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 * $FreeBSD$
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 nipq = 0; /* total # of reass queues */
123 static int maxnipq;
124 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
125 &maxnipq, 0,
126 "Maximum number of IPv4 fragment reassembly queue entries");
127
128 static int maxfragsperpacket;
129 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
130 &maxfragsperpacket, 0,
131 "Maximum number of IPv4 fragments allowed per packet");
132
133 static int ip_sendsourcequench = 0;
134 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
135 &ip_sendsourcequench, 0,
136 "Enable the transmission of source quench packets");
137
138 int ip_do_randomid = 0;
139 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
140 &ip_do_randomid, 0,
141 "Assign random ip_id values");
142
143 /*
144 * XXX - Setting ip_checkinterface mostly implements the receive side of
145 * the Strong ES model described in RFC 1122, but since the routing table
146 * and transmit implementation do not implement the Strong ES model,
147 * setting this to 1 results in an odd hybrid.
148 *
149 * XXX - ip_checkinterface currently must be disabled if you use ipnat
150 * to translate the destination address to another local interface.
151 *
152 * XXX - ip_checkinterface must be disabled if you add IP aliases
153 * to the loopback interface instead of the interface where the
154 * packets for those addresses are received.
155 */
156 static int ip_checkinterface = 0;
157 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
158 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
159
160 #ifdef DIAGNOSTIC
161 static int ipprintfs = 0;
162 #endif
163
164 struct pfil_head inet_pfil_hook;
165
166 static struct ifqueue ipintrq;
167 static int ipqmaxlen = IFQ_MAXLEN;
168
169 extern struct domain inetdomain;
170 extern struct protosw inetsw[];
171 u_char ip_protox[IPPROTO_MAX];
172 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
173 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
174 u_long in_ifaddrhmask; /* mask for hash table */
175
176 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
177 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
178 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
179 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
180
181 struct ipstat ipstat;
182 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
183 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
184
185 /* Packet reassembly stuff */
186 #define IPREASS_NHASH_LOG2 6
187 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
188 #define IPREASS_HMASK (IPREASS_NHASH - 1)
189 #define IPREASS_HASH(x,y) \
190 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
191
192 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
193 struct mtx ipqlock;
194 struct callout ipport_tick_callout;
195
196 #define IPQ_LOCK() mtx_lock(&ipqlock)
197 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
198 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
199 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
200
201 #ifdef IPCTL_DEFMTU
202 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
203 &ip_mtu, 0, "Default MTU");
204 #endif
205
206 #ifdef IPSTEALTH
207 int ipstealth = 0;
208 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
209 &ipstealth, 0, "");
210 #endif
211
212 /*
213 * ipfw_ether and ipfw_bridge hooks.
214 * XXX: Temporary until those are converted to pfil_hooks as well.
215 */
216 ip_fw_chk_t *ip_fw_chk_ptr = NULL;
217 ip_dn_io_t *ip_dn_io_ptr = NULL;
218 int fw_enable = 1;
219 int fw_one_pass = 1;
220
221 /*
222 * XXX this is ugly. IP options source routing magic.
223 */
224 struct ipoptrt {
225 struct in_addr dst; /* final destination */
226 char nop; /* one NOP to align */
227 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
228 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
229 };
230
231 struct ipopt_tag {
232 struct m_tag tag;
233 int ip_nhops;
234 struct ipoptrt ip_srcrt;
235 };
236
237 static void save_rte(struct mbuf *, u_char *, struct in_addr);
238 static int ip_dooptions(struct mbuf *m, int);
239 static void ip_forward(struct mbuf *m, int srcrt);
240 static void ip_freef(struct ipqhead *, struct ipq *);
241
242 /*
243 * IP initialization: fill in IP protocol switch table.
244 * All protocols not implemented in kernel go to raw IP protocol handler.
245 */
246 void
247 ip_init()
248 {
249 register struct protosw *pr;
250 register int i;
251
252 TAILQ_INIT(&in_ifaddrhead);
253 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
254 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
255 if (pr == NULL)
256 panic("ip_init: PF_INET not found");
257
258 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
259 for (i = 0; i < IPPROTO_MAX; i++)
260 ip_protox[i] = pr - inetsw;
261 /*
262 * Cycle through IP protocols and put them into the appropriate place
263 * in ip_protox[].
264 */
265 for (pr = inetdomain.dom_protosw;
266 pr < inetdomain.dom_protoswNPROTOSW; pr++)
267 if (pr->pr_domain->dom_family == PF_INET &&
268 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
269 /* Be careful to only index valid IP protocols. */
270 if (pr->pr_protocol && pr->pr_protocol < IPPROTO_MAX)
271 ip_protox[pr->pr_protocol] = pr - inetsw;
272 }
273
274 /* Initialize packet filter hooks. */
275 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
276 inet_pfil_hook.ph_af = AF_INET;
277 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
278 printf("%s: WARNING: unable to register pfil hook, "
279 "error %d\n", __func__, i);
280
281 /* Initialize IP reassembly queue. */
282 IPQ_LOCK_INIT();
283 for (i = 0; i < IPREASS_NHASH; i++)
284 TAILQ_INIT(&ipq[i]);
285 maxnipq = nmbclusters / 32;
286 maxfragsperpacket = 16;
287
288 /* Start ipport_tick. */
289 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
290 ipport_tick(NULL);
291 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
292 SHUTDOWN_PRI_DEFAULT);
293
294 /* Initialize various other remaining things. */
295 ip_id = time_second & 0xffff;
296 ipintrq.ifq_maxlen = ipqmaxlen;
297 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
298 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
299 }
300
301 void ip_fini(xtp)
302 void *xtp;
303 {
304 callout_stop(&ipport_tick_callout);
305 }
306
307 /*
308 * Ip input routine. Checksum and byte swap header. If fragmented
309 * try to reassemble. Process options. Pass to next level.
310 */
311 void
312 ip_input(struct mbuf *m)
313 {
314 struct ip *ip = NULL;
315 struct in_ifaddr *ia = NULL;
316 struct ifaddr *ifa;
317 int checkif, hlen = 0;
318 u_short sum;
319 int dchg = 0; /* dest changed after fw */
320 struct in_addr odst; /* original dst address */
321 #ifdef FAST_IPSEC
322 struct m_tag *mtag;
323 struct tdb_ident *tdbi;
324 struct secpolicy *sp;
325 int s, error;
326 #endif /* FAST_IPSEC */
327
328 M_ASSERTPKTHDR(m);
329
330 if (m->m_flags & M_FASTFWD_OURS) {
331 /*
332 * Firewall or NAT changed destination to local.
333 * We expect ip_len and ip_off to be in host byte order.
334 */
335 m->m_flags &= ~M_FASTFWD_OURS;
336 /* Set up some basics that will be used later. */
337 ip = mtod(m, struct ip *);
338 hlen = ip->ip_hl << 2;
339 goto ours;
340 }
341
342 ipstat.ips_total++;
343
344 if (m->m_pkthdr.len < sizeof(struct ip))
345 goto tooshort;
346
347 if (m->m_len < sizeof (struct ip) &&
348 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
349 ipstat.ips_toosmall++;
350 return;
351 }
352 ip = mtod(m, struct ip *);
353
354 if (ip->ip_v != IPVERSION) {
355 ipstat.ips_badvers++;
356 goto bad;
357 }
358
359 hlen = ip->ip_hl << 2;
360 if (hlen < sizeof(struct ip)) { /* minimum header length */
361 ipstat.ips_badhlen++;
362 goto bad;
363 }
364 if (hlen > m->m_len) {
365 if ((m = m_pullup(m, hlen)) == NULL) {
366 ipstat.ips_badhlen++;
367 return;
368 }
369 ip = mtod(m, struct ip *);
370 }
371
372 /* 127/8 must not appear on wire - RFC1122 */
373 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
374 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
375 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
376 ipstat.ips_badaddr++;
377 goto bad;
378 }
379 }
380
381 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
382 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
383 } else {
384 if (hlen == sizeof(struct ip)) {
385 sum = in_cksum_hdr(ip);
386 } else {
387 sum = in_cksum(m, hlen);
388 }
389 }
390 if (sum) {
391 ipstat.ips_badsum++;
392 goto bad;
393 }
394
395 #ifdef ALTQ
396 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
397 /* packet is dropped by traffic conditioner */
398 return;
399 #endif
400
401 /*
402 * Convert fields to host representation.
