast routing check handle mcast pkts
1488 */
1489 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1490 break;
1491
1492 case IPOPT_RR:
1493 #ifdef IPSTEALTH
1494 if (ipstealth && pass == 0)
1495 break;
1496 #endif
1497 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1498 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1499 goto bad;
1500 }
1501 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1502 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1503 goto bad;
1504 }
1505 /*
1506 * If no space remains, ignore.
1507 */
1508 off--; /* 0 origin */
1509 if (off > optlen - (int)sizeof(struct in_addr))
1510 break;
1511 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1512 sizeof(ipaddr.sin_addr));
1513 /*
1514 * locate outgoing interface; if we're the destination,
1515 * use the incoming interface (should be same).
1516 */
1517 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1518 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
1519 type = ICMP_UNREACH;
1520 code = ICMP_UNREACH_HOST;
1521 goto bad;
1522 }
1523 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1524 sizeof(struct in_addr));
1525 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1526 break;
1527
1528 case IPOPT_TS:
1529 #ifdef IPSTEALTH
1530 if (ipstealth && pass == 0)
1531 break;
1532 #endif
1533 code = cp - (u_char *)ip;
1534 if (optlen < 4 || optlen > 40) {
1535 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1536 goto bad;
1537 }
1538 if ((off = cp[IPOPT_OFFSET]) < 5) {
1539 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1540 goto bad;
1541 }
1542 if (off > optlen - (int)sizeof(int32_t)) {
1543 cp[IPOPT_OFFSET + 1] += (1 << 4);
1544 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1545 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1546 goto bad;
1547 }
1548 break;
1549 }
1550 off--; /* 0 origin */
1551 sin = (struct in_addr *)(cp + off);
1552 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1553
1554 case IPOPT_TS_TSONLY:
1555 break;
1556
1557 case IPOPT_TS_TSANDADDR:
1558 if (off + sizeof(n_time) +
1559 sizeof(struct in_addr) > optlen) {
1560 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1561 goto bad;
1562 }
1563 ipaddr.sin_addr = dst;
1564 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1565 m->m_pkthdr.rcvif);
1566 if (ia == NULL)
1567 continue;
1568 (void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1569 sizeof(struct in_addr));
1570 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1571 off += sizeof(struct in_addr);
1572 break;
1573
1574 case IPOPT_TS_PRESPEC:
1575 if (off + sizeof(n_time) +
1576 sizeof(struct in_addr) > optlen) {
1577 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1578 goto bad;
1579 }
1580 (void)memcpy(&ipaddr.sin_addr, sin,
1581 sizeof(struct in_addr));
1582 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1583 continue;
1584 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1585 off += sizeof(struct in_addr);
1586 break;
1587
1588 default:
1589 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1590 goto bad;
1591 }
1592 ntime = iptime();
1593 (void)memcpy(cp + off, &ntime, sizeof(n_time));
1594 cp[IPOPT_OFFSET] += sizeof(n_time);
1595 }
1596 }
1597 if (forward && ipforwarding) {
1598 ip_forward(m, 1);
1599 return (1);
1600 }
1601 return (0);
1602 bad:
1603 icmp_error(m, type, code, 0, 0);
1604 ipstat.ips_badoptions++;
1605 return (1);
1606 }
1607
1608 /*
1609 * Given address of next destination (final or next hop),
1610 * return internet address info of interface to be used to get there.
1611 */
1612 struct in_ifaddr *
1613 ip_rtaddr(dst)
1614 struct in_addr dst;
1615 {
1616 struct route sro;
1617 struct sockaddr_in *sin;
1618 struct in_ifaddr *ifa;
1619
1620 bzero(&sro, sizeof(sro));
1621 sin = (struct sockaddr_in *)&sro.ro_dst;
1622 sin->sin_family = AF_INET;
1623 sin->sin_len = sizeof(*sin);
1624 sin->sin_addr = dst;
1625 rtalloc_ign(&sro, RTF_CLONING);
1626
1627 if (sro.ro_rt == NULL)
1628 return (NULL);
1629
1630 ifa = ifatoia(sro.ro_rt->rt_ifa);
1631 RTFREE(sro.ro_rt);
1632 return (ifa);
1633 }
1634
1635 /*
1636 * Save incoming source route for use in replies,
1637 * to be picked up later by ip_srcroute if the receiver is interested.
