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 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
34 * $FreeBSD$
35 * $ANA: ip_input.c,v 1.5 1996/09/18 14:34:59 wollman Exp $
36 */
37
38 #define _IP_VHL
39
40 #include "opt_bootp.h"
41 #include "opt_ipfw.h"
42 #include "opt_ipdn.h"
43 #include "opt_ipdivert.h"
44 #include "opt_ipfilter.h"
45 #include "opt_ipstealth.h"
46
47 #include <stddef.h>
48
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/mbuf.h>
52 #include <sys/malloc.h>
53 #include <sys/domain.h>
54 #include <sys/protosw.h>
55 #include <sys/socket.h>
56 #include <sys/time.h>
57 #include <sys/kernel.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60
61 #include <net/if.h>
62 #include <net/if_var.h>
63 #include <net/if_dl.h>
64 #include <net/route.h>
65 #include <net/netisr.h>
66
67 #include <netinet/in.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/in_var.h>
70 #include <netinet/ip.h>
71 #include <netinet/in_pcb.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/ip_icmp.h>
74 #include <machine/in_cksum.h>
75
76 #include <sys/socketvar.h>
77
78 #ifdef IPFIREWALL
79 #include <netinet/ip_fw.h>
80 #endif
81
82 #ifdef DUMMYNET
83 #include <netinet/ip_dummynet.h>
84 #endif
85
86 int rsvp_on = 0;
87 static int ip_rsvp_on;
88 struct socket *ip_rsvpd;
89
90 int ipforwarding = 0;
91 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
92 &ipforwarding, 0, "");
93
94 static int ipsendredirects = 1; /* XXX */
95 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
96 &ipsendredirects, 0, "");
97
98 int ip_defttl = IPDEFTTL;
99 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
100 &ip_defttl, 0, "");
101
102 static int ip_dosourceroute = 0;
103 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
104 &ip_dosourceroute, 0, "");
105
106 static int ip_acceptsourceroute = 0;
107 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
108 CTLFLAG_RW, &ip_acceptsourceroute, 0, "");
109 #ifdef DIAGNOSTIC
110 static int ipprintfs = 0;
111 #endif
112
113 extern struct domain inetdomain;
114 extern struct protosw inetsw[];
115 u_char ip_protox[IPPROTO_MAX];
116 static int ipqmaxlen = IFQ_MAXLEN;
117 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
118 struct ifqueue ipintrq;
119 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RD,
120 &ipintrq.ifq_maxlen, 0, "");
121 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
122 &ipintrq.ifq_drops, 0, "");
123
124 struct ipstat ipstat;
125 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD,
126 &ipstat, ipstat, "");
127
128 /* Packet reassembly stuff */
129 #define IPREASS_NHASH_LOG2 6
130 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
131 #define IPREASS_HMASK (IPREASS_NHASH - 1)
132 #define IPREASS_HASH(x,y) \
133 ((((x) & 0xF | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
134
135 static struct ipq ipq[IPREASS_NHASH];
136 static int nipq = 0; /* total # of reass queues */
137 static int maxnipq;
138
139 #ifdef IPCTL_DEFMTU
140 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
141 &ip_mtu, 0, "");
142 #endif
143
144 #ifdef IPSTEALTH
145 static int ipstealth = 0;
146 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
147 &ipstealth, 0, "");
148 #endif
149
150 #if !defined(COMPAT_IPFW) || COMPAT_IPFW == 1
151 #undef COMPAT_IPFW
152 #define COMPAT_IPFW 1
153 #else
154 #undef COMPAT_IPFW
155 #endif
156
157 #ifdef COMPAT_IPFW
158
159 #include <netinet/ip_fw.h>
160
161 /* Firewall hooks */
162 ip_fw_chk_t *ip_fw_chk_ptr;
163 ip_fw_ctl_t *ip_fw_ctl_ptr;
164 int fw_enable = 1 ;
165
166 #ifdef DUMMYNET
167 ip_dn_ctl_t *ip_dn_ctl_ptr;
168 #endif
169
170 #endif
171
172 #if defined(IPFILTER_LKM) || defined(IPFILTER)
173 int iplattach __P((void));
174 int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL;
175 #endif
176
177
178 static int ip_nfragpackets = 0;
179 static int ip_maxfragpackets; /* initialized in ip_init() */
180 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
181 &ip_maxfragpackets, 0,
182 "Maximum number of IPv4 fragment reassembly queue entries");
183
184 /*
185 * We need to save the IP options in case a protocol wants to respond
186 * to an incoming packet over the same route if the packet got here
187 * using IP source routing. This allows connection establishment and
188 * maintenance when the remote end is on a network that is not known
189 * to us.
190 */
191 static int ip_nhops = 0;
192 static struct ip_srcrt {
193 struct in_addr dst; /* final destination */
194 char nop; /* one NOP to align */
195 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
196 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
197 } ip_srcrt;
198
199 #ifdef IPDIVERT
200 /*
201 * Shared variable between ip_input() and ip_reass() to communicate
202 * about which packets, once assembled from fragments, get diverted,
203 * and to which port.
204 */
205 static u_short frag_divert_port;
206 #endif
207
208 struct sockaddr_in *ip_fw_fwd_addr;
209
210 static void save_rte __P((u_char *, struct in_addr));
211 static int ip_dooptions __P((struct mbuf *));
212 static void ip_forward __P((struct mbuf *, int));
213 static void ip_freef __P((struct ipq *));
214 static struct ip *
215 ip_reass __P((struct mbuf *, struct ipq *, struct ipq *));
216 static struct in_ifaddr *
217 ip_rtaddr __P((struct in_addr));
218 static void ipintr __P((void));
219 /*
220 * IP initialization: fill in IP protocol switch table.
221 * All protocols not implemented in kernel go to raw IP protocol handler.
