1/*
2 * Copyright (c) 1982, 1986, 1988, 1990, 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_output.c 8.3 (Berkeley) 1/21/94
34 * $FreeBSD: head/sys/netinet/ip_output.c 98666 2002-06-23 09:15:43Z luigi $
35 */
36
37#define _IP_VHL
38
39#include "opt_ipfw.h"
40#include "opt_ipdn.h"
41#include "opt_ipdivert.h"
42#include "opt_ipfilter.h"
43#include "opt_ipsec.h"
44#include "opt_pfil_hooks.h"
45#include "opt_random_ip_id.h"
46
47#include <sys/param.h>
48#include <sys/systm.h>
49#include <sys/kernel.h>
50#include <sys/malloc.h>
51#include <sys/mbuf.h>
52#include <sys/protosw.h>
53#include <sys/socket.h>
54#include <sys/socketvar.h>
55
56#include <net/if.h>
57#include <net/route.h>
58
59#include <netinet/in.h>
60#include <netinet/in_systm.h>
61#include <netinet/ip.h>
62#include <netinet/in_pcb.h>
63#include <netinet/in_var.h>
64#include <netinet/ip_var.h>
65
66#include <machine/in_cksum.h>
67
68static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
69
70#ifdef IPSEC
71#include <netinet6/ipsec.h>
72#include <netkey/key.h>
73#ifdef IPSEC_DEBUG
74#include <netkey/key_debug.h>
75#else
76#define KEYDEBUG(lev,arg)
77#endif
78#endif /*IPSEC*/
79
80#include <netinet/ip_fw.h>
81#include <netinet/ip_dummynet.h>
82
83#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\
84 x, (ntohl(a.s_addr)>>24)&0xFF,\
85 (ntohl(a.s_addr)>>16)&0xFF,\
86 (ntohl(a.s_addr)>>8)&0xFF,\
87 (ntohl(a.s_addr))&0xFF, y);
88
89u_short ip_id;
90
91static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
92static struct ifnet *ip_multicast_if(struct in_addr *, int *);
93static void ip_mloopback
94 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
95static int ip_getmoptions
96 (struct sockopt *, struct ip_moptions *);
97static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
98static int ip_setmoptions
99 (struct sockopt *, struct ip_moptions **);
100
101int ip_optcopy(struct ip *, struct ip *);
102
103
104extern struct protosw inetsw[];
105
106/*
107 * IP output. The packet in mbuf chain m contains a skeletal IP
108 * header (with len, off, ttl, proto, tos, src, dst).
109 * The mbuf chain containing the packet will be freed.
110 * The mbuf opt, if present, will not be freed.
111 */
112int
113ip_output(m0, opt, ro, flags, imo)
114 struct mbuf *m0;
115 struct mbuf *opt;
116 struct route *ro;
117 int flags;
118 struct ip_moptions *imo;
119{
120 struct ip *ip, *mhip;
121 struct ifnet *ifp = NULL; /* keep compiler happy */
122 struct mbuf *m;
123 int hlen = sizeof (struct ip);
124 int len, off, error = 0;
125 struct sockaddr_in *dst = NULL; /* keep compiler happy */
126 struct in_ifaddr *ia;
127 int isbroadcast, sw_csum;
128 struct in_addr pkt_dst;
129#ifdef IPSEC
130 struct route iproute;
131 struct socket *so = NULL;
132 struct secpolicy *sp = NULL;
133#endif
134 struct ip_fw_args args;
135 int src_was_INADDR_ANY = 0; /* as the name says... */
136#ifdef PFIL_HOOKS
137 struct packet_filter_hook *pfh;
138 struct mbuf *m1;
139 int rv;
140#endif /* PFIL_HOOKS */
141
142 args.eh = NULL;
143 args.rule = NULL;
144 args.next_hop = NULL;
145 args.divert_rule = 0; /* divert cookie */
146
147 /* Grab info from MT_TAG mbufs prepended to the chain. */
148 for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) {
149 switch(m0->m_tag_id) {
150 default:
151 printf("ip_output: unrecognised MT_TAG tag %d\n",
152 m0->m_tag_id);
153 break;
154
155 case PACKET_TAG_DUMMYNET:
156 /*
157 * the packet was already tagged, so part of the
158 * processing was already done, and we need to go down.