403 */
404 ip->ip_len = ntohs(ip->ip_len);
405 if (ip->ip_len < hlen) {
406 ipstat.ips_badlen++;
407 goto bad;
408 }
409 ip->ip_off = ntohs(ip->ip_off);
410
411 /*
412 * Check that the amount of data in the buffers
413 * is as at least much as the IP header would have us expect.
414 * Trim mbufs if longer than we expect.
415 * Drop packet if shorter than we expect.
416 */
417 if (m->m_pkthdr.len < ip->ip_len) {
418 tooshort:
419 ipstat.ips_tooshort++;
420 goto bad;
421 }
422 if (m->m_pkthdr.len > ip->ip_len) {
423 if (m->m_len == m->m_pkthdr.len) {
424 m->m_len = ip->ip_len;
425 m->m_pkthdr.len = ip->ip_len;
426 } else
427 m_adj(m, ip->ip_len - m->m_pkthdr.len);
428 }
429 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
430 /*
431 * Bypass packet filtering for packets from a tunnel (gif).
432 */
433 if (ipsec_getnhist(m))
434 goto passin;
435 #endif
436 #if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF)
437 /*
438 * Bypass packet filtering for packets from a tunnel (gif).
439 */
440 if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL)
441 goto passin;
442 #endif
443
444 /*
445 * Run through list of hooks for input packets.
446 *
447 * NB: Beware of the destination address changing (e.g.
448 * by NAT rewriting). When this happens, tell
449 * ip_forward to do the right thing.
450 */
451
452 /* Jump over all PFIL processing if hooks are not active. */
453 if (inet_pfil_hook.ph_busy_count == -1)
454 goto passin;
455
456 odst = ip->ip_dst;
457 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
458 PFIL_IN, NULL) != 0)
459 return;
460 if (m == NULL) /* consumed by filter */
461 return;
462
463 ip = mtod(m, struct ip *);
464 dchg = (odst.s_addr != ip->ip_dst.s_addr);
465
466 #ifdef IPFIREWALL_FORWARD
467 if (m->m_flags & M_FASTFWD_OURS) {
468 m->m_flags &= ~M_FASTFWD_OURS;
469 goto ours;
470 }
471 #ifndef IPFIREWALL_FORWARD_EXTENDED
472 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
473 #else
474 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
475 /*
476 * Directly ship on the packet. This allows to forward packets
477 * that were destined for us to some other directly connected
478 * host.
479 */
480 ip_forward(m, dchg);
481 return;
482 }
483 #endif /* IPFIREWALL_FORWARD_EXTENDED */
484 #endif /* IPFIREWALL_FORWARD */
485
486 passin:
487 /*
488 * Process options and, if not destined for us,
489 * ship it on. ip_dooptions returns 1 when an
490 * error was detected (causing an icmp message
491 * to be sent and the original packet to be freed).
492 */
493 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
494 return;
495
496 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
497 * matter if it is destined to another node, or whether it is
498 * a multicast one, RSVP wants it! and prevents it from being forwarded
499 * anywhere else. Also checks if the rsvp daemon is running before
500 * grabbing the packet.
501 */
502 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
503 goto ours;
504
505 /*
506 * Check our list of addresses, to see if the packet is for us.
507 * If we don't have any addresses, assume any unicast packet
508 * we receive might be for us (and let the upper layers deal
509 * with it).
510 */
511 if (TAILQ_EMPTY(&in_ifaddrhead) &&
512 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
513 goto ours;
514
515 /*
516 * Enable a consistency check between the destination address
517 * and the arrival interface for a unicast packet (the RFC 1122
518 * strong ES model) if IP forwarding is disabled and the packet
519 * is not locally generated and the packet is not subject to
520 * 'ipfw fwd'.
521 *
522 * XXX - Checking also should be disabled if the destination
523 * address is ipnat'ed to a different interface.
524 *
525 * XXX - Checking is incompatible with IP aliases added
526 * to the loopback interface instead of the interface where
527 * the packets are received.
528 *
529 * XXX - This is the case for carp vhost IPs as well so we
530 * insert a workaround. If the packet got here, we already
531 * checked with carp_iamatch() and carp_forus().
532 */
533 checkif = ip_checkinterface && (ipforwarding == 0) &&
534 m->m_pkthdr.rcvif != NULL &&
535 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
536 #ifdef DEV_CARP
537 !m->m_pkthdr.rcvif->if_carp &&
538 #endif
539 (dchg == 0);
540
541 /*
542 * Check for exact addresses in the hash bucket.
543 */
544 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
545 /*
546 * If the address matches, verify that the packet
547 * arrived via the correct interface if checking is
548 * enabled.
549 */
550 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
551 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
552 goto ours;
553 }
554 /*
555 * Check for broadcast addresses.
556 *
557 * Only accept broadcast packets that arrive via the matching
558 * interface. Reception of forwarded directed broadcasts would
559 * be handled via ip_forward() and ether_output() with the loopback
560 * into the stack for SIMPLEX interfaces handled by ether_output().
561 */
562 if (m->m_pkthdr.rcvif != NULL &&
563 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
564 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
565 if (ifa->ifa_addr->sa_family != AF_INET)
566 continue;
567 ia = ifatoia(ifa);
568 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
569 ip->ip_dst.s_addr)
570 goto ours;
571 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
572 goto ours;
573 #ifdef BOOTP_COMPAT
574 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
575 goto ours;
576 #endif
577 }
578 }
579 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
580 struct in_multi *inm;
581 if (ip_mrouter) {
582 /*
583 * If we are acting as a multicast router, all
584 * incoming multicast packets are passed to the
585 * kernel-level multicast forwarding function.
586 * The packet is returned (relatively) intact; if
587 * ip_mforward() returns a non-zero value, the packet
588 * must be discarded, else it may be accepted below.
589 */
590 if (ip_mforward &&
591 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
592 ipstat.ips_cantforward++;
593 m_freem(m);
594 return;
595 }
596
597 /*
598 * The process-level routing daemon needs to receive
599 * all multicast IGMP packets, whether or not this
600 * host belongs to their destination groups.
601 */
602 if (ip->ip_p == IPPROTO_IGMP)
603 goto ours;
604 ipstat.ips_forward++;
605 }
606 /*
607 * See if we belong to the destination multicast group on the
608 * arrival interface.
609 */
610 IN_MULTI_LOCK();
611 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
612 IN_MULTI_UNLOCK();
613 if (inm == NULL) {
614 ipstat.ips_notmember++;
615 m_freem(m);
616 return;
617 }
618 goto ours;
619 }
620 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
621 goto ours;
622 if (ip->ip_dst.s_addr == INADDR_ANY)
623 goto ours;
624
625 /*
626 * FAITH(Firewall Aided Internet Translator)
627 */
628 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
629 if (ip_keepfaith) {
630 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
631 goto ours;
632 }
633 m_freem(m);
634 return;
635 }
636
637 /*
638 * Not for us; forward if possible and desirable.