1638 */
1639 static void
1640 save_rte(m, option, dst)
1641 struct mbuf *m;
1642 u_char *option;
1643 struct in_addr dst;
1644 {
1645 unsigned olen;
1646 struct ipopt_tag *opts;
1647
1648 opts = (struct ipopt_tag *)m_tag_get(PACKET_TAG_IPOPTIONS,
1649 sizeof(struct ipopt_tag), M_NOWAIT);
1650 if (opts == NULL)
1651 return;
1652
1653 olen = option[IPOPT_OLEN];
1654 #ifdef DIAGNOSTIC
1655 if (ipprintfs)
1656 printf("save_rte: olen %d\n", olen);
1657 #endif
1658 if (olen > sizeof(opts->ip_srcrt) - (1 + sizeof(dst))) {
1659 m_tag_free((struct m_tag *)opts);
1660 return;
1661 }
1662 bcopy(option, opts->ip_srcrt.srcopt, olen);
1663 opts->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1664 opts->ip_srcrt.dst = dst;
1665 m_tag_prepend(m, (struct m_tag *)opts);
1666 }
1667
1668 /*
1669 * Retrieve incoming source route for use in replies,
1670 * in the same form used by setsockopt.
1671 * The first hop is placed before the options, will be removed later.
1672 */
1673 struct mbuf *
1674 ip_srcroute(m0)
1675 struct mbuf *m0;
1676 {
1677 register struct in_addr *p, *q;
1678 register struct mbuf *m;
1679 struct ipopt_tag *opts;
1680
1681 opts = (struct ipopt_tag *)m_tag_find(m0, PACKET_TAG_IPOPTIONS, NULL);
1682 if (opts == NULL)
1683 return (NULL);
1684
1685 if (opts->ip_nhops == 0)
1686 return (NULL);
1687 m = m_get(M_DONTWAIT, MT_DATA);
1688 if (m == NULL)
1689 return (NULL);
1690
1691 #define OPTSIZ (sizeof(opts->ip_srcrt.nop) + sizeof(opts->ip_srcrt.srcopt))
1692
1693 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1694 m->m_len = opts->ip_nhops * sizeof(struct in_addr) +
1695 sizeof(struct in_addr) + OPTSIZ;
1696 #ifdef DIAGNOSTIC
1697 if (ipprintfs)
1698 printf("ip_srcroute: nhops %d mlen %d", opts->ip_nhops, m->m_len);
1699 #endif
1700
1701 /*
1702 * First save first hop for return route
1703 */
1704 p = &(opts->ip_srcrt.route[opts->ip_nhops - 1]);
1705 *(mtod(m, struct in_addr *)) = *p--;
1706 #ifdef DIAGNOSTIC
1707 if (ipprintfs)
1708 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1709 #endif
1710
1711 /*
1712 * Copy option fields and padding (nop) to mbuf.
1713 */
1714 opts->ip_srcrt.nop = IPOPT_NOP;
1715 opts->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1716 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1717 &(opts->ip_srcrt.nop), OPTSIZ);
1718 q = (struct in_addr *)(mtod(m, caddr_t) +
1719 sizeof(struct in_addr) + OPTSIZ);
1720 #undef OPTSIZ
1721 /*
1722 * Record return path as an IP source route,
1723 * reversing the path (pointers are now aligned).
1724 */
1725 while (p >= opts->ip_srcrt.route) {
1726 #ifdef DIAGNOSTIC
1727 if (ipprintfs)
1728 printf(" %lx", (u_long)ntohl(q->s_addr));
1729 #endif
1730 *q++ = *p--;
1731 }
1732 /*
1733 * Last hop goes to final destination.