159 * Get parameters from the header.
160 */
161 args.rule = ((struct dn_pkt *)m0)->rule;
162 opt = NULL ;
163 ro = & ( ((struct dn_pkt *)m0)->ro ) ;
164 imo = NULL ;
165 dst = ((struct dn_pkt *)m0)->dn_dst ;
166 ifp = ((struct dn_pkt *)m0)->ifp ;
167 flags = ((struct dn_pkt *)m0)->flags ;
168 break;
169
170 case PACKET_TAG_DIVERT:
171 args.divert_rule = (int)m0->m_data & 0xffff;
172 break;
173
174 case PACKET_TAG_IPFORWARD:
175 args.next_hop = (struct sockaddr_in *)m0->m_data;
176 break;
177 }
178 }
179 m = m0;
180
181 KASSERT(!m || (m->m_flags & M_PKTHDR) != 0, ("ip_output: no HDR"));
182
183 KASSERT(ro != NULL, ("ip_output: no route, proto %d",
184 mtod(m, struct ip *)->ip_p));
185
186#ifdef IPSEC
187 so = ipsec_getsocket(m);
188 (void)ipsec_setsocket(m, NULL);
189#endif
190 if (args.rule != NULL) { /* dummynet already saw us */
191 ip = mtod(m, struct ip *);
192 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
193 ia = ifatoia(ro->ro_rt->rt_ifa);
194 goto sendit;
195 }
196
197 if (opt) {
198 m = ip_insertoptions(m, opt, &len);
199 hlen = len;
200 }
201 ip = mtod(m, struct ip *);
202 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
203
204 /*
205 * Fill in IP header.
206 */
207 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
208 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
209 ip->ip_off &= IP_DF;
210#ifdef RANDOM_IP_ID
211 ip->ip_id = ip_randomid();
212#else
213 ip->ip_id = htons(ip_id++);
214#endif
215 ipstat.ips_localout++;
216 } else {
217 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
218 }
219
220 dst = (struct sockaddr_in *)&ro->ro_dst;
221 /*
222 * If there is a cached route,
223 * check that it is to the same destination
224 * and is still up. If not, free it and try again.
225 * The address family should also be checked in case of sharing the
226 * cache with IPv6.
227 */
228 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
229 dst->sin_family != AF_INET ||
230 dst->sin_addr.s_addr != pkt_dst.s_addr)) {
231 RTFREE(ro->ro_rt);
232 ro->ro_rt = (struct rtentry *)0;
233 }
234 if (ro->ro_rt == 0) {
235 bzero(dst, sizeof(*dst));
236 dst->sin_family = AF_INET;
237 dst->sin_len = sizeof(*dst);
238 dst->sin_addr = pkt_dst;
239 }
240 /*
241 * If routing to interface only,
242 * short circuit routing lookup.
243 */
244 if (flags & IP_ROUTETOIF) {
245 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
246 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
247 ipstat.ips_noroute++;
248 error = ENETUNREACH;
249 goto bad;
250 }
251 ifp = ia->ia_ifp;
252 ip->ip_ttl = 1;
253 isbroadcast = in_broadcast(dst->sin_addr, ifp);
254 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
255 imo != NULL && imo->imo_multicast_ifp != NULL) {
256 /*
257 * Bypass the normal routing lookup for multicast
258 * packets if the interface is specified.
259 */
260 ifp = imo->imo_multicast_ifp;
261 IFP_TO_IA(ifp, ia);
262 isbroadcast = 0; /* fool gcc */
263 } else {
264 /*
265 * If this is the case, we probably don't want to allocate
266 * a protocol-cloned route since we didn't get one from the
267 * ULP. This lets TCP do its thing, while not burdening
268 * forwarding or ICMP with the overhead of cloning a route.