639 */
640 if (ipforwarding == 0) {
641 ipstat.ips_cantforward++;
642 m_freem(m);
643 } else {
644 #ifdef IPSEC
645 /*
646 * Enforce inbound IPsec SPD.
647 */
648 if (ipsec4_in_reject(m, NULL)) {
649 ipsecstat.in_polvio++;
650 goto bad;
651 }
652 #endif /* IPSEC */
653 #ifdef FAST_IPSEC
654 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
655 s = splnet();
656 if (mtag != NULL) {
657 tdbi = (struct tdb_ident *)(mtag + 1);
658 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
659 } else {
660 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
661 IP_FORWARDING, &error);
662 }
663 if (sp == NULL) { /* NB: can happen if error */
664 splx(s);
665 /*XXX error stat???*/
666 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
667 goto bad;
668 }
669
670 /*
671 * Check security policy against packet attributes.
672 */
673 error = ipsec_in_reject(sp, m);
674 KEY_FREESP(&sp);
675 splx(s);
676 if (error) {
677 ipstat.ips_cantforward++;
678 goto bad;
679 }
680 #endif /* FAST_IPSEC */
681 ip_forward(m, dchg);
682 }
683 return;
684
685 ours:
686 #ifdef IPSTEALTH
687 /*
688 * IPSTEALTH: Process non-routing options only
689 * if the packet is destined for us.
690 */
691 if (ipstealth && hlen > sizeof (struct ip) &&
692 ip_dooptions(m, 1))
693 return;
694 #endif /* IPSTEALTH */
695
696 /* Count the packet in the ip address stats */
697 if (ia != NULL) {
698 ia->ia_ifa.if_ipackets++;
699 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
700 }
701
702 /*
703 * Attempt reassembly; if it succeeds, proceed.
704 * ip_reass() will return a different mbuf.
705 */
706 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
707 m = ip_reass(m);
708 if (m == NULL)
709 return;
710 ip = mtod(m, struct ip *);
711 /* Get the header length of the reassembled packet */
712 hlen = ip->ip_hl << 2;
713 }
714
715 /*
716 * Further protocols expect the packet length to be w/o the
717 * IP header.
718 */
719 ip->ip_len -= hlen;
720
721 #ifdef IPSEC
722 /*
723 * enforce IPsec policy checking if we are seeing last header.
724 * note that we do not visit this with protocols with pcb layer
725 * code - like udp/tcp/raw ip.
726 */
727 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
728 ipsec4_in_reject(m, NULL)) {
729 ipsecstat.in_polvio++;
730 goto bad;
731 }
732 #endif
733 #if FAST_IPSEC
734 /*
735 * enforce IPsec policy checking if we are seeing last header.
736 * note that we do not visit this with protocols with pcb layer
737 * code - like udp/tcp/raw ip.
738 */
739 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
740 /*
741 * Check if the packet has already had IPsec processing
742 * done. If so, then just pass it along. This tag gets
743 * set during AH, ESP, etc. input handling, before the
744 * packet is returned to the ip input queue for delivery.
745 */
746 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
747 s = splnet();
748 if (mtag != NULL) {
749 tdbi = (struct tdb_ident *)(mtag + 1);
750 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
751 } else {
752 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
753 IP_FORWARDING, &error);
754 }
755 if (sp != NULL) {
756 /*
757 * Check security policy against packet attributes.
758 */
759 error = ipsec_in_reject(sp, m);
760 KEY_FREESP(&sp);
761 } else {
762 /* XXX error stat??? */
763 error = EINVAL;
764 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
765 goto bad;
766 }
767 splx(s);
768 if (error)
769 goto bad;
770 }
771 #endif /* FAST_IPSEC */
772
773 /*
774 * Switch out to protocol's input routine.
775 */
776 ipstat.ips_delivered++;
777
778 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
779 return;
780 bad:
781 m_freem(m);
782 }
783
784 /*
785 * Take incoming datagram fragment and try to reassemble it into
786 * whole datagram. If the argument is the first fragment or one
787 * in between the function will return NULL and store the mbuf
788 * in the fragment chain. If the argument is the last fragment
789 * the packet will be reassembled and the pointer to the new
790 * mbuf returned for further processing. Only m_tags attached
791 * to the first packet/fragment are preserved.
792 * The IP header is *NOT* adjusted out of iplen.
793 */
794
795 struct mbuf *
796 ip_reass(struct mbuf *m)
797 {
798 struct ip *ip;
799 struct mbuf *p, *q, *nq, *t;
800 struct ipq *fp = NULL;
801 struct ipqhead *head;
802 int i, hlen, next;
803 u_int8_t ecn, ecn0;
804 u_short hash;
805
806 /* If maxnipq is 0, never accept fragments. */
807 if (maxnipq == 0) {
808 ipstat.ips_fragments++;
809 ipstat.ips_fragdropped++;
810 m_freem(m);
811 return (NULL);
812 }
813
814 ip = mtod(m, struct ip *);
815 hlen = ip->ip_hl << 2;
816
817 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
818 head = &ipq[hash];
819 IPQ_LOCK();
820
821 /*
822 * Look for queue of fragments
823 * of this datagram.
824 */
825 TAILQ_FOREACH(fp, head, ipq_list)
826 if (ip->ip_id == fp->ipq_id &&
827 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
828 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
829 #ifdef MAC
830 mac_fragment_match(m, fp) &&
831 #endif
832 ip->ip_p == fp->ipq_p)
833 goto found;
834
835 fp = NULL;
836
837 /*
838 * Enforce upper bound on number of fragmented packets
839 * for which we attempt reassembly;
840 * If maxnipq is -1, accept all fragments without limitation.
841 */
842 if ((nipq > maxnipq) && (maxnipq > 0)) {
843 /*
844 * drop something from the tail of the current queue
845 * before proceeding further
846 */
847 struct ipq *q = TAILQ_LAST(head, ipqhead);
848 if (q == NULL) { /* gak */
849 for (i = 0; i < IPREASS_NHASH; i++) {
850 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
851 if (r) {
852 ipstat.ips_fragtimeout += r->ipq_nfrags;
853 ip_freef(&ipq[i], r);
854 break;
855 }
856 }
857 } else {
858 ipstat.ips_fragtimeout += q->ipq_nfrags;
859 ip_freef(head, q);
860 }
861 }
862
863 found:
864 /*
865 * Adjust ip_len to not reflect header,
866 * convert offset of this to bytes.
867 */
868 ip->ip_len -= hlen;
869 if (ip->ip_off & IP_MF) {
870 /*
871 * Make sure that fragments have a data length
872 * that's a non-zero multiple of 8 bytes.
873 */
874 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
875 ipstat.ips_toosmall++; /* XXX */
876 goto dropfrag;
877 }
878 m->m_flags |= M_FRAG;
879 } else
880 m->m_flags &= ~M_FRAG;
881 ip->ip_off <<= 3;
882
883
884 /*
885 * Attempt reassembly; if it succeeds, proceed.
886 * ip_reass() will return a different mbuf.
887 */
888 ipstat.ips_fragments++;
889 m->m_pkthdr.header = ip;
890
891 /* Previous ip_reass() started here. */
892 /*
893 * Presence of header sizes in mbufs
894 * would confuse code below.
895 */
896 m->m_data += hlen;
897 m->m_len -= hlen;
898
899 /*
900 * If first fragment to arrive, create a reassembly queue.