222 */
223 void
224 ip_init()
225 {
226 register struct protosw *pr;
227 register int i;
228
229 TAILQ_INIT(&in_ifaddrhead);
230 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
231 if (pr == 0)
232 panic("ip_init");
233 for (i = 0; i < IPPROTO_MAX; i++)
234 ip_protox[i] = pr - inetsw;
235 for (pr = inetdomain.dom_protosw;
236 pr < inetdomain.dom_protoswNPROTOSW; pr++)
237 if (pr->pr_domain->dom_family == PF_INET &&
238 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
239 ip_protox[pr->pr_protocol] = pr - inetsw;
240
241 for (i = 0; i < IPREASS_NHASH; i++)
242 ipq[i].next = ipq[i].prev = &ipq[i];
243
244 maxnipq = nmbclusters / 4;
245 ip_maxfragpackets = nmbclusters / 4;
246
247 ip_id = time_second & 0xffff;
248 ipintrq.ifq_maxlen = ipqmaxlen;
249 #ifdef DUMMYNET
250 ip_dn_init();
251 #endif
252 #ifdef IPNAT
253 ip_nat_init();
254 #endif
255 #ifdef IPFILTER
256 iplattach();
257 #endif
258
259 }
260
261 static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
262 static struct route ipforward_rt;
263
264 /*
265 * Ip input routine. Checksum and byte swap header. If fragmented
266 * try to reassemble. Process options. Pass to next level.
267 */
268 void
269 ip_input(struct mbuf *m)
270 {
271 struct ip *ip;
272 struct ipq *fp;
273 struct in_ifaddr *ia;
274 int i, hlen, mff;
275 u_short sum;
276 #ifndef IPDIVERT /* dummy variable for the firewall code to play with */
277 u_short ip_divert_cookie = 0 ;
278 #endif
279 #ifdef COMPAT_IPFW
280 struct ip_fw_chain *rule = NULL ;
281 #endif
282
283 #if defined(IPFIREWALL) && defined(DUMMYNET)
284 /*
285 * dummynet packet are prepended a vestigial mbuf with
286 * m_type = MT_DUMMYNET and m_data pointing to the matching
287 * rule.
288 */
289 if (m->m_type == MT_DUMMYNET) {
290 rule = (struct ip_fw_chain *)(m->m_data) ;
291 m = m->m_next ;
292 ip = mtod(m, struct ip *);
293 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
294 goto iphack ;
295 } else
296 rule = NULL ;
297 #endif
298
299 #ifdef DIAGNOSTIC
300 if (m == NULL || (m->m_flags & M_PKTHDR) == 0)
301 panic("ip_input no HDR");
302 #endif
303 /*
304 * If no IP addresses have been set yet but the interfaces
305 * are receiving, can't do anything with incoming packets yet.
306 * XXX This is broken! We should be able to receive broadcasts
307 * and multicasts even without any local addresses configured.
308 */
309 if (TAILQ_EMPTY(&in_ifaddrhead))
310 goto bad;
311 ipstat.ips_total++;
312
313 if (m->m_pkthdr.len < sizeof(struct ip))
314 goto tooshort;
315
316 if (m->m_len < sizeof (struct ip) &&
317 (m = m_pullup(m, sizeof (struct ip))) == 0) {
318 ipstat.ips_toosmall++;
319 return;
320 }
321 ip = mtod(m, struct ip *);
322
323 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
324 ipstat.ips_badvers++;
325 goto bad;
326 }
327
328 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
329 if (hlen < sizeof(struct ip)) { /* minimum header length */
330 ipstat.ips_badhlen++;
331 goto bad;
332 }
333 if (hlen > m->m_len) {
334 if ((m = m_pullup(m, hlen)) == 0) {
335 ipstat.ips_badhlen++;
336 return;
337 }
338 ip = mtod(m, struct ip *);
339 }
340 if (hlen == sizeof(struct ip)) {
341 sum = in_cksum_hdr(ip);
342 } else {
343 sum = in_cksum(m, hlen);
344 }
345 if (sum) {
346 ipstat.ips_badsum++;
347 goto bad;
348 }
349
350 /*
351 * Convert fields to host representation.
352 */
353 NTOHS(ip->ip_len);
354 if (ip->ip_len < hlen) {
355 ipstat.ips_badlen++;
356 goto bad;
357 }
358 NTOHS(ip->ip_id);
359 NTOHS(ip->ip_off);
360
361 /*
362 * Check that the amount of data in the buffers
363 * is as at least much as the IP header would have us expect.
364 * Trim mbufs if longer than we expect.
365 * Drop packet if shorter than we expect.
366 */
367 if (m->m_pkthdr.len < ip->ip_len) {
368 tooshort:
369 ipstat.ips_tooshort++;
370 goto bad;
371 }
372 if (m->m_pkthdr.len > ip->ip_len) {
373 if (m->m_len == m->m_pkthdr.len) {
374 m->m_len = ip->ip_len;
375 m->m_pkthdr.len = ip->ip_len;
376 } else
377 m_adj(m, ip->ip_len - m->m_pkthdr.len);
378 }
379 /*
380 * IpHack's section.
381 * Right now when no processing on packet has done
382 * and it is still fresh out of network we do our black
383 * deals with it.
384 * - Firewall: deny/allow/divert
385 * - Xlate: translate packet's addr/port (NAT).
386 * - Pipe: pass pkt through dummynet.
387 * - Wrap: fake packet's addr/port <unimpl.>
388 * - Encapsulate: put it in another IP and send out. <unimp.>
389 */
390
391 #if defined(IPFIREWALL) && defined(DUMMYNET)
392 iphack:
393 #endif
394 #if defined(IPFILTER) || defined(IPFILTER_LKM)
395 /*
396 * Check if we want to allow this packet to be processed.
397 * Consider it to be bad if not.
398 */
399 if (fr_checkp) {
400 struct mbuf *m1 = m;
401
402 if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1)
403 return;
404 ip = mtod(m = m1, struct ip *);
405 }
406 #endif
407 #ifdef COMPAT_IPFW
408 if (fw_enable && ip_fw_chk_ptr) {
409 #ifdef IPFIREWALL_FORWARD
410 /*
411 * If we've been forwarded from the output side, then
412 * skip the firewall a second time
413 */
414 if (ip_fw_fwd_addr)
415 goto ours;
416 #endif /* IPFIREWALL_FORWARD */
417 i = (*ip_fw_chk_ptr)(&ip, hlen, NULL, &ip_divert_cookie,
418 &m, &rule, &ip_fw_fwd_addr);
419 /*
420 * see the comment in ip_output for the return values
421 * produced by the firewall.