269 * Of course, we still want to do any cloning requested by
270 * the link layer, as this is probably required in all cases
271 * for correct operation (as it is for ARP).
272 */
273 if (ro->ro_rt == 0)
274 rtalloc_ign(ro, RTF_PRCLONING);
275 if (ro->ro_rt == 0) {
276 ipstat.ips_noroute++;
277 error = EHOSTUNREACH;
278 goto bad;
279 }
280 ia = ifatoia(ro->ro_rt->rt_ifa);
281 ifp = ro->ro_rt->rt_ifp;
282 ro->ro_rt->rt_use++;
283 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
284 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
285 if (ro->ro_rt->rt_flags & RTF_HOST)
286 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
287 else
288 isbroadcast = in_broadcast(dst->sin_addr, ifp);
289 }
290 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
291 struct in_multi *inm;
292
293 m->m_flags |= M_MCAST;
294 /*
295 * IP destination address is multicast. Make sure "dst"
296 * still points to the address in "ro". (It may have been
297 * changed to point to a gateway address, above.)
298 */
299 dst = (struct sockaddr_in *)&ro->ro_dst;
300 /*
301 * See if the caller provided any multicast options
302 */
303 if (imo != NULL) {
304 ip->ip_ttl = imo->imo_multicast_ttl;
305 if (imo->imo_multicast_vif != -1)
306 ip->ip_src.s_addr =
307 ip_mcast_src(imo->imo_multicast_vif);
308 } else
309 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
310 /*
311 * Confirm that the outgoing interface supports multicast.
312 */
313 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
314 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
315 ipstat.ips_noroute++;
316 error = ENETUNREACH;
317 goto bad;
318 }
319 }
320 /*
321 * If source address not specified yet, use address
322 * of outgoing interface.
323 */
324 if (ip->ip_src.s_addr == INADDR_ANY) {
325 /* Interface may have no addresses. */
326 if (ia != NULL)
327 ip->ip_src = IA_SIN(ia)->sin_addr;
328 }
329
330 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
331 /*
332 * XXX
333 * delayed checksums are not currently
334 * compatible with IP multicast routing
335 */
336 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
337 in_delayed_cksum(m);
338 m->m_pkthdr.csum_flags &=
339 ~CSUM_DELAY_DATA;
340 }
341 }
342 IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
343 if (inm != NULL &&
344 (imo == NULL || imo->imo_multicast_loop)) {
345 /*
346 * If we belong to the destination multicast group
347 * on the outgoing interface, and the caller did not
348 * forbid loopback, loop back a copy.
349 */
350 ip_mloopback(ifp, m, dst, hlen);
351 }
352 else {
353 /*
354 * If we are acting as a multicast router, perform
355 * multicast forwarding as if the packet had just
356 * arrived on the interface to which we are about
357 * to send. The multicast forwarding function
358 * recursively calls this function, using the
359 * IP_FORWARDING flag to prevent infinite recursion.
360 *
361 * Multicasts that are looped back by ip_mloopback(),
362 * above, will be forwarded by the ip_input() routine,
363 * if necessary.
364 */
365 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
366 /*
367 * Check if rsvp daemon is running. If not, don't
368 * set ip_moptions. This ensures that the packet
369 * is multicast and not just sent down one link
370 * as prescribed by rsvpd.
371 */
372 if (!rsvp_on)
373 imo = NULL;
374 if (ip_mforward(ip, ifp, m, imo) != 0) {
375 m_freem(m);
376 goto done;
377 }
378 }
379 }
380
381 /*
382 * Multicasts with a time-to-live of zero may be looped-
383 * back, above, but must not be transmitted on a network.
384 * Also, multicasts addressed to the loopback interface
385 * are not sent -- the above call to ip_mloopback() will
386 * loop back a copy if this host actually belongs to the
387 * destination group on the loopback interface.