1734 */
1735 *q = opts->ip_srcrt.dst;
1736 #ifdef DIAGNOSTIC
1737 if (ipprintfs)
1738 printf(" %lx\n", (u_long)ntohl(q->s_addr));
1739 #endif
1740 m_tag_delete(m0, (struct m_tag *)opts);
1741 return (m);
1742 }
1743
1744 /*
1745 * Strip out IP options, at higher
1746 * level protocol in the kernel.
1747 * Second argument is buffer to which options
1748 * will be moved, and return value is their length.
1749 * XXX should be deleted; last arg currently ignored.
1750 */
1751 void
1752 ip_stripoptions(m, mopt)
1753 register struct mbuf *m;
1754 struct mbuf *mopt;
1755 {
1756 register int i;
1757 struct ip *ip = mtod(m, struct ip *);
1758 register caddr_t opts;
1759 int olen;
1760
1761 olen = (ip->ip_hl << 2) - sizeof (struct ip);
1762 opts = (caddr_t)(ip + 1);
1763 i = m->m_len - (sizeof (struct ip) + olen);
1764 bcopy(opts + olen, opts, (unsigned)i);
1765 m->m_len -= olen;
1766 if (m->m_flags & M_PKTHDR)
1767 m->m_pkthdr.len -= olen;
1768 ip->ip_v = IPVERSION;
1769 ip->ip_hl = sizeof(struct ip) >> 2;
1770 }
1771
1772 u_char inetctlerrmap[PRC_NCMDS] = {
1773 0, 0, 0, 0,
1774 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1775 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1776 EMSGSIZE, EHOSTUNREACH, 0, 0,
1777 0, 0, EHOSTUNREACH, 0,
1778 ENOPROTOOPT, ECONNREFUSED
1779 };
1780
1781 /*
1782 * Forward a packet. If some error occurs return the sender
1783 * an icmp packet. Note we can't always generate a meaningful
1784 * icmp message because icmp doesn't have a large enough repertoire
1785 * of codes and types.
1786 *
1787 * If not forwarding, just drop the packet. This could be confusing
1788 * if ipforwarding was zero but some routing protocol was advancing
1789 * us as a gateway to somewhere. However, we must let the routing
1790 * protocol deal with that.
1791 *
1792 * The srcrt parameter indicates whether the packet is being forwarded
1793 * via a source route.
1794 */
1795 void
1796 ip_forward(struct mbuf *m, int srcrt)
1797 {
1798 struct ip *ip = mtod(m, struct ip *);
1799 struct in_ifaddr *ia = NULL;
1800 struct mbuf *mcopy;
1801 struct in_addr dest;
1802 int error, type = 0, code = 0, mtu = 0;
1803
1804 #ifdef DIAGNOSTIC
1805 if (ipprintfs)
1806 printf("forward: src %lx dst %lx ttl %x\n",
1807 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1808 ip->ip_ttl);
1809 #endif
1810
1811
1812 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1813 ipstat.ips_cantforward++;
1814 m_freem(m);
1815 return;
1816 }
1817 #ifdef IPSTEALTH
1818 if (!ipstealth) {
1819 #endif
1820 if (ip->ip_ttl <= IPTTLDEC) {
1821 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1822 0, 0);
1823 return;
1824 }
1825 #ifdef IPSTEALTH
1826 }
1827 #endif
1828
1829 if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) {
1830 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1831 return;
1832 }
1833
1834 /*
1835 * Save the IP header and at most 8 bytes of the payload,
1836 * in case we need to generate an ICMP message to the src.
1837 *
1838 * XXX this can be optimized a lot by saving the data in a local
1839 * buffer on the stack (72 bytes at most), and only allocating the
1840 * mbuf if really necessary. The vast majority of the packets
1841 * are forwarded without having to send an ICMP back (either
1842 * because unnecessary, or because rate limited), so we are
1843 * really we are wasting a lot of work here.
1844 *
1845 * We don't use m_copy() because it might return a reference
1846 * to a shared cluster. Both this function and ip_output()
1847 * assume exclusive access to the IP header in `m', so any
1848 * data in a cluster may change before we reach icmp_error().