901 */
902 if (fp == NULL) {
903 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
904 goto dropfrag;
905 fp = mtod(t, struct ipq *);
906 #ifdef MAC
907 if (mac_init_ipq(fp, M_NOWAIT) != 0) {
908 m_free(t);
909 goto dropfrag;
910 }
911 mac_create_ipq(m, fp);
912 #endif
913 TAILQ_INSERT_HEAD(head, fp, ipq_list);
914 nipq++;
915 fp->ipq_nfrags = 1;
916 fp->ipq_ttl = IPFRAGTTL;
917 fp->ipq_p = ip->ip_p;
918 fp->ipq_id = ip->ip_id;
919 fp->ipq_src = ip->ip_src;
920 fp->ipq_dst = ip->ip_dst;
921 fp->ipq_frags = m;
922 m->m_nextpkt = NULL;
923 goto inserted;
924 } else {
925 fp->ipq_nfrags++;
926 #ifdef MAC
927 mac_update_ipq(m, fp);
928 #endif
929 }
930
931 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
932
933 /*
934 * Handle ECN by comparing this segment with the first one;
935 * if CE is set, do not lose CE.
936 * drop if CE and not-ECT are mixed for the same packet.
937 */
938 ecn = ip->ip_tos & IPTOS_ECN_MASK;
939 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
940 if (ecn == IPTOS_ECN_CE) {
941 if (ecn0 == IPTOS_ECN_NOTECT)
942 goto dropfrag;
943 if (ecn0 != IPTOS_ECN_CE)
944 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
945 }
946 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
947 goto dropfrag;
948
949 /*
950 * Find a segment which begins after this one does.
951 */
952 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
953 if (GETIP(q)->ip_off > ip->ip_off)
954 break;
955
956 /*
957 * If there is a preceding segment, it may provide some of
958 * our data already. If so, drop the data from the incoming
959 * segment. If it provides all of our data, drop us, otherwise
960 * stick new segment in the proper place.
961 *
962 * If some of the data is dropped from the the preceding
963 * segment, then it's checksum is invalidated.
964 */
965 if (p) {
966 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
967 if (i > 0) {
968 if (i >= ip->ip_len)
969 goto dropfrag;
970 m_adj(m, i);
971 m->m_pkthdr.csum_flags = 0;
972 ip->ip_off += i;
973 ip->ip_len -= i;
974 }
975 m->m_nextpkt = p->m_nextpkt;
976 p->m_nextpkt = m;
977 } else {
978 m->m_nextpkt = fp->ipq_frags;
979 fp->ipq_frags = m;
980 }
981
982 /*
983 * While we overlap succeeding segments trim them or,
984 * if they are completely covered, dequeue them.
985 */
986 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
987 q = nq) {
988 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
989 if (i < GETIP(q)->ip_len) {
990 GETIP(q)->ip_len -= i;
991 GETIP(q)->ip_off += i;
992 m_adj(q, i);
993 q->m_pkthdr.csum_flags = 0;
994 break;
995 }
996 nq = q->m_nextpkt;
997 m->m_nextpkt = nq;
998 ipstat.ips_fragdropped++;
999 fp->ipq_nfrags--;
1000 m_freem(q);
1001 }
1002
1003 inserted:
1004
1005 /*
1006 * Check for complete reassembly and perform frag per packet
1007 * limiting.
1008 *
1009 * Frag limiting is performed here so that the nth frag has
1010 * a chance to complete the packet before we drop the packet.
1011 * As a result, n+1 frags are actually allowed per packet, but
1012 * only n will ever be stored. (n = maxfragsperpacket.)
1013 *
1014 */
1015 next = 0;
1016 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1017 if (GETIP(q)->ip_off != next) {
1018 if (fp->ipq_nfrags > maxfragsperpacket) {
1019 ipstat.ips_fragdropped += fp->ipq_nfrags;
1020 ip_freef(head, fp);
1021 }
1022 goto done;
1023 }
1024 next += GETIP(q)->ip_len;
1025 }
1026 /* Make sure the last packet didn't have the IP_MF flag */
1027 if (p->m_flags & M_FRAG) {
1028 if (fp->ipq_nfrags > maxfragsperpacket) {
1029 ipstat.ips_fragdropped += fp->ipq_nfrags;
1030 ip_freef(head, fp);
1031 }
1032 goto done;
1033 }
1034
1035 /*
1036 * Reassembly is complete. Make sure the packet is a sane size.
1037 */
1038 q = fp->ipq_frags;
1039 ip = GETIP(q);
1040 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1041 ipstat.ips_toolong++;
1042 ipstat.ips_fragdropped += fp->ipq_nfrags;
1043 ip_freef(head, fp);
1044 goto done;
1045 }
1046
1047 /*
1048 * Concatenate fragments.
1049 */
1050 m = q;
1051 t = m->m_next;
1052 m->m_next = NULL;
1053 m_cat(m, t);
1054 nq = q->m_nextpkt;
1055 q->m_nextpkt = NULL;
1056 for (q = nq; q != NULL; q = nq) {
1057 nq = q->m_nextpkt;
1058 q->m_nextpkt = NULL;
1059 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1060 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1061 m_cat(m, q);
1062 }
1063 /*
1064 * In order to do checksumming faster we do 'end-around carry' here
1065 * (and not in for{} loop), though it implies we are not going to
1066 * reassemble more than 64k fragments.
1067 */
1068 m->m_pkthdr.csum_data =
1069 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1070 #ifdef MAC
1071 mac_create_datagram_from_ipq(fp, m);
1072 mac_destroy_ipq(fp);
1073 #endif
1074
1075 /*
1076 * Create header for new ip packet by modifying header of first
1077 * packet; dequeue and discard fragment reassembly header.
1078 * Make header visible.
1079 */
1080 ip->ip_len = (ip->ip_hl << 2) + next;
1081 ip->ip_src = fp->ipq_src;
1082 ip->ip_dst = fp->ipq_dst;
1083 TAILQ_REMOVE(head, fp, ipq_list);
1084 nipq--;
1085 (void) m_free(dtom(fp));
1086 m->m_len += (ip->ip_hl << 2);
1087 m->m_data -= (ip->ip_hl << 2);
1088 /* some debugging cruft by sklower, below, will go away soon */
1089 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1090 m_fixhdr(m);
1091 ipstat.ips_reassembled++;
1092 IPQ_UNLOCK();
1093 return (m);
1094
1095 dropfrag:
1096 ipstat.ips_fragdropped++;
1097 if (fp != NULL)
1098 fp->ipq_nfrags--;
1099 m_freem(m);
1100 done:
1101 IPQ_UNLOCK();
1102 return (NULL);
1103
1104 #undef GETIP
1105 }
1106
1107 /*
1108 * Free a fragment reassembly header and all
1109 * associated datagrams.
1110 */
1111 static void
1112 ip_freef(fhp, fp)
1113 struct ipqhead *fhp;
1114 struct ipq *fp;
1115 {
1116 register struct mbuf *q;
1117
1118 IPQ_LOCK_ASSERT();
1119
1120 while (fp->ipq_frags) {
1121 q = fp->ipq_frags;
1122 fp->ipq_frags = q->m_nextpkt;
1123 m_freem(q);
1124 }
1125 TAILQ_REMOVE(fhp, fp, ipq_list);
1126 (void) m_free(dtom(fp));
1127 nipq--;
1128 }
1129
1130 /*
1131 * IP timer processing;
1132 * if a timer expires on a reassembly
1133 * queue, discard it.
1134 */
1135 void
1136 ip_slowtimo()
1137 {
1138 register struct ipq *fp;
1139 int i;
1140
1141 IPQ_LOCK();
1142 for (i = 0; i < IPREASS_NHASH; i++) {
1143 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1144 struct ipq *fpp;
1145
1146 fpp = fp;
1147 fp = TAILQ_NEXT(fp, ipq_list);
1148 if(--fpp->ipq_ttl == 0) {
1149 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1150 ip_freef(&ipq[i], fpp);
1151 }
1152 }
1153 }
1154 /*
1155 * If we are over the maximum number of fragments
1156 * (due to the limit being lowered), drain off
1157 * enough to get down to the new limit.