388 */
389 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
390 m_freem(m);
391 goto done;
392 }
393
394 goto sendit;
395 }
396#ifndef notdef
397 /*
398 * If the source address is not specified yet, use the address
399 * of the outoing interface. In case, keep note we did that, so
400 * if the the firewall changes the next-hop causing the output
401 * interface to change, we can fix that.
402 */
403 if (ip->ip_src.s_addr == INADDR_ANY) {
404 /* Interface may have no addresses. */
405 if (ia != NULL) {
406 ip->ip_src = IA_SIN(ia)->sin_addr;
407 src_was_INADDR_ANY = 1;
408 }
409 }
410#endif /* notdef */
411 /*
412 * Verify that we have any chance at all of being able to queue
413 * the packet or packet fragments
414 */
415 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
416 ifp->if_snd.ifq_maxlen) {
417 error = ENOBUFS;
418 ipstat.ips_odropped++;
419 goto bad;
420 }
421
422 /*
423 * Look for broadcast address and
424 * verify user is allowed to send
425 * such a packet.
426 */
427 if (isbroadcast) {
428 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
429 error = EADDRNOTAVAIL;
430 goto bad;
431 }
432 if ((flags & IP_ALLOWBROADCAST) == 0) {
433 error = EACCES;
434 goto bad;
435 }
436 /* don't allow broadcast messages to be fragmented */
437 if ((u_short)ip->ip_len > ifp->if_mtu) {
438 error = EMSGSIZE;
439 goto bad;
440 }
441 m->m_flags |= M_BCAST;
442 } else {
443 m->m_flags &= ~M_BCAST;
444 }
445
446sendit:
447#ifdef IPSEC
448 /* get SP for this packet */
449 if (so == NULL)
450 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
451 else
452 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
453
454 if (sp == NULL) {
455 ipsecstat.out_inval++;
456 goto bad;
457 }
458
459 error = 0;
460
461 /* check policy */
462 switch (sp->policy) {
463 case IPSEC_POLICY_DISCARD:
464 /*
465 * This packet is just discarded.
466 */
467 ipsecstat.out_polvio++;
468 goto bad;
469
470 case IPSEC_POLICY_BYPASS:
471 case IPSEC_POLICY_NONE:
472 /* no need to do IPsec. */
473 goto skip_ipsec;
474
475 case IPSEC_POLICY_IPSEC:
476 if (sp->req == NULL) {
477 /* acquire a policy */
478 error = key_spdacquire(sp);
479 goto bad;
480 }
481 break;
482
483 case IPSEC_POLICY_ENTRUST:
484 default:
485 printf("ip_output: Invalid policy found. %d\n", sp->policy);
486 }
487 {
488 struct ipsec_output_state state;
489 bzero(&state, sizeof(state));
490 state.m = m;
491 if (flags & IP_ROUTETOIF) {
492 state.ro = &iproute;
493 bzero(&iproute, sizeof(iproute));
494 } else
495 state.ro = ro;
496 state.dst = (struct sockaddr *)dst;
497
498 ip->ip_sum = 0;
499
500 /*
501 * XXX
502 * delayed checksums are not currently compatible with IPsec
503 */
504 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
505 in_delayed_cksum(m);
506 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
507 }
508
509 ip->ip_len = htons(ip->ip_len);
510 ip->ip_off = htons(ip->ip_off);
511
512 error = ipsec4_output(&state, sp, flags);
513
514 m = state.m;
515 if (flags & IP_ROUTETOIF) {
516 /*
517 * if we have tunnel mode SA, we may need to ignore
518 * IP_ROUTETOIF.