1849 */
1850 MGET(mcopy, M_DONTWAIT, m->m_type);
1851 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1852 /*
1853 * It's probably ok if the pkthdr dup fails (because
1854 * the deep copy of the tag chain failed), but for now
1855 * be conservative and just discard the copy since
1856 * code below may some day want the tags.
1857 */
1858 m_free(mcopy);
1859 mcopy = NULL;
1860 }
1861 if (mcopy != NULL) {
1862 mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1863 (int)ip->ip_len);
1864 mcopy->m_pkthdr.len = mcopy->m_len;
1865 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1866 }
1867
1868 #ifdef IPSTEALTH
1869 if (!ipstealth) {
1870 #endif
1871 ip->ip_ttl -= IPTTLDEC;
1872 #ifdef IPSTEALTH
1873 }
1874 #endif
1875
1876 /*
1877 * If forwarding packet using same interface that it came in on,
1878 * perhaps should send a redirect to sender to shortcut a hop.
1879 * Only send redirect if source is sending directly to us,
1880 * and if packet was not source routed (or has any options).
1881 * Also, don't send redirect if forwarding using a default route
1882 * or a route modified by a redirect.
1883 */
1884 dest.s_addr = 0;
1885 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1886 struct sockaddr_in *sin;
1887 struct route ro;
1888 struct rtentry *rt;
1889
1890 bzero(&ro, sizeof(ro));
1891 sin = (struct sockaddr_in *)&ro.ro_dst;
1892 sin->sin_family = AF_INET;
1893 sin->sin_len = sizeof(*sin);
1894 sin->sin_addr = ip->ip_dst;
1895 rtalloc_ign(&ro, RTF_CLONING);
1896
1897 rt = ro.ro_rt;
1898
1899 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1900 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1901 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1902 u_long src = ntohl(ip->ip_src.s_addr);
1903
1904 if (RTA(rt) &&
1905 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1906 if (rt->rt_flags & RTF_GATEWAY)
1907 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1908 else
1909 dest.s_addr = ip->ip_dst.s_addr;
1910 /* Router requirements says to only send host redirects */
1911 type = ICMP_REDIRECT;
1912 code = ICMP_REDIRECT_HOST;
1913 #ifdef DIAGNOSTIC
1914 if (ipprintfs)
1915 printf("redirect (%d) to %lx\n", code, (u_long)dest.s_addr);
1916 #endif
1917 }
1918 }
1919 if (rt)
1920 RTFREE(rt);
1921 }
1922
1923 error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
1924 if (error)
1925 ipstat.ips_cantforward++;
1926 else {
1927 ipstat.ips_forward++;
1928 if (type)
1929 ipstat.ips_redirectsent++;
1930 else {
1931 if (mcopy)
1932 m_freem(mcopy);
1933 return;
1934 }
1935 }
1936 if (mcopy == NULL)
1937 return;
1938
1939 switch (error) {
1940
1941 case 0: /* forwarded, but need redirect */
1942 /* type, code set above */
1943 break;
1944
1945 case ENETUNREACH: /* shouldn't happen, checked above */
1946 case EHOSTUNREACH:
1947 case ENETDOWN:
1948 case EHOSTDOWN:
1949 default:
1950 type = ICMP_UNREACH;
1951 code = ICMP_UNREACH_HOST;
1952 break;
1953
1954 case EMSGSIZE:
1955 type = ICMP_UNREACH;
1956 code = ICMP_UNREACH_NEEDFRAG;
1957 #if defined(IPSEC) || defined(FAST_IPSEC)
1958 /*
1959 * If the packet is routed over IPsec tunnel, tell the
1960 * originator the tunnel MTU.
1961 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1962 * XXX quickhack!!!