422 */
423 if (!m) /* packet discarded by firewall */
424 return ;
425 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */
426 goto pass ;
427 #ifdef DUMMYNET
428 if (i & 0x10000) {
429 /* send packet to the appropriate pipe */
430 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule);
431 return ;
432 }
433 #endif
434 #ifdef IPDIVERT
435 if (i > 0 && i < 0x10000) {
436 /* Divert packet */
437 frag_divert_port = i & 0xffff ;
438 goto ours;
439 }
440 #endif
441 #ifdef IPFIREWALL_FORWARD
442 if (i == 0 && ip_fw_fwd_addr != NULL)
443 goto pass ;
444 #endif
445 /*
446 * if we get here, the packet must be dropped
447 */
448 m_freem(m);
449 return;
450 }
451 pass:
452
453 #endif /* !COMPAT_IPFW */
454
455 /*
456 * Process options and, if not destined for us,
457 * ship it on. ip_dooptions returns 1 when an
458 * error was detected (causing an icmp message
459 * to be sent and the original packet to be freed).
460 */
461 ip_nhops = 0; /* for source routed packets */
462 if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
463 #ifdef IPFIREWALL_FORWARD
464 ip_fw_fwd_addr = NULL;
465 #endif
466 return;
467 }
468
469 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
470 * matter if it is destined to another node, or whether it is
471 * a multicast one, RSVP wants it! and prevents it from being forwarded
472 * anywhere else. Also checks if the rsvp daemon is running before
473 * grabbing the packet.
474 */
475 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
476 goto ours;
477
478 /*
479 * Check our list of addresses, to see if the packet is for us.
480 */
481 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia;
482 ia = TAILQ_NEXT(ia, ia_link)) {
483 #define satosin(sa) ((struct sockaddr_in *)(sa))
484
485 #ifdef BOOTP_COMPAT
486 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
487 goto ours;
488 #endif
489 #ifdef IPFIREWALL_FORWARD
490 /*
491 * If the addr to forward to is one of ours, we pretend to
492 * be the destination for this packet.
493 */
494 if (ip_fw_fwd_addr == NULL) {
495 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
496 goto ours;
497 } else if (IA_SIN(ia)->sin_addr.s_addr ==
498 ip_fw_fwd_addr->sin_addr.s_addr)
499 goto ours;
500 #else
501 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
502 goto ours;
503 #endif
504 if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) {
505 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
506 ip->ip_dst.s_addr)
507 goto ours;
508 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr)
509 goto ours;
510 }
511 }
512 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
513 struct in_multi *inm;
514 if (ip_mrouter) {
515 /*
516 * If we are acting as a multicast router, all
517 * incoming multicast packets are passed to the
518 * kernel-level multicast forwarding function.
519 * The packet is returned (relatively) intact; if
520 * ip_mforward() returns a non-zero value, the packet
521 * must be discarded, else it may be accepted below.
522 *
523 * (The IP ident field is put in the same byte order
524 * as expected when ip_mforward() is called from
525 * ip_output().)
526 */
527 ip->ip_id = htons(ip->ip_id);
528 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
529 ipstat.ips_cantforward++;
530 m_freem(m);
531 return;
532 }
533 ip->ip_id = ntohs(ip->ip_id);
534
535 /*
536 * The process-level routing demon needs to receive
537 * all multicast IGMP packets, whether or not this
538 * host belongs to their destination groups.
539 */
540 if (ip->ip_p == IPPROTO_IGMP)
541 goto ours;
542 ipstat.ips_forward++;
543 }
544 /*
545 * See if we belong to the destination multicast group on the
546 * arrival interface.
547 */
548 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
549 if (inm == NULL) {
550 ipstat.ips_notmember++;
551 m_freem(m);
552 return;
553 }
554 goto ours;
555 }
556 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
557 goto ours;
558 if (ip->ip_dst.s_addr == INADDR_ANY)
559 goto ours;
560
561 /*
562 * Not for us; forward if possible and desirable.
563 */
564 if (ipforwarding == 0) {
565 ipstat.ips_cantforward++;
566 m_freem(m);
567 } else
568 ip_forward(m, 0);
569 #ifdef IPFIREWALL_FORWARD
570 ip_fw_fwd_addr = NULL;
571 #endif
572 return;
573
574 ours:
575
576 /*
577 * If offset or IP_MF are set, must reassemble.
578 * Otherwise, nothing need be done.
579 * (We could look in the reassembly queue to see
580 * if the packet was previously fragmented,
581 * but it's not worth the time; just let them time out.)
582 */
583 if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) {
584 if (m->m_flags & M_EXT) { /* XXX */
585 if ((m = m_pullup(m, hlen)) == 0) {
586 ipstat.ips_toosmall++;
587 #ifdef IPDIVERT
588 frag_divert_port = 0;
589 ip_divert_cookie = 0;
590 #endif
591 #ifdef IPFIREWALL_FORWARD
592 ip_fw_fwd_addr = NULL;
593 #endif
594 return;
595 }
596 ip = mtod(m, struct ip *);
597 }
598 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
599 /*
600 * Look for queue of fragments
601 * of this datagram.
602 */
603 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
604 if (ip->ip_id == fp->ipq_id &&
605 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
606 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
607 ip->ip_p == fp->ipq_p)
608 goto found;
609
610 fp = 0;
611
612 /* check if there's a place for the new queue */
613 if (nipq > maxnipq) {
614 /*
615 * drop something from the tail of the current queue
616 * before proceeding further
617 */
618 if (ipq[sum].prev == &ipq[sum]) { /* gak */
619 for (i = 0; i < IPREASS_NHASH; i++) {
620 if (ipq[i].prev != &ipq[i]) {
621 ip_freef(ipq[i].prev);
622 break;
623 }
624 }
625 } else
626 ip_freef(ipq[sum].prev);
627 }
628 found:
629 /*
630 * Adjust ip_len to not reflect header,
631 * set ip_mff if more fragments are expected,
632 * convert offset of this to bytes.
633 */
634 ip->ip_len -= hlen;
635 mff = (ip->ip_off & IP_MF) != 0;
636 if (mff) {
637 /*
638 * Make sure that fragments have a data length
639 * that's a non-zero multiple of 8 bytes.
640 */
641 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
642 ipstat.ips_toosmall++; /* XXX */
643 goto bad;
644 }
645 m->m_flags |= M_FRAG;
646 }
647 ip->ip_off <<= 3;
648
649 /*
650 * If datagram marked as having more fragments
651 * or if this is not the first fragment,
652 * attempt reassembly; if it succeeds, proceed.