1158 */
1159 if (maxnipq >= 0 && nipq > maxnipq) {
1160 for (i = 0; i < IPREASS_NHASH; i++) {
1161 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1162 ipstat.ips_fragdropped +=
1163 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1164 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1165 }
1166 }
1167 }
1168 IPQ_UNLOCK();
1169 }
1170
1171 /*
1172 * Drain off all datagram fragments.
1173 */
1174 void
1175 ip_drain()
1176 {
1177 int i;
1178
1179 IPQ_LOCK();
1180 for (i = 0; i < IPREASS_NHASH; i++) {
1181 while(!TAILQ_EMPTY(&ipq[i])) {
1182 ipstat.ips_fragdropped +=
1183 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1184 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1185 }
1186 }
1187 IPQ_UNLOCK();
1188 in_rtqdrain();
1189 }
1190
1191 /*
1192 * Do option processing on a datagram,
1193 * possibly discarding it if bad options are encountered,
1194 * or forwarding it if source-routed.
1195 * The pass argument is used when operating in the IPSTEALTH
1196 * mode to tell what options to process:
1197 * [LS]SRR (pass 0) or the others (pass 1).
1198 * The reason for as many as two passes is that when doing IPSTEALTH,
1199 * non-routing options should be processed only if the packet is for us.
1200 * Returns 1 if packet has been forwarded/freed,
1201 * 0 if the packet should be processed further.
1202 */
1203 static int
1204 ip_dooptions(struct mbuf *m, int pass)
1205 {
1206 struct ip *ip = mtod(m, struct ip *);
1207 u_char *cp;
1208 struct in_ifaddr *ia;
1209 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1210 struct in_addr *sin, dst;
1211 n_time ntime;
1212 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1213
1214 /* ignore or reject packets with IP options */
1215 if (ip_doopts == 0)
1216 return 0;
1217 else if (ip_doopts == 2) {
1218 type = ICMP_UNREACH;
1219 code = ICMP_UNREACH_FILTER_PROHIB;
1220 goto bad;
1221 }
1222
1223 dst = ip->ip_dst;
1224 cp = (u_char *)(ip + 1);
1225 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1226 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1227 opt = cp[IPOPT_OPTVAL];
1228 if (opt == IPOPT_EOL)
1229 break;
1230 if (opt == IPOPT_NOP)
1231 optlen = 1;
1232 else {
1233 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1234 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1235 goto bad;
1236 }
1237 optlen = cp[IPOPT_OLEN];
1238 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1239 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1240 goto bad;
1241 }
1242 }
1243 switch (opt) {
1244
1245 default:
1246 break;
1247
1248 /*
1249 * Source routing with record.
1250 * Find interface with current destination address.
1251 * If none on this machine then drop if strictly routed,
1252 * or do nothing if loosely routed.
1253 * Record interface address and bring up next address
1254 * component. If strictly routed make sure next
1255 * address is on directly accessible net.
1256 */
1257 case IPOPT_LSRR:
1258 case IPOPT_SSRR:
1259 #ifdef IPSTEALTH
1260 if (ipstealth && pass > 0)
1261 break;
1262 #endif
1263 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1264 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1265 goto bad;
1266 }
1267 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1268 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1269 goto bad;
1270 }
1271 ipaddr.sin_addr = ip->ip_dst;
1272 ia = (struct in_ifaddr *)
1273 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1274 if (ia == NULL) {
1275 if (opt == IPOPT_SSRR) {
1276 type = ICMP_UNREACH;
1277 code = ICMP_UNREACH_SRCFAIL;
1278 goto bad;
1279 }
1280 if (!ip_dosourceroute)
1281 goto nosourcerouting;
1282 /*
1283 * Loose routing, and not at next destination
1284 * yet; nothing to do except forward.
1285 */
1286 break;
1287 }
1288 off--; /* 0 origin */
1289 if (off > optlen - (int)sizeof(struct in_addr)) {
1290 /*
1291 * End of source route. Should be for us.
1292 */
1293 if (!ip_acceptsourceroute)
1294 goto nosourcerouting;
1295 save_rte(m, cp, ip->ip_src);
1296 break;
1297 }
1298 #ifdef IPSTEALTH
1299 if (ipstealth)
1300 goto dropit;
1301 #endif
1302 if (!ip_dosourceroute) {
1303 if (ipforwarding) {
1304 char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1305 /*
1306 * Acting as a router, so generate ICMP
1307 */
1308 nosourcerouting:
1309 strcpy(buf, inet_ntoa(ip->ip_dst));
1310 log(LOG_WARNING,
1311 "attempted source route from %s to %s\n",
1312 inet_ntoa(ip->ip_src), buf);
1313 type = ICMP_UNREACH;
1314 code = ICMP_UNREACH_SRCFAIL;
1315 goto bad;
1316 } else {
1317 /*
1318 * Not acting as a router, so silently drop.
1319 */
1320 #ifdef IPSTEALTH
1321 dropit:
1322 #endif
1323 ipstat.ips_cantforward++;
1324 m_freem(m);
1325 return (1);
1326 }
1327 }
1328
1329 /*
1330 * locate outgoing interface
1331 */
1332 (void)memcpy(&ipaddr.sin_addr, cp + off,
1333 sizeof(ipaddr.sin_addr));
1334
1335 if (opt == IPOPT_SSRR) {
1336 #define INA struct in_ifaddr *
1337 #define SA struct sockaddr *
1338 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == NULL)
1339 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1340 } else
1341 ia = ip_rtaddr(ipaddr.sin_addr);
1342 if (ia == NULL) {
1343 type = ICMP_UNREACH;
1344 code = ICMP_UNREACH_SRCFAIL;
1345 goto bad;
1346 }
1347 ip->ip_dst = ipaddr.sin_addr;
1348 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1349 sizeof(struct in_addr));
1350 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1351 /*
1352 * Let ip_intr's mcast routing check handle mcast pkts
1353 */
1354 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1355 break;
1356
1357 case IPOPT_RR:
1358 #ifdef IPSTEALTH
1359 if (ipstealth && pass == 0)
1360 break;
1361 #endif
1362 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1363 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1364 goto bad;
1365 }
1366 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1367 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1368 goto bad;
1369 }
1370 /*
1371 * If no space remains, ignore.
1372 */
1373 off--; /* 0 origin */
1374 if (off > optlen - (int)sizeof(struct in_addr))
1375 break;
1376 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1377 sizeof(ipaddr.sin_addr));
1378 /*
1379 * locate outgoing interface; if we're the destination,
1380 * use the incoming interface (should be same).