1963 */
1964 {
1965 struct secpolicy *sp = NULL;
1966 int ipsecerror;
1967 int ipsechdr;
1968 struct route *ro;
1969
1970 #ifdef IPSEC
1971 sp = ipsec4_getpolicybyaddr(mcopy,
1972 IPSEC_DIR_OUTBOUND,
1973 IP_FORWARDING,
1974 &ipsecerror);
1975 #else /* FAST_IPSEC */
1976 sp = ipsec_getpolicybyaddr(mcopy,
1977 IPSEC_DIR_OUTBOUND,
1978 IP_FORWARDING,
1979 &ipsecerror);
1980 #endif
1981 if (sp != NULL) {
1982 /* count IPsec header size */
1983 ipsechdr = ipsec4_hdrsiz(mcopy,
1984 IPSEC_DIR_OUTBOUND,
1985 NULL);
1986
1987 /*
1988 * find the correct route for outer IPv4
1989 * header, compute tunnel MTU.
1990 */
1991 if (sp->req != NULL
1992 && sp->req->sav != NULL
1993 && sp->req->sav->sah != NULL) {
1994 ro = &sp->req->sav->sah->sa_route;
1995 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1996 mtu =
1997 ro->ro_rt->rt_rmx.rmx_mtu ?
1998 ro->ro_rt->rt_rmx.rmx_mtu :
1999 ro->ro_rt->rt_ifp->if_mtu;
2000 mtu -= ipsechdr;
2001 }
2002 }
2003
2004 #ifdef IPSEC
2005 key_freesp(sp);
2006 #else /* FAST_IPSEC */
2007 KEY_FREESP(&sp);
2008 #endif
2009 }
2010 }
2011 #endif /*IPSEC || FAST_IPSEC*/
2012 /*
2013 * If the MTU wasn't set before use the interface mtu or
2014 * fall back to the next smaller mtu step compared to the
2015 * current packet size.
2016 */
2017 if (mtu == 0) {
2018 if (ia != NULL)
2019 mtu = ia->ia_ifp->if_mtu;
2020 else
2021 mtu = ip_next_mtu(ip->ip_len, 0);
2022 }
2023 ipstat.ips_cantfrag++;
2024 break;
2025
2026 case ENOBUFS:
2027 /*
2028 * A router should not generate ICMP_SOURCEQUENCH as
2029 * required in RFC1812 Requirements for IP Version 4 Routers.
2030 * Source quench could be a big problem under DoS attacks,
2031 * or if the underlying interface is rate-limited.
2032 * Those who need source quench packets may re-enable them
2033 * via the net.inet.ip.sendsourcequench sysctl.
2034 */
2035 if (ip_sendsourcequench == 0) {
2036 m_freem(mcopy);
2037 return;
2038 } else {
2039 type = ICMP_SOURCEQUENCH;
2040 code = 0;
2041 }
2042 break;
2043
2044 case EACCES: /* ipfw denied packet */
2045 m_freem(mcopy);
2046 return;
2047 }
2048 icmp_error(mcopy, type, code, dest.s_addr, mtu);
2049 }
2050
2051 void
2052 ip_savecontrol(inp, mp, ip, m)
2053 register struct inpcb *inp;
2054 register struct mbuf **mp;
2055 register struct ip *ip;
2056 register struct mbuf *m;
2057 {
2058 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
2059 struct bintime bt;
2060
2061 bintime(&bt);
2062 if (inp->inp_socket->so_options & SO_BINTIME) {
2063 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
2064 SCM_BINTIME, SOL_SOCKET);
2065 if (*mp)
2066 mp = &(*mp)->m_next;
2067 }
2068 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2069 struct timeval tv;
2070
2071 bintime2timeval(&bt, &tv);
2072 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2073 SCM_TIMESTAMP, SOL_SOCKET);
2074 if (*mp)
2075 mp = &(*mp)->m_next;
2076 }
2077 }
2078 if (inp->inp_flags & INP_RECVDSTADDR) {
2079 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2080 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2081 if (*mp)
2082 mp = &(*mp)->m_next;
2083 }
2084 if (inp->inp_flags & INP_RECVTTL) {
2085 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2086 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2087 if (*mp)
2088 mp = &(*mp)->m_next;
2089 }
2090 #ifdef notyet
2091 /* XXX
2092 * Moving these out of udp_input() made them even more broken
2093 * than they already were.