653 */
654 if (mff || ip->ip_off) {
655 ipstat.ips_fragments++;
656 m->m_pkthdr.header = ip;
657 ip = ip_reass(m, fp, &ipq[sum]);
658 if (ip == 0) {
659 #ifdef IPFIREWALL_FORWARD
660 ip_fw_fwd_addr = NULL;
661 #endif
662 return;
663 }
664 /* Get the length of the reassembled packets header */
665 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
666 ipstat.ips_reassembled++;
667 m = dtom(ip);
668 #ifdef IPDIVERT
669 if (frag_divert_port) {
670 ip->ip_len += hlen;
671 HTONS(ip->ip_len);
672 HTONS(ip->ip_off);
673 HTONS(ip->ip_id);
674 ip->ip_sum = 0;
675 ip->ip_sum = in_cksum_hdr(ip);
676 NTOHS(ip->ip_id);
677 NTOHS(ip->ip_off);
678 NTOHS(ip->ip_len);
679 ip->ip_len -= hlen;
680 }
681 #endif
682 } else
683 if (fp)
684 ip_freef(fp);
685 } else
686 ip->ip_len -= hlen;
687
688 #ifdef IPDIVERT
689 /*
690 * Divert reassembled packets to the divert protocol if required
691 * If divert port is null then cookie should be too,
692 * so we shouldn't need to clear them here. Assume ip_divert does so.
693 */
694 if (frag_divert_port) {
695 ipstat.ips_delivered++;
696 ip_divert_port = frag_divert_port;
697 frag_divert_port = 0;
698 (*inetsw[ip_protox[IPPROTO_DIVERT]].pr_input)(m, hlen);
699 return;
700 }
701
702 /* Don't let packets divert themselves */
703 if (ip->ip_p == IPPROTO_DIVERT) {
704 ipstat.ips_noproto++;
705 goto bad;
706 }
707
708 #endif
709
710 /*
711 * Switch out to protocol's input routine.
712 */
713 ipstat.ips_delivered++;
714 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
715 #ifdef IPFIREWALL_FORWARD
716 ip_fw_fwd_addr = NULL; /* tcp needed it */
717 #endif
718 return;
719 bad:
720 #ifdef IPFIREWALL_FORWARD
721 ip_fw_fwd_addr = NULL;
722 #endif
723 m_freem(m);
724 }
725
726 /*
727 * IP software interrupt routine - to go away sometime soon
728 */
729 static void
730 ipintr(void)
731 {
732 int s;
733 struct mbuf *m;
734
735 while(1) {
736 s = splimp();
737 IF_DEQUEUE(&ipintrq, m);
738 splx(s);
739 if (m == 0)
740 return;
741 ip_input(m);
742 }
743 }
744
745 NETISR_SET(NETISR_IP, ipintr);
746
747 /*
748 * Take incoming datagram fragment and try to
749 * reassemble it into whole datagram. If a chain for
750 * reassembly of this datagram already exists, then it
751 * is given as fp; otherwise have to make a chain.
752 */
753 static struct ip *
754 ip_reass(m, fp, where)
755 register struct mbuf *m;
756 register struct ipq *fp;
757 struct ipq *where;
758 {
759 struct ip *ip = mtod(m, struct ip *);
760 register struct mbuf *p = 0, *q, *nq;
761 struct mbuf *t;
762 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
763 int i, next;
764
765 /*
766 * Presence of header sizes in mbufs
767 * would confuse code below.
768 */
769 m->m_data += hlen;
770 m->m_len -= hlen;
771
772 /*
773 * If first fragment to arrive, create a reassembly queue.
774 */
775 if (fp == 0) {
776 /*
777 * Enforce upper bound on number of fragmented packets
778 * for which we attempt reassembly;
779 * If maxfrag is 0, never accept fragments.
780 * If maxfrag is -1, accept all fragments without limitation.
781 */
782 if ((ip_maxfragpackets >= 0) && (ip_nfragpackets >= ip_maxfragpackets))
783 goto dropfrag;
784 ip_nfragpackets++;
785 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
786 goto dropfrag;
787 fp = mtod(t, struct ipq *);
788 insque(fp, where);
789 nipq++;
790 fp->ipq_ttl = IPFRAGTTL;
791 fp->ipq_p = ip->ip_p;
792 fp->ipq_id = ip->ip_id;
793 fp->ipq_src = ip->ip_src;
794 fp->ipq_dst = ip->ip_dst;
795 fp->ipq_frags = m;
796 m->m_nextpkt = NULL;
797 #ifdef IPDIVERT
798 fp->ipq_divert = 0;
799 fp->ipq_div_cookie = 0;
800 #endif
801 goto inserted;
802 }
803
804 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
805
806 /*
807 * Find a segment which begins after this one does.
808 */
809 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
810 if (GETIP(q)->ip_off > ip->ip_off)
811 break;
812
813 /*
814 * If there is a preceding segment, it may provide some of
815 * our data already. If so, drop the data from the incoming
816 * segment. If it provides all of our data, drop us, otherwise
817 * stick new segment in the proper place.
818 */
819 if (p) {
820 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
821 if (i > 0) {
822 if (i >= ip->ip_len)
823 goto dropfrag;
824 m_adj(dtom(ip), i);
825 ip->ip_off += i;
826 ip->ip_len -= i;
827 }
828 m->m_nextpkt = p->m_nextpkt;
829 p->m_nextpkt = m;
830 } else {
831 m->m_nextpkt = fp->ipq_frags;
832 fp->ipq_frags = m;
833 }
834
835 /*
836 * While we overlap succeeding segments trim them or,
837 * if they are completely covered, dequeue them.
838 */
839 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
840 q = nq) {
841 i = (ip->ip_off + ip->ip_len) -
842 GETIP(q)->ip_off;
843 if (i < GETIP(q)->ip_len) {
844 GETIP(q)->ip_len -= i;
845 GETIP(q)->ip_off += i;
846 m_adj(q, i);
847 break;
848 }
849 nq = q->m_nextpkt;
850 m->m_nextpkt = nq;
851 m_freem(q);
852 }
853
854 inserted:
855
856 #ifdef IPDIVERT
857 /*
858 * Any fragment diverting causes the whole packet to divert
859 */
860 if (frag_divert_port) {
861 fp->ipq_divert = frag_divert_port;
862 fp->ipq_div_cookie = ip_divert_cookie;
863 }
864 frag_divert_port = 0;
865 ip_divert_cookie = 0;
866 #endif
867
868 /*
869 * Check for complete reassembly.