519 */
520 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
521 flags &= ~IP_ROUTETOIF;
522 ro = state.ro;
523 }
524 } else
525 ro = state.ro;
526 dst = (struct sockaddr_in *)state.dst;
527 if (error) {
528 /* mbuf is already reclaimed in ipsec4_output. */
529 m0 = NULL;
530 switch (error) {
531 case EHOSTUNREACH:
532 case ENETUNREACH:
533 case EMSGSIZE:
534 case ENOBUFS:
535 case ENOMEM:
536 break;
537 default:
538 printf("ip4_output (ipsec): error code %d\n", error);
539 /*fall through*/
540 case ENOENT:
541 /* don't show these error codes to the user */
542 error = 0;
543 break;
544 }
545 goto bad;
546 }
547 }
548
549 /* be sure to update variables that are affected by ipsec4_output() */
550 ip = mtod(m, struct ip *);
551#ifdef _IP_VHL
552 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
553#else
554 hlen = ip->ip_hl << 2;
555#endif
556 if (ro->ro_rt == NULL) {
557 if ((flags & IP_ROUTETOIF) == 0) {
558 printf("ip_output: "
559 "can't update route after IPsec processing\n");
560 error = EHOSTUNREACH; /*XXX*/
561 goto bad;
562 }
563 } else {
564 ia = ifatoia(ro->ro_rt->rt_ifa);
565 ifp = ro->ro_rt->rt_ifp;
566 }
567
568 /* make it flipped, again. */
569 ip->ip_len = ntohs(ip->ip_len);
570 ip->ip_off = ntohs(ip->ip_off);
571skip_ipsec:
572#endif /*IPSEC*/
573
574 /*
575 * IpHack's section.
576 * - Xlate: translate packet's addr/port (NAT).
577 * - Firewall: deny/allow/etc.
578 * - Wrap: fake packet's addr/port <unimpl.>
579 * - Encapsulate: put it in another IP and send out. <unimp.>
580 */
581#ifdef PFIL_HOOKS
582 /*
583 * Run through list of hooks for output packets.
584 */
585 m1 = m;
586 pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
587 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
588 if (pfh->pfil_func) {
589 rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1);
590 if (rv) {
591 error = EHOSTUNREACH;
592 goto done;
593 }
594 m = m1;
595 if (m == NULL)
596 goto done;
597 ip = mtod(m, struct ip *);
598 }
599#endif /* PFIL_HOOKS */
600
601 /*
602 * Check with the firewall...
603 * but not if we are already being fwd'd from a firewall.
604 */
605 if (fw_enable && IPFW_LOADED && !args.next_hop) {
606 struct sockaddr_in *old = dst;
607
608 args.m = m;
609 args.next_hop = dst;
610 args.oif = ifp;
611 off = ip_fw_chk_ptr(&args);
612 m = args.m;
613 dst = args.next_hop;
614
615 /*
616 * On return we must do the following:
617 * m == NULL -> drop the pkt (old interface, deprecated)
618 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface)
619 * 1<=off<= 0xffff -> DIVERT
620 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe
621 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet
622 * dst != old -> IPFIREWALL_FORWARD
623 * off==0, dst==old -> accept
624 * If some of the above modules are not compiled in, then
625 * we should't have to check the corresponding condition
626 * (because the ipfw control socket should not accept
627 * unsupported rules), but better play safe and drop
628 * packets in case of doubt.
629 */
630 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
631 if (m)
632 m_freem(m);
633 error = EACCES;
634 goto done;
635 }
636 ip = mtod(m, struct ip *);
637 if (off == 0 && dst == old) /* common case */
638 goto pass;
639 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) {
640 /*
641 * pass the pkt to dummynet. Need to include
642 * pipe number, m, ifp, ro, dst because these are
643 * not recomputed in the next pass.
644 * All other parameters have been already used and
645 * so they are not needed anymore.
646 * XXX note: if the ifp or ro entry are deleted
647 * while a pkt is in dummynet, we are in trouble!
648 */
649 args.ro = ro;
650 args.dst = dst;
651 args.flags = flags;
652
653 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT,
654 &args);
655 goto done;
656 }
657#ifdef IPDIVERT
658 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) {
659 struct mbuf *clone = NULL;
660
661 /* Clone packet if we're doing a 'tee' */
662 if ((off & IP_FW_PORT_TEE_FLAG) != 0)
663 clone = m_dup(m, M_DONTWAIT);
664
665 /*
666 * XXX
667 * delayed checksums are not currently compatible
668 * with divert sockets.