2094 */
2095 /* options were tossed already */
2096 if (inp->inp_flags & INP_RECVOPTS) {
2097 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2098 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2099 if (*mp)
2100 mp = &(*mp)->m_next;
2101 }
2102 /* ip_srcroute doesn't do what we want here, need to fix */
2103 if (inp->inp_flags & INP_RECVRETOPTS) {
2104 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2105 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2106 if (*mp)
2107 mp = &(*mp)->m_next;
2108 }
2109 #endif
2110 if (inp->inp_flags & INP_RECVIF) {
2111 struct ifnet *ifp;
2112 struct sdlbuf {
2113 struct sockaddr_dl sdl;
2114 u_char pad[32];
2115 } sdlbuf;
2116 struct sockaddr_dl *sdp;
2117 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2118
2119 if (((ifp = m->m_pkthdr.rcvif))
2120 && ( ifp->if_index && (ifp->if_index <= if_index))) {
2121 sdp = (struct sockaddr_dl *)
2122 (ifaddr_byindex(ifp->if_index)->ifa_addr);
2123 /*
2124 * Change our mind and don't try copy.
2125 */
2126 if ((sdp->sdl_family != AF_LINK)
2127 || (sdp->sdl_len > sizeof(sdlbuf))) {
2128 goto makedummy;
2129 }
2130 bcopy(sdp, sdl2, sdp->sdl_len);
2131 } else {
2132 makedummy:
2133 sdl2->sdl_len
2134 = offsetof(struct sockaddr_dl, sdl_data[0]);
2135 sdl2->sdl_family = AF_LINK;
2136 sdl2->sdl_index = 0;
2137 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2138 }
2139 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2140 IP_RECVIF, IPPROTO_IP);
2141 if (*mp)
2142 mp = &(*mp)->m_next;
2143 }
2144 }
2145
2146 /*
2147 * XXX these routines are called from the upper part of the kernel.
2148 * They need to be locked when we remove Giant.
2149 *
2150 * They could also be moved to ip_mroute.c, since all the RSVP
2151 * handling is done there already.
2152 */
2153 static int ip_rsvp_on;
2154 struct socket *ip_rsvpd;
2155 int
2156 ip_rsvp_init(struct socket *so)
2157 {
2158 if (so->so_type != SOCK_RAW ||
2159 so->so_proto->pr_protocol != IPPROTO_RSVP)
2160 return EOPNOTSUPP;
2161
2162 if (ip_rsvpd != NULL)
2163 return EADDRINUSE;
2164
2165 ip_rsvpd = so;
2166 /*
2167 * This may seem silly, but we need to be sure we don't over-increment
2168 * the RSVP counter, in case something slips up.
2169 */
2170 if (!ip_rsvp_on) {
2171 ip_rsvp_on = 1;
2172 rsvp_on++;
2173 }
2174
2175 return 0;
2176 }
2177
2178 int
2179 ip_rsvp_done(void)
2180 {
2181 ip_rsvpd = NULL;
2182 /*
2183 * This may seem silly, but we need to be sure we don't over-decrement
2184 * the RSVP counter, in case something slips up.
2185 */
2186 if (ip_rsvp_on) {
2187 ip_rsvp_on = 0;
2188 rsvp_on--;
2189 }
2190 return 0;
2191 }
2192
2193 void
2194 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
2195 {
2196 if (rsvp_input_p) { /* call the real one if loaded */
2197 rsvp_input_p(m, off);
2198 return;
2199 }
2200
2201 /* Can still get packets with rsvp_on = 0 if there is a local member
2202 * of the group to which the RSVP packet is addressed. But in this
2203 * case we want to throw the packet away.
2204 */
2205
2206 if (!rsvp_on) {
2207 m_freem(m);
2208 return;
2209 }
2210
2211 if (ip_rsvpd != NULL) {
2212 rip_input(m, off);
2213 return;
2214 }
2215 /* Drop the packet */
2216 m_freem(m);
2217 }
Cache object: 87e6af6746942e715bc5ab56017b1a39
|