1381 */
1382 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1383 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
1384 type = ICMP_UNREACH;
1385 code = ICMP_UNREACH_HOST;
1386 goto bad;
1387 }
1388 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1389 sizeof(struct in_addr));
1390 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1391 break;
1392
1393 case IPOPT_TS:
1394 #ifdef IPSTEALTH
1395 if (ipstealth && pass == 0)
1396 break;
1397 #endif
1398 code = cp - (u_char *)ip;
1399 if (optlen < 4 || optlen > 40) {
1400 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1401 goto bad;
1402 }
1403 if ((off = cp[IPOPT_OFFSET]) < 5) {
1404 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1405 goto bad;
1406 }
1407 if (off > optlen - (int)sizeof(int32_t)) {
1408 cp[IPOPT_OFFSET + 1] += (1 << 4);
1409 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1410 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1411 goto bad;
1412 }
1413 break;
1414 }
1415 off--; /* 0 origin */
1416 sin = (struct in_addr *)(cp + off);
1417 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1418
1419 case IPOPT_TS_TSONLY:
1420 break;
1421
1422 case IPOPT_TS_TSANDADDR:
1423 if (off + sizeof(n_time) +
1424 sizeof(struct in_addr) > optlen) {
1425 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1426 goto bad;
1427 }
1428 ipaddr.sin_addr = dst;
1429 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1430 m->m_pkthdr.rcvif);
1431 if (ia == NULL)
1432 continue;
1433 (void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1434 sizeof(struct in_addr));
1435 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1436 off += sizeof(struct in_addr);
1437 break;
1438
1439 case IPOPT_TS_PRESPEC:
1440 if (off + sizeof(n_time) +
1441 sizeof(struct in_addr) > optlen) {
1442 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1443 goto bad;
1444 }
1445 (void)memcpy(&ipaddr.sin_addr, sin,
1446 sizeof(struct in_addr));
1447 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1448 continue;
1449 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1450 off += sizeof(struct in_addr);
1451 break;
1452
1453 default:
1454 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1455 goto bad;
1456 }
1457 ntime = iptime();
1458 (void)memcpy(cp + off, &ntime, sizeof(n_time));
1459 cp[IPOPT_OFFSET] += sizeof(n_time);
1460 }
1461 }
1462 if (forward && ipforwarding) {
1463 ip_forward(m, 1);
1464 return (1);
1465 }
1466 return (0);
1467 bad:
1468 icmp_error(m, type, code, 0, 0);
1469 ipstat.ips_badoptions++;
1470 return (1);
1471 }
1472
1473 /*
1474 * Given address of next destination (final or next hop),
1475 * return internet address info of interface to be used to get there.
1476 */
1477 struct in_ifaddr *
1478 ip_rtaddr(dst)
1479 struct in_addr dst;
1480 {
1481 struct route sro;
1482 struct sockaddr_in *sin;
1483 struct in_ifaddr *ifa;
1484
1485 bzero(&sro, sizeof(sro));
1486 sin = (struct sockaddr_in *)&sro.ro_dst;
1487 sin->sin_family = AF_INET;
1488 sin->sin_len = sizeof(*sin);
1489 sin->sin_addr = dst;
1490 rtalloc_ign(&sro, RTF_CLONING);
1491
1492 if (sro.ro_rt == NULL)
1493 return (NULL);
1494
1495 ifa = ifatoia(sro.ro_rt->rt_ifa);
1496 RTFREE(sro.ro_rt);
1497 return (ifa);
1498 }
1499
1500 /*
1501 * Save incoming source route for use in replies,
1502 * to be picked up later by ip_srcroute if the receiver is interested.
1503 */
1504 static void
1505 save_rte(m, option, dst)
1506 struct mbuf *m;
1507 u_char *option;
1508 struct in_addr dst;
1509 {
1510 unsigned olen;
1511 struct ipopt_tag *opts;
1512
1513 opts = (struct ipopt_tag *)m_tag_get(PACKET_TAG_IPOPTIONS,
1514 sizeof(struct ipopt_tag), M_NOWAIT);
1515 if (opts == NULL)
1516 return;
1517
1518 olen = option[IPOPT_OLEN];
1519 #ifdef DIAGNOSTIC
1520 if (ipprintfs)
1521 printf("save_rte: olen %d\n", olen);
1522 #endif
1523 if (olen > sizeof(opts->ip_srcrt) - (1 + sizeof(dst)))
1524 return;
1525 bcopy(option, opts->ip_srcrt.srcopt, olen);
1526 opts->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1527 opts->ip_srcrt.dst = dst;
1528 m_tag_prepend(m, (struct m_tag *)opts);
1529 }
1530
1531 /*
1532 * Retrieve incoming source route for use in replies,
1533 * in the same form used by setsockopt.
1534 * The first hop is placed before the options, will be removed later.
1535 */
1536 struct mbuf *
1537 ip_srcroute(m0)
1538 struct mbuf *m0;
1539 {
1540 register struct in_addr *p, *q;
1541 register struct mbuf *m;
1542 struct ipopt_tag *opts;
1543
1544 opts = (struct ipopt_tag *)m_tag_find(m0, PACKET_TAG_IPOPTIONS, NULL);
1545 if (opts == NULL)
1546 return (NULL);
1547
1548 if (opts->ip_nhops == 0)
1549 return (NULL);
1550 m = m_get(M_DONTWAIT, MT_HEADER);
1551 if (m == NULL)
1552 return (NULL);
1553
1554 #define OPTSIZ (sizeof(opts->ip_srcrt.nop) + sizeof(opts->ip_srcrt.srcopt))
1555
1556 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1557 m->m_len = opts->ip_nhops * sizeof(struct in_addr) +
1558 sizeof(struct in_addr) + OPTSIZ;
1559 #ifdef DIAGNOSTIC
1560 if (ipprintfs)
1561 printf("ip_srcroute: nhops %d mlen %d", opts->ip_nhops, m->m_len);
1562 #endif
1563
1564 /*
1565 * First save first hop for return route
1566 */
1567 p = &(opts->ip_srcrt.route[opts->ip_nhops - 1]);
1568 *(mtod(m, struct in_addr *)) = *p--;
1569 #ifdef DIAGNOSTIC
1570 if (ipprintfs)
1571 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1572 #endif
1573
1574 /*
1575 * Copy option fields and padding (nop) to mbuf.
1576 */
1577 opts->ip_srcrt.nop = IPOPT_NOP;
1578 opts->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1579 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1580 &(opts->ip_srcrt.nop), OPTSIZ);
1581 q = (struct in_addr *)(mtod(m, caddr_t) +
1582 sizeof(struct in_addr) + OPTSIZ);
1583 #undef OPTSIZ
1584 /*
1585 * Record return path as an IP source route,
1586 * reversing the path (pointers are now aligned).
1587 */
1588 while (p >= opts->ip_srcrt.route) {
1589 #ifdef DIAGNOSTIC
1590 if (ipprintfs)
1591 printf(" %lx", (u_long)ntohl(q->s_addr));
1592 #endif
1593 *q++ = *p--;
1594 }
1595 /*
1596 * Last hop goes to final destination.
1597 */
1598 *q = opts->ip_srcrt.dst;
1599 #ifdef DIAGNOSTIC
1600 if (ipprintfs)
1601 printf(" %lx\n", (u_long)ntohl(q->s_addr));
1602 #endif
1603 m_tag_delete(m0, (struct m_tag *)opts);
1604 return (m);
1605 }
1606
1607 /*
1608 * Strip out IP options, at higher
1609 * level protocol in the kernel.
1610 * Second argument is buffer to which options
1611 * will be moved, and return value is their length.
1612 * XXX should be deleted; last arg currently ignored.
1613 */
1614 void
1615 ip_stripoptions(m, mopt)
1616 register struct mbuf *m;
1617 struct mbuf *mopt;
1618 {
1619 register int i;
1620 struct ip *ip = mtod(m, struct ip *);
1621 register caddr_t opts;
1622 int olen;
1623
1624 olen = (ip->ip_hl << 2) - sizeof (struct ip);
1625 opts = (caddr_t)(ip + 1);
1626 i = m->m_len - (sizeof (struct ip) + olen);
1627 bcopy(opts + olen, opts, (unsigned)i);
1628 m->m_len -= olen;
1629 if (m->m_flags & M_PKTHDR)
1630 m->m_pkthdr.len -= olen;
1631 ip->ip_v = IPVERSION;
1632 ip->ip_hl = sizeof(struct ip) >> 2;
1633 }
1634
1635 u_char inetctlerrmap[PRC_NCMDS] = {
1636 0, 0, 0, 0,
1637 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1638 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1639 EMSGSIZE, EHOSTUNREACH, 0, 0,
1640 0, 0, EHOSTUNREACH, 0,
1641 ENOPROTOOPT, ECONNREFUSED
1642 };
1643
1644 /*
1645 * Forward a packet. If some error occurs return the sender
1646 * an icmp packet. Note we can't always generate a meaningful
1647 * icmp message because icmp doesn't have a large enough repertoire
1648 * of codes and types.