669 */
670 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
671 in_delayed_cksum(m);
672 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
673 }
674
675 /* Restore packet header fields to original values */
676 ip->ip_len = htons(ip->ip_len);
677 ip->ip_off = htons(ip->ip_off);
678
679 /* Deliver packet to divert input routine */
680 divert_packet(m, 0, off & 0xffff, args.divert_rule);
681
682 /* If 'tee', continue with original packet */
683 if (clone != NULL) {
684 m = clone;
685 ip = mtod(m, struct ip *);
686 goto pass;
687 }
688 goto done;
689 }
690#endif
691
692 /* IPFIREWALL_FORWARD */
693 /*
694 * Check dst to make sure it is directly reachable on the
695 * interface we previously thought it was.
696 * If it isn't (which may be likely in some situations) we have
697 * to re-route it (ie, find a route for the next-hop and the
698 * associated interface) and set them here. This is nested
699 * forwarding which in most cases is undesirable, except where
700 * such control is nigh impossible. So we do it here.
701 * And I'm babbling.
702 */
703 if (off == 0 && old != dst) { /* FORWARD, dst has changed */
704#if 0
705 /*
706 * XXX To improve readability, this block should be
707 * changed into a function call as below:
708 */
709 error = ip_ipforward(&m, &dst, &ifp);
710 if (error)
711 goto bad;
712 if (m == NULL) /* ip_input consumed the mbuf */
713 goto done;
714#else
715 struct in_ifaddr *ia;
716
717 /*
718 * XXX sro_fwd below is static, and a pointer
719 * to it gets passed to routines downstream.
720 * This could have surprisingly bad results in
721 * practice, because its content is overwritten
722 * by subsequent packets.
723 */
724 /* There must be a better way to do this next line... */
725 static struct route sro_fwd;
726 struct route *ro_fwd = &sro_fwd;
727
728#if 0
729 print_ip("IPFIREWALL_FORWARD: New dst ip: ",
730 dst->sin_addr, "\n");
731#endif
732
733 /*
734 * We need to figure out if we have been forwarded
735 * to a local socket. If so, then we should somehow
736 * "loop back" to ip_input, and get directed to the
737 * PCB as if we had received this packet. This is
738 * because it may be dificult to identify the packets
739 * you want to forward until they are being output
740 * and have selected an interface. (e.g. locally
741 * initiated packets) If we used the loopback inteface,
742 * we would not be able to control what happens
743 * as the packet runs through ip_input() as
744 * it is done through a ISR.
745 */
746 LIST_FOREACH(ia,
747 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) {
748 /*
749 * If the addr to forward to is one
750 * of ours, we pretend to
751 * be the destination for this packet.
752 */
753 if (IA_SIN(ia)->sin_addr.s_addr ==
754 dst->sin_addr.s_addr)
755 break;
756 }
757 if (ia) { /* tell ip_input "dont filter" */
758 struct m_hdr tag;
759
760 tag.mh_type = MT_TAG;
761 tag.mh_flags = PACKET_TAG_IPFORWARD;
762 tag.mh_data = (caddr_t)args.next_hop;
763 tag.mh_next = m;
764
765 if (m->m_pkthdr.rcvif == NULL)
766 m->m_pkthdr.rcvif = ifunit("lo0");
767 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
768 m->m_pkthdr.csum_flags |=
769 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
770 m0->m_pkthdr.csum_data = 0xffff;
771 }
772 m->m_pkthdr.csum_flags |=
773 CSUM_IP_CHECKED | CSUM_IP_VALID;
774 ip->ip_len = htons(ip->ip_len);
775 ip->ip_off = htons(ip->ip_off);
776 ip_input((struct mbuf *)&tag);
777 goto done;
778 }
779 /* Some of the logic for this was
780 * nicked from above.