870 */
871 next = 0;
872 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
873 if (GETIP(q)->ip_off != next)
874 return (0);
875 next += GETIP(q)->ip_len;
876 }
877 /* Make sure the last packet didn't have the IP_MF flag */
878 if (p->m_flags & M_FRAG)
879 return (0);
880
881 /*
882 * Reassembly is complete. Make sure the packet is a sane size.
883 */
884 q = fp->ipq_frags;
885 ip = GETIP(q);
886 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
887 ipstat.ips_toolong++;
888 ip_freef(fp);
889 return (0);
890 }
891
892 /*
893 * Concatenate fragments.
894 */
895 m = q;
896 t = m->m_next;
897 m->m_next = 0;
898 m_cat(m, t);
899 nq = q->m_nextpkt;
900 q->m_nextpkt = 0;
901 for (q = nq; q != NULL; q = nq) {
902 nq = q->m_nextpkt;
903 q->m_nextpkt = NULL;
904 m_cat(m, q);
905 }
906
907 #ifdef IPDIVERT
908 /*
909 * extract divert port for packet, if any
910 */
911 frag_divert_port = fp->ipq_divert;
912 ip_divert_cookie = fp->ipq_div_cookie;
913 #endif
914
915 /*
916 * Create header for new ip packet by
917 * modifying header of first packet;
918 * dequeue and discard fragment reassembly header.
919 * Make header visible.
920 */
921 ip->ip_len = next;
922 ip->ip_src = fp->ipq_src;
923 ip->ip_dst = fp->ipq_dst;
924 remque(fp);
925 nipq--;
926 (void) m_free(dtom(fp));
927 ip_nfragpackets--;
928 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
929 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
930 /* some debugging cruft by sklower, below, will go away soon */
931 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
932 register int plen = 0;
933 for (t = m; m; m = m->m_next)
934 plen += m->m_len;
935 t->m_pkthdr.len = plen;
936 }
937 return (ip);
938
939 dropfrag:
940 #ifdef IPDIVERT
941 frag_divert_port = 0;
942 ip_divert_cookie = 0;
943 #endif
944 ipstat.ips_fragdropped++;
945 m_freem(m);
946 return (0);
947
948 #undef GETIP
949 }
950
951 /*
952 * Free a fragment reassembly header and all
953 * associated datagrams.
954 */
955 static void
956 ip_freef(fp)
957 struct ipq *fp;
958 {
959 register struct mbuf *q;
960
961 while (fp->ipq_frags) {
962 q = fp->ipq_frags;
963 fp->ipq_frags = q->m_nextpkt;
964 m_freem(q);
965 }
966 remque(fp);
967 (void) m_free(dtom(fp));
968 ip_nfragpackets--;
969 nipq--;
970 }
971
972 /*
973 * IP timer processing;
974 * if a timer expires on a reassembly
975 * queue, discard it.
976 */
977 void
978 ip_slowtimo()
979 {
980 register struct ipq *fp;
981 int s = splnet();
982 int i;
983
984 for (i = 0; i < IPREASS_NHASH; i++) {
985 fp = ipq[i].next;
986 if (fp == 0)
987 continue;
988 while (fp != &ipq[i]) {
989 --fp->ipq_ttl;
990 fp = fp->next;
991 if (fp->prev->ipq_ttl == 0) {
992 ipstat.ips_fragtimeout++;
993 ip_freef(fp->prev);
994 }
995 }
996 }
997 /*
998 * If we are over the maximum number of fragments
999 * (due to the limit being lowered), drain off
1000 * enough to get down to the new limit.
1001 */
1002 for (i = 0; i < IPREASS_NHASH; i++) {
1003 if (ip_maxfragpackets >= 0) {
1004 while ((ip_nfragpackets > ip_maxfragpackets) &&
1005 (ipq[i].next != &ipq[i])) {
1006 ipstat.ips_fragdropped++;
1007 ip_freef(ipq[i].next);
1008 }
1009 }
1010 }
1011 ipflow_slowtimo();
1012 splx(s);
1013 }
1014
1015 /*
1016 * Drain off all datagram fragments.
1017 */
1018 void
1019 ip_drain()
1020 {
1021 int i;
1022
1023 for (i = 0; i < IPREASS_NHASH; i++) {
1024 while (ipq[i].next != &ipq[i]) {
1025 ipstat.ips_fragdropped++;
1026 ip_freef(ipq[i].next);
1027 }
1028 }
1029 in_rtqdrain();
1030 }
1031
1032 /*
1033 * Do option processing on a datagram,
1034 * possibly discarding it if bad options are encountered,
1035 * or forwarding it if source-routed.
1036 * Returns 1 if packet has been forwarded/freed,
1037 * 0 if the packet should be processed further.
1038 */
1039 static int
1040 ip_dooptions(m)
1041 struct mbuf *m;
1042 {
1043 register struct ip *ip = mtod(m, struct ip *);
1044 register u_char *cp;
1045 register struct ip_timestamp *ipt;
1046 register struct in_ifaddr *ia;
1047 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1048 struct in_addr *sin, dst;
1049 n_time ntime;
1050
1051 dst = ip->ip_dst;
1052 cp = (u_char *)(ip + 1);
1053 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1054 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1055 opt = cp[IPOPT_OPTVAL];
1056 if (opt == IPOPT_EOL)
1057 break;
1058 if (opt == IPOPT_NOP)
1059 optlen = 1;
1060 else {
1061 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1062 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1063 goto bad;
1064 }
1065 optlen = cp[IPOPT_OLEN];
1066 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1067 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1068 goto bad;
1069 }
1070 }
1071 switch (opt) {
1072
1073 default:
1074 break;
1075
1076 /*
1077 * Source routing with record.
1078 * Find interface with current destination address.
1079 * If none on this machine then drop if strictly routed,
1080 * or do nothing if loosely routed.
1081 * Record interface address and bring up next address
1082 * component. If strictly routed make sure next
1083 * address is on directly accessible net.