1649 *
1650 * If not forwarding, just drop the packet. This could be confusing
1651 * if ipforwarding was zero but some routing protocol was advancing
1652 * us as a gateway to somewhere. However, we must let the routing
1653 * protocol deal with that.
1654 *
1655 * The srcrt parameter indicates whether the packet is being forwarded
1656 * via a source route.
1657 */
1658 void
1659 ip_forward(struct mbuf *m, int srcrt)
1660 {
1661 struct ip *ip = mtod(m, struct ip *);
1662 struct in_ifaddr *ia = NULL;
1663 int error, type = 0, code = 0;
1664 struct mbuf *mcopy;
1665 struct in_addr dest;
1666 struct ifnet *destifp, dummyifp;
1667
1668 #ifdef DIAGNOSTIC
1669 if (ipprintfs)
1670 printf("forward: src %lx dst %lx ttl %x\n",
1671 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1672 ip->ip_ttl);
1673 #endif
1674
1675
1676 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1677 ipstat.ips_cantforward++;
1678 m_freem(m);
1679 return;
1680 }
1681 #ifdef IPSTEALTH
1682 if (!ipstealth) {
1683 #endif
1684 if (ip->ip_ttl <= IPTTLDEC) {
1685 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1686 0, 0);
1687 return;
1688 }
1689 #ifdef IPSTEALTH
1690 }
1691 #endif
1692
1693 if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) {
1694 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1695 return;
1696 }
1697
1698 /*
1699 * Save the IP header and at most 8 bytes of the payload,
1700 * in case we need to generate an ICMP message to the src.
1701 *
1702 * XXX this can be optimized a lot by saving the data in a local
1703 * buffer on the stack (72 bytes at most), and only allocating the
1704 * mbuf if really necessary. The vast majority of the packets
1705 * are forwarded without having to send an ICMP back (either
1706 * because unnecessary, or because rate limited), so we are
1707 * really we are wasting a lot of work here.
1708 *
1709 * We don't use m_copy() because it might return a reference
1710 * to a shared cluster. Both this function and ip_output()
1711 * assume exclusive access to the IP header in `m', so any
1712 * data in a cluster may change before we reach icmp_error().
1713 */
1714 MGET(mcopy, M_DONTWAIT, m->m_type);
1715 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1716 /*
1717 * It's probably ok if the pkthdr dup fails (because
1718 * the deep copy of the tag chain failed), but for now
1719 * be conservative and just discard the copy since
1720 * code below may some day want the tags.
1721 */
1722 m_free(mcopy);
1723 mcopy = NULL;
1724 }
1725 if (mcopy != NULL) {
1726 mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1727 (int)ip->ip_len);
1728 mcopy->m_pkthdr.len = mcopy->m_len;
1729 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1730 }
1731
1732 #ifdef IPSTEALTH
1733 if (!ipstealth) {
1734 #endif
1735 ip->ip_ttl -= IPTTLDEC;
1736 #ifdef IPSTEALTH
1737 }
1738 #endif
1739
1740 /*
1741 * If forwarding packet using same interface that it came in on,
1742 * perhaps should send a redirect to sender to shortcut a hop.
1743 * Only send redirect if source is sending directly to us,
1744 * and if packet was not source routed (or has any options).
1745 * Also, don't send redirect if forwarding using a default route
1746 * or a route modified by a redirect.
1747 */
1748 dest.s_addr = 0;
1749 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1750 struct sockaddr_in *sin;
1751 struct route ro;
1752 struct rtentry *rt;
1753
1754 bzero(&ro, sizeof(ro));
1755 sin = (struct sockaddr_in *)&ro.ro_dst;
1756 sin->sin_family = AF_INET;
1757 sin->sin_len = sizeof(*sin);
1758 sin->sin_addr = ip->ip_dst;
1759 rtalloc_ign(&ro, RTF_CLONING);
1760
1761 rt = ro.ro_rt;
1762
1763 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1764 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1765 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1766 u_long src = ntohl(ip->ip_src.s_addr);
1767
1768 if (RTA(rt) &&
1769 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1770 if (rt->rt_flags & RTF_GATEWAY)
1771 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1772 else
1773 dest.s_addr = ip->ip_dst.s_addr;
1774 /* Router requirements says to only send host redirects */
1775 type = ICMP_REDIRECT;
1776 code = ICMP_REDIRECT_HOST;
1777 #ifdef DIAGNOSTIC
1778 if (ipprintfs)
1779 printf("redirect (%d) to %lx\n", code, (u_long)dest.s_addr);
1780 #endif
1781 }
1782 }
1783 if (rt)
1784 RTFREE(rt);
1785 }
1786
1787 error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
1788 if (error)
1789 ipstat.ips_cantforward++;
1790 else {
1791 ipstat.ips_forward++;
1792 if (type)
1793 ipstat.ips_redirectsent++;
1794 else {
1795 if (mcopy)
1796 m_freem(mcopy);
1797 return;
1798 }
1799 }
1800 if (mcopy == NULL)
1801 return;
1802 destifp = NULL;
1803
1804 switch (error) {
1805
1806 case 0: /* forwarded, but need redirect */
1807 /* type, code set above */
1808 break;
1809
1810 case ENETUNREACH: /* shouldn't happen, checked above */
1811 case EHOSTUNREACH:
1812 case ENETDOWN:
1813 case EHOSTDOWN:
1814 default:
1815 type = ICMP_UNREACH;
1816 code = ICMP_UNREACH_HOST;
1817 break;
1818
1819 case EMSGSIZE:
1820 type = ICMP_UNREACH;
1821 code = ICMP_UNREACH_NEEDFRAG;
1822 #if defined(IPSEC) || defined(FAST_IPSEC)
1823 /*
1824 * If the packet is routed over IPsec tunnel, tell the
1825 * originator the tunnel MTU.
1826 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1827 * XXX quickhack!!!
1828 */
1829 {
1830 struct secpolicy *sp = NULL;
1831 int ipsecerror;
1832 int ipsechdr;
1833 struct route *ro;
1834
1835 #ifdef IPSEC
1836 sp = ipsec4_getpolicybyaddr(mcopy,
1837 IPSEC_DIR_OUTBOUND,
1838 IP_FORWARDING,
1839 &ipsecerror);
1840 #else /* FAST_IPSEC */
1841 sp = ipsec_getpolicybyaddr(mcopy,
1842 IPSEC_DIR_OUTBOUND,
1843 IP_FORWARDING,
1844 &ipsecerror);
1845 #endif
1846 if (sp != NULL) {
1847 /* count IPsec header size */
1848 ipsechdr = ipsec4_hdrsiz(mcopy,
1849 IPSEC_DIR_OUTBOUND,
1850 NULL);
1851
1852 /*
1853 * find the correct route for outer IPv4
1854 * header, compute tunnel MTU.
1855 *
1856 * XXX BUG ALERT
1857 * The "dummyifp" code relies upon the fact
1858 * that icmp_error() touches only ifp->if_mtu.