1084 */
1085 case IPOPT_LSRR:
1086 case IPOPT_SSRR:
1087 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1088 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1089 goto bad;
1090 }
1091 ipaddr.sin_addr = ip->ip_dst;
1092 ia = (struct in_ifaddr *)
1093 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1094 if (ia == 0) {
1095 if (opt == IPOPT_SSRR) {
1096 type = ICMP_UNREACH;
1097 code = ICMP_UNREACH_SRCFAIL;
1098 goto bad;
1099 }
1100 if (!ip_dosourceroute)
1101 goto nosourcerouting;
1102 /*
1103 * Loose routing, and not at next destination
1104 * yet; nothing to do except forward.
1105 */
1106 break;
1107 }
1108 off--; /* 0 origin */
1109 if (off > optlen - (int)sizeof(struct in_addr)) {
1110 /*
1111 * End of source route. Should be for us.
1112 */
1113 if (!ip_acceptsourceroute)
1114 goto nosourcerouting;
1115 save_rte(cp, ip->ip_src);
1116 break;
1117 }
1118
1119 if (!ip_dosourceroute) {
1120 if (ipforwarding) {
1121 char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1122 /*
1123 * Acting as a router, so generate ICMP
1124 */
1125 nosourcerouting:
1126 strcpy(buf, inet_ntoa(ip->ip_dst));
1127 log(LOG_WARNING,
1128 "attempted source route from %s to %s\n",
1129 inet_ntoa(ip->ip_src), buf);
1130 type = ICMP_UNREACH;
1131 code = ICMP_UNREACH_SRCFAIL;
1132 goto bad;
1133 } else {
1134 /*
1135 * Not acting as a router, so silently drop.
1136 */
1137 ipstat.ips_cantforward++;
1138 m_freem(m);
1139 return (1);
1140 }
1141 }
1142
1143 /*
1144 * locate outgoing interface
1145 */
1146 (void)memcpy(&ipaddr.sin_addr, cp + off,
1147 sizeof(ipaddr.sin_addr));
1148
1149 if (opt == IPOPT_SSRR) {
1150 #define INA struct in_ifaddr *
1151 #define SA struct sockaddr *
1152 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1153 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1154 } else
1155 ia = ip_rtaddr(ipaddr.sin_addr);
1156 if (ia == 0) {
1157 type = ICMP_UNREACH;
1158 code = ICMP_UNREACH_SRCFAIL;
1159 goto bad;
1160 }
1161 ip->ip_dst = ipaddr.sin_addr;
1162 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1163 sizeof(struct in_addr));
1164 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1165 /*
1166 * Let ip_intr's mcast routing check handle mcast pkts
1167 */
1168 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1169 break;
1170
1171 case IPOPT_RR:
1172 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1173 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1174 goto bad;
1175 }
1176 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1177 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1178 goto bad;
1179 }
1180 /*
1181 * If no space remains, ignore.
1182 */
1183 off--; /* 0 origin */
1184 if (off > optlen - (int)sizeof(struct in_addr))
1185 break;
1186 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1187 sizeof(ipaddr.sin_addr));
1188 /*
1189 * locate outgoing interface; if we're the destination,
1190 * use the incoming interface (should be same).
1191 */
1192 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1193 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1194 type = ICMP_UNREACH;
1195 code = ICMP_UNREACH_HOST;
1196 goto bad;
1197 }
1198 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1199 sizeof(struct in_addr));
1200 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1201 break;
1202
1203 case IPOPT_TS:
1204 code = cp - (u_char *)ip;
1205 ipt = (struct ip_timestamp *)cp;
1206 if (ipt->ipt_len < 5)
1207 goto bad;
1208 if (ipt->ipt_ptr >
1209 ipt->ipt_len - (int)sizeof(int32_t)) {
1210 if (++ipt->ipt_oflw == 0)
1211 goto bad;
1212 break;
1213 }
1214 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
1215 switch (ipt->ipt_flg) {
1216
1217 case IPOPT_TS_TSONLY:
1218 break;
1219
1220 case IPOPT_TS_TSANDADDR:
1221 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1222 sizeof(struct in_addr) > ipt->ipt_len)
1223 goto bad;
1224 ipaddr.sin_addr = dst;
1225 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1226 m->m_pkthdr.rcvif);
1227 if (ia == 0)
1228 continue;
1229 (void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1230 sizeof(struct in_addr));
1231 ipt->ipt_ptr += sizeof(struct in_addr);
1232 break;
1233
1234 case IPOPT_TS_PRESPEC:
1235 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1236 sizeof(struct in_addr) > ipt->ipt_len)
1237 goto bad;
1238 (void)memcpy(&ipaddr.sin_addr, sin,
1239 sizeof(struct in_addr));
1240 if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1241 continue;
1242 ipt->ipt_ptr += sizeof(struct in_addr);
1243 break;
1244
1245 default:
1246 goto bad;
1247 }
1248 ntime = iptime();
1249 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime,
1250 sizeof(n_time));
1251 ipt->ipt_ptr += sizeof(n_time);
1252 }
1253 }
1254 if (forward && ipforwarding) {
1255 ip_forward(m, 1);
1256 return (1);
1257 }
1258 return (0);
1259 bad:
1260 ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2; /* XXX icmp_error adds in hdr length */
1261 icmp_error(m, type, code, 0, 0);
1262 ipstat.ips_badoptions++;
1263 return (1);
1264 }
1265
1266 /*
1267 * Given address of next destination (final or next hop),
1268 * return internet address info of interface to be used to get there.
1269 */
1270 static struct in_ifaddr *
1271 ip_rtaddr(dst)
1272 struct in_addr dst;
1273 {
1274 register struct sockaddr_in *sin;
1275
1276 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
1277
1278 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
1279 if (ipforward_rt.ro_rt) {
1280 RTFREE(ipforward_rt.ro_rt);
1281 ipforward_rt.ro_rt = 0;
1282 }
1283 sin->sin_family = AF_INET;
1284 sin->sin_len = sizeof(*sin);
1285 sin->sin_addr = dst;
1286
1287 rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1288 }
1289 if (ipforward_rt.ro_rt == 0)
1290 return ((struct in_ifaddr *)0);
1291 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
1292 }
1293
1294 /*
1295 * Save incoming source route for use in replies,
1296 * to be picked up later by ip_srcroute if the receiver is interested.
1297 */
1298 void
1299 save_rte(option, dst)
1300 u_char *option;
1301 struct in_addr dst;
1302 {
1303 unsigned olen;
1304
1305 olen = option[IPOPT_OLEN];
1306 #ifdef DIAGNOSTIC
1307 if (ipprintfs)
1308 printf("save_rte: olen %d\n", olen);
1309 #endif
1310 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1311 return;
1312 bcopy(option, ip_srcrt.srcopt, olen);
1313 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1314 ip_srcrt.dst = dst;
1315 }
1316
1317 /*
1318 * Retrieve incoming source route for use in replies,
1319 * in the same form used by setsockopt.
1320 * The first hop is placed before the options, will be removed later.
1321 */
1322 struct mbuf *
1323 ip_srcroute()
1324 {
1325 register struct in_addr *p, *q;
1326 register struct mbuf *m;
1327
1328 if (ip_nhops == 0)
1329 return ((struct mbuf *)0);
1330 m = m_get(M_DONTWAIT, MT_HEADER);
1331 if (m == 0)
1332 return ((struct mbuf *)0);
1333
1334 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1335
1336 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1337 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1338 OPTSIZ;
1339 #ifdef DIAGNOSTIC
1340 if (ipprintfs)
1341 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1342 #endif
1343
1344 /*
1345 * First save first hop for return route
1346 */
1347 p = &ip_srcrt.route[ip_nhops - 1];
1348 *(mtod(m, struct in_addr *)) = *p--;
1349 #ifdef DIAGNOSTIC
1350 if (ipprintfs)
1351 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1352 #endif
1353
1354 /*
1355 * Copy option fields and padding (nop) to mbuf.
1356 */
1357 ip_srcrt.nop = IPOPT_NOP;
1358 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1359 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1360 &ip_srcrt.nop, OPTSIZ);
1361 q = (struct in_addr *)(mtod(m, caddr_t) +
1362 sizeof(struct in_addr) + OPTSIZ);
1363 #undef OPTSIZ
1364 /*
1365 * Record return path as an IP source route,
1366 * reversing the path (pointers are now aligned).
1367 */
1368 while (p >= ip_srcrt.route) {
1369 #ifdef DIAGNOSTIC
1370 if (ipprintfs)
1371 printf(" %lx", (u_long)ntohl(q->s_addr));
1372 #endif
1373 *q++ = *p--;
1374 }
1375 /*
1376 * Last hop goes to final destination.
1377 */
1378 *q = ip_srcrt.dst;
1379 #ifdef DIAGNOSTIC
1380 if (ipprintfs)
1381 printf(" %lx\n", (u_long)ntohl(q->s_addr));
1382 #endif
1383 return (m);
1384 }
1385
1386 /*
1387 * Strip out IP options, at higher
1388 * level protocol in the kernel.
1389 * Second argument is buffer to which options
1390 * will be moved, and return value is their length.
1391 * XXX should be deleted; last arg currently ignored.
1392 */
1393 void
1394 ip_stripoptions(m, mopt)
1395 register struct mbuf *m;
1396 struct mbuf *mopt;
1397 {
1398 register int i;
1399 struct ip *ip = mtod(m, struct ip *);
1400 register caddr_t opts;
1401 int olen;
1402
1403 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1404 opts = (caddr_t)(ip + 1);
1405 i = m->m_len - (sizeof (struct ip) + olen);
1406 bcopy(opts + olen, opts, (unsigned)i);
1407 m->m_len -= olen;
1408 if (m->m_flags & M_PKTHDR)
1409 m->m_pkthdr.len -= olen;
1410 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1411 }
1412
1413 u_char inetctlerrmap[PRC_NCMDS] = {
1414 0, 0, 0, 0,
1415 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1416 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1417 EMSGSIZE, EHOSTUNREACH, 0, 0,
1418 0, 0, 0, 0,
1419 ENOPROTOOPT
1420 };
1421
1422 /*
1423 * Forward a packet. If some error occurs return the sender
1424 * an icmp packet. Note we can't always generate a meaningful
1425 * icmp message because icmp doesn't have a large enough repertoire
1426 * of codes and types.
1427 *
1428 * If not forwarding, just drop the packet. This could be confusing
1429 * if ipforwarding was zero but some routing protocol was advancing
1430 * us as a gateway to somewhere. However, we must let the routing
1431 * protocol deal with that.
1432 *
1433 * The srcrt parameter indicates whether the packet is being forwarded
1434 * via a source route.
1435 */
1436 static void
1437 ip_forward(m, srcrt)
1438 struct mbuf *m;
1439 int srcrt;
1440 {
1441 register struct ip *ip = mtod(m, struct ip *);
1442 register struct sockaddr_in *sin;
1443 register struct rtentry *rt;
1444 int error, type = 0, code = 0;
1445 struct mbuf *mcopy;
1446 n_long dest;
1447 struct ifnet *destifp;
1448
1449 dest = 0;
1450 #ifdef DIAGNOSTIC
1451 if (ipprintfs)
1452 printf("forward: src %lx dst %lx ttl %x\n",
1453 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1454 ip->ip_ttl);
1455 #endif
1456
1457
1458 if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) {
1459 ipstat.ips_cantforward++;
1460 m_freem(m);
1461 return;
1462 }
1463 HTONS(ip->ip_id);
1464 #ifdef IPSTEALTH
1465 if (!ipstealth) {
1466 #endif
1467 if (ip->ip_ttl <= IPTTLDEC) {
1468 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1469 dest, 0);
1470 return;
1471 }
1472 ip->ip_ttl -= IPTTLDEC;
1473 #ifdef IPSTEALTH
1474 }
1475 #endif
1476
1477 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1478 if ((rt = ipforward_rt.ro_rt) == 0 ||
1479 ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
1480 if (ipforward_rt.ro_rt) {
1481 RTFREE(ipforward_rt.ro_rt);
1482 ipforward_rt.ro_rt = 0;
1483 }
1484 sin->sin_family = AF_INET;
1485 sin->sin_len = sizeof(*sin);
1486 sin->sin_addr = ip->ip_dst;
1487
1488 rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1489 if (ipforward_rt.ro_rt == 0) {
1490 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1491 return;
1492 }
1493 rt = ipforward_rt.ro_rt;
1494 }
1495
1496 /*
1497 * Save at most 64 bytes of the packet in case
1498 * we need to generate an ICMP message to the src.
1499 */
1500 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64));
1501
1502 /*
1503 * If forwarding packet using same interface that it came in on,
1504 * perhaps should send a redirect to sender to shortcut a hop.
1505 * Only send redirect if source is sending directly to us,
1506 * and if packet was not source routed (or has any options).