1859 */
1860 /*XXX*/
1861 destifp = NULL;
1862 if (sp->req != NULL
1863 && sp->req->sav != NULL
1864 && sp->req->sav->sah != NULL) {
1865 ro = &sp->req->sav->sah->sa_route;
1866 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1867 dummyifp.if_mtu =
1868 ro->ro_rt->rt_rmx.rmx_mtu ?
1869 ro->ro_rt->rt_rmx.rmx_mtu :
1870 ro->ro_rt->rt_ifp->if_mtu;
1871 dummyifp.if_mtu -= ipsechdr;
1872 destifp = &dummyifp;
1873 }
1874 }
1875
1876 #ifdef IPSEC
1877 key_freesp(sp);
1878 #else /* FAST_IPSEC */
1879 KEY_FREESP(&sp);
1880 #endif
1881 ipstat.ips_cantfrag++;
1882 break;
1883 } else
1884 #endif /*IPSEC || FAST_IPSEC*/
1885 /*
1886 * When doing source routing 'ia' can be NULL. Fall back
1887 * to the minimum guaranteed routeable packet size and use
1888 * the same hack as IPSEC to setup a dummyifp for icmp.
1889 */
1890 if (ia == NULL) {
1891 dummyifp.if_mtu = IP_MSS;
1892 destifp = &dummyifp;
1893 } else
1894 destifp = ia->ia_ifp;
1895 #if defined(IPSEC) || defined(FAST_IPSEC)
1896 }
1897 #endif /*IPSEC || FAST_IPSEC*/
1898 ipstat.ips_cantfrag++;
1899 break;
1900
1901 case ENOBUFS:
1902 /*
1903 * A router should not generate ICMP_SOURCEQUENCH as
1904 * required in RFC1812 Requirements for IP Version 4 Routers.
1905 * Source quench could be a big problem under DoS attacks,
1906 * or if the underlying interface is rate-limited.
1907 * Those who need source quench packets may re-enable them
1908 * via the net.inet.ip.sendsourcequench sysctl.
1909 */
1910 if (ip_sendsourcequench == 0) {
1911 m_freem(mcopy);
1912 return;
1913 } else {
1914 type = ICMP_SOURCEQUENCH;
1915 code = 0;
1916 }
1917 break;
1918
1919 case EACCES: /* ipfw denied packet */
1920 m_freem(mcopy);
1921 return;
1922 }
1923 icmp_error(mcopy, type, code, dest.s_addr, destifp);
1924 }
1925
1926 void
1927 ip_savecontrol(inp, mp, ip, m)
1928 register struct inpcb *inp;
1929 register struct mbuf **mp;
1930 register struct ip *ip;
1931 register struct mbuf *m;
1932 {
1933 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1934 struct bintime bt;
1935
1936 bintime(&bt);
1937 if (inp->inp_socket->so_options & SO_BINTIME) {
1938 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1939 SCM_BINTIME, SOL_SOCKET);
1940 if (*mp)
1941 mp = &(*mp)->m_next;
1942 }
1943 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1944 struct timeval tv;
1945
1946 bintime2timeval(&bt, &tv);
1947 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1948 SCM_TIMESTAMP, SOL_SOCKET);
1949 if (*mp)
1950 mp = &(*mp)->m_next;
1951 }
1952 }
1953 if (inp->inp_flags & INP_RECVDSTADDR) {
1954 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1955 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1956 if (*mp)
1957 mp = &(*mp)->m_next;
1958 }
1959 if (inp->inp_flags & INP_RECVTTL) {
1960 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1961 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1962 if (*mp)
1963 mp = &(*mp)->m_next;
1964 }
1965 #ifdef notyet
1966 /* XXX
1967 * Moving these out of udp_input() made them even more broken
1968 * than they already were.
1969 */
1970 /* options were tossed already */
1971 if (inp->inp_flags & INP_RECVOPTS) {
1972 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1973 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1974 if (*mp)
1975 mp = &(*mp)->m_next;
1976 }
1977 /* ip_srcroute doesn't do what we want here, need to fix */
1978 if (inp->inp_flags & INP_RECVRETOPTS) {
1979 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1980 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1981 if (*mp)
1982 mp = &(*mp)->m_next;
1983 }
1984 #endif
1985 if (inp->inp_flags & INP_RECVIF) {
1986 struct ifnet *ifp;
1987 struct sdlbuf {
1988 struct sockaddr_dl sdl;
1989 u_char pad[32];
1990 } sdlbuf;
1991 struct sockaddr_dl *sdp;
1992 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1993
1994 if (((ifp = m->m_pkthdr.rcvif))
1995 && ( ifp->if_index && (ifp->if_index <= if_index))) {
1996 sdp = (struct sockaddr_dl *)
1997 (ifaddr_byindex(ifp->if_index)->ifa_addr);
1998 /*
1999 * Change our mind and don't try copy.
2000 */
2001 if ((sdp->sdl_family != AF_LINK)
2002 || (sdp->sdl_len > sizeof(sdlbuf))) {
2003 goto makedummy;
2004 }
2005 bcopy(sdp, sdl2, sdp->sdl_len);
2006 } else {
2007 makedummy:
2008 sdl2->sdl_len
2009 = offsetof(struct sockaddr_dl, sdl_data[0]);
2010 sdl2->sdl_family = AF_LINK;
2011 sdl2->sdl_index = 0;
2012 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2013 }
2014 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2015 IP_RECVIF, IPPROTO_IP);
2016 if (*mp)
2017 mp = &(*mp)->m_next;
2018 }
2019 }
2020
2021 /*
2022 * XXX these routines are called from the upper part of the kernel.
2023 * They need to be locked when we remove Giant.
2024 *
2025 * They could also be moved to ip_mroute.c, since all the RSVP
2026 * handling is done there already.
2027 */
2028 static int ip_rsvp_on;
2029 struct socket *ip_rsvpd;
2030 int
2031 ip_rsvp_init(struct socket *so)
2032 {
2033 if (so->so_type != SOCK_RAW ||
2034 so->so_proto->pr_protocol != IPPROTO_RSVP)
2035 return EOPNOTSUPP;
2036
2037 if (ip_rsvpd != NULL)
2038 return EADDRINUSE;
2039
2040 ip_rsvpd = so;
2041 /*
2042 * This may seem silly, but we need to be sure we don't over-increment
2043 * the RSVP counter, in case something slips up.
2044 */
2045 if (!ip_rsvp_on) {
2046 ip_rsvp_on = 1;
2047 rsvp_on++;
2048 }
2049
2050 return 0;
2051 }
2052
2053 int
2054 ip_rsvp_done(void)
2055 {
2056 ip_rsvpd = NULL;
2057 /*
2058 * This may seem silly, but we need to be sure we don't over-decrement
2059 * the RSVP counter, in case something slips up.
2060 */
2061 if (ip_rsvp_on) {
2062 ip_rsvp_on = 0;
2063 rsvp_on--;
2064 }
2065 return 0;
2066 }
2067
2068 void
2069 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
2070 {
2071 if (rsvp_input_p) { /* call the real one if loaded */
2072 rsvp_input_p(m, off);
2073 return;
2074 }
2075
2076 /* Can still get packets with rsvp_on = 0 if there is a local member
2077 * of the group to which the RSVP packet is addressed. But in this
2078 * case we want to throw the packet away.
2079 */
2080
2081 if (!rsvp_on) {
2082 m_freem(m);
2083 return;
2084 }
2085
2086 if (ip_rsvpd != NULL) {
2087 rip_input(m, off);
2088 return;
2089 }
2090 /* Drop the packet */
2091 m_freem(m);
2092 }
Cache object: 67238427b45fbf95507153e0efdceaec
|