1507 * Also, don't send redirect if forwarding using a default route
1508 * or a route modified by a redirect.
1509 */
1510 #define satosin(sa) ((struct sockaddr_in *)(sa))
1511 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1512 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1513 satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1514 ipsendredirects && !srcrt) {
1515 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1516 u_long src = ntohl(ip->ip_src.s_addr);
1517
1518 if (RTA(rt) &&
1519 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1520 if (rt->rt_flags & RTF_GATEWAY)
1521 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1522 else
1523 dest = ip->ip_dst.s_addr;
1524 /* Router requirements says to only send host redirects */
1525 type = ICMP_REDIRECT;
1526 code = ICMP_REDIRECT_HOST;
1527 #ifdef DIAGNOSTIC
1528 if (ipprintfs)
1529 printf("redirect (%d) to %lx\n", code, (u_long)dest);
1530 #endif
1531 }
1532 }
1533
1534 error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1535 IP_FORWARDING, 0);
1536 if (error)
1537 ipstat.ips_cantforward++;
1538 else {
1539 ipstat.ips_forward++;
1540 if (type)
1541 ipstat.ips_redirectsent++;
1542 else {
1543 if (mcopy) {
1544 ipflow_create(&ipforward_rt, mcopy);
1545 m_freem(mcopy);
1546 }
1547 return;
1548 }
1549 }
1550 if (mcopy == NULL)
1551 return;
1552 destifp = NULL;
1553
1554 switch (error) {
1555
1556 case 0: /* forwarded, but need redirect */
1557 /* type, code set above */
1558 break;
1559
1560 case ENETUNREACH: /* shouldn't happen, checked above */
1561 case EHOSTUNREACH:
1562 case ENETDOWN:
1563 case EHOSTDOWN:
1564 default:
1565 type = ICMP_UNREACH;
1566 code = ICMP_UNREACH_HOST;
1567 break;
1568
1569 case EMSGSIZE:
1570 type = ICMP_UNREACH;
1571 code = ICMP_UNREACH_NEEDFRAG;
1572 if (ipforward_rt.ro_rt)
1573 destifp = ipforward_rt.ro_rt->rt_ifp;
1574 ipstat.ips_cantfrag++;
1575 break;
1576
1577 case ENOBUFS:
1578 type = ICMP_SOURCEQUENCH;
1579 code = 0;
1580 break;
1581
1582 case EACCES: /* ipfw denied packet */
1583 m_freem(mcopy);
1584 return;
1585 }
1586 icmp_error(mcopy, type, code, dest, destifp);
1587 }
1588
1589 void
1590 ip_savecontrol(inp, mp, ip, m)
1591 register struct inpcb *inp;
1592 register struct mbuf **mp;
1593 register struct ip *ip;
1594 register struct mbuf *m;
1595 {
1596 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1597 struct timeval tv;
1598
1599 microtime(&tv);
1600 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1601 SCM_TIMESTAMP, SOL_SOCKET);
1602 if (*mp)
1603 mp = &(*mp)->m_next;
1604 }
1605 if (inp->inp_flags & INP_RECVDSTADDR) {
1606 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1607 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1608 if (*mp)
1609 mp = &(*mp)->m_next;
1610 }
1611 #ifdef notyet
1612 /* XXX
1613 * Moving these out of udp_input() made them even more broken
1614 * than they already were.
1615 */
1616 /* options were tossed already */
1617 if (inp->inp_flags & INP_RECVOPTS) {
1618 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1619 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1620 if (*mp)
1621 mp = &(*mp)->m_next;
1622 }
1623 /* ip_srcroute doesn't do what we want here, need to fix */
1624 if (inp->inp_flags & INP_RECVRETOPTS) {
1625 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1626 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1627 if (*mp)
1628 mp = &(*mp)->m_next;
1629 }
1630 #endif
1631 if (inp->inp_flags & INP_RECVIF) {
1632 struct ifnet *ifp;
1633 struct sdlbuf {
1634 struct sockaddr_dl sdl;
1635 u_char pad[32];
1636 } sdlbuf;
1637 struct sockaddr_dl *sdp;
1638 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1639
1640 if (((ifp = m->m_pkthdr.rcvif))
1641 && ( ifp->if_index && (ifp->if_index <= if_index))) {
1642 sdp = (struct sockaddr_dl *)(ifnet_addrs
1643 [ifp->if_index - 1]->ifa_addr);
1644 /*
1645 * Change our mind and don't try copy.
1646 */
1647 if ((sdp->sdl_family != AF_LINK)
1648 || (sdp->sdl_len > sizeof(sdlbuf))) {
1649 goto makedummy;
1650 }
1651 bcopy(sdp, sdl2, sdp->sdl_len);
1652 } else {
1653 makedummy:
1654 sdl2->sdl_len
1655 = offsetof(struct sockaddr_dl, sdl_data[0]);
1656 sdl2->sdl_family = AF_LINK;
1657 sdl2->sdl_index = 0;
1658 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1659 }
1660 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1661 IP_RECVIF, IPPROTO_IP);
1662 if (*mp)
1663 mp = &(*mp)->m_next;
1664 }
1665 }
1666
1667 int
1668 ip_rsvp_init(struct socket *so)
1669 {
1670 if (so->so_type != SOCK_RAW ||
1671 so->so_proto->pr_protocol != IPPROTO_RSVP)
1672 return EOPNOTSUPP;
1673
1674 if (ip_rsvpd != NULL)
1675 return EADDRINUSE;
1676
1677 ip_rsvpd = so;
1678 /*
1679 * This may seem silly, but we need to be sure we don't over-increment
1680 * the RSVP counter, in case something slips up.
1681 */
1682 if (!ip_rsvp_on) {
1683 ip_rsvp_on = 1;
1684 rsvp_on++;
1685 }
1686
1687 return 0;
1688 }
1689
1690 int
1691 ip_rsvp_done(void)
1692 {
1693 ip_rsvpd = NULL;
1694 /*
1695 * This may seem silly, but we need to be sure we don't over-decrement
1696 * the RSVP counter, in case something slips up.
1697 */
1698 if (ip_rsvp_on) {
1699 ip_rsvp_on = 0;
1700 rsvp_on--;
1701 }
1702 return 0;
1703 }
Cache object: 8f517656d206fc6a1f86d2ef81e275b4
|