781 *
782 * This rewrites the cached route in a local PCB.
783 * Is this what we want to do?
784 */
785 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst));
786
787 ro_fwd->ro_rt = 0;
788 rtalloc_ign(ro_fwd, RTF_PRCLONING);
789
790 if (ro_fwd->ro_rt == 0) {
791 ipstat.ips_noroute++;
792 error = EHOSTUNREACH;
793 goto bad;
794 }
795
796 ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
797 ifp = ro_fwd->ro_rt->rt_ifp;
798 ro_fwd->ro_rt->rt_use++;
799 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
800 dst = (struct sockaddr_in *)
801 ro_fwd->ro_rt->rt_gateway;
802 if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
803 isbroadcast =
804 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
805 else
806 isbroadcast = in_broadcast(dst->sin_addr, ifp);
807 if (ro->ro_rt)
808 RTFREE(ro->ro_rt);
809 ro->ro_rt = ro_fwd->ro_rt;
810 dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
811
812#endif /* ... block to be put into a function */
813 /*
814 * If we added a default src ip earlier,
815 * which would have been gotten from the-then
816 * interface, do it again, from the new one.
817 */
818 if (src_was_INADDR_ANY)
819 ip->ip_src = IA_SIN(ia)->sin_addr;
820 goto pass ;
821 }
822
823 /*
824 * if we get here, none of the above matches, and
825 * we have to drop the pkt
826 */
827 m_freem(m);
828 error = EACCES; /* not sure this is the right error msg */
829 goto done;
830 }
831
832pass:
833 /* 127/8 must not appear on wire - RFC1122. */
834 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
835 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
836 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
837 ipstat.ips_badaddr++;
838 error = EADDRNOTAVAIL;
839 goto bad;
840 }
841 }
842
843 m->m_pkthdr.csum_flags |= CSUM_IP;
844 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
845 if (sw_csum & CSUM_DELAY_DATA) {
846 in_delayed_cksum(m);
847 sw_csum &= ~CSUM_DELAY_DATA;
848 }
849 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
850
851 /*
852 * If small enough for interface, or the interface will take
853 * care of the fragmentation for us, can just send directly.
854 */
855 if ((u_short)ip->ip_len <= ifp->if_mtu ||
856 ifp->if_hwassist & CSUM_FRAGMENT) {
857 ip->ip_len = htons(ip->ip_len);
858 ip->ip_off = htons(ip->ip_off);
859 ip->ip_sum = 0;
860 if (sw_csum & CSUM_DELAY_IP) {
861 if (ip->ip_vhl == IP_VHL_BORING) {
862 ip->ip_sum = in_cksum_hdr(ip);
863 } else {
864 ip->ip_sum = in_cksum(m, hlen);
865 }
866 }
867
868 /* Record statistics for this interface address. */
869 if (!(flags & IP_FORWARDING) && ia) {
870 ia->ia_ifa.if_opackets++;
871 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
872 }
873
874#ifdef IPSEC
875 /* clean ipsec history once it goes out of the node */
876 ipsec_delaux(m);
877#endif
878
879 error = (*ifp->if_output)(ifp, m,
880 (struct sockaddr *)dst, ro->ro_rt);
881 goto done;
882 }
883 /*
884 * Too large for interface; fragment if possible.
885 * Must be able to put at least 8 bytes per fragment.
886 */
887 if (ip->ip_off & IP_DF) {
888 error = EMSGSIZE;
889 /*
890 * This case can happen if the user changed the MTU
891 * of an interface after enabling IP on it. Because
892 * most netifs don't keep track of routes pointing to
893 * them, there is no way for one to update all its
894 * routes when the MTU is changed.
895 */
896 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST))
897 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU)
898 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
899 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
900 }
901 ipstat.ips_cantfrag++;
902 goto bad;
903 }
904 len = (ifp->if_mtu - hlen) &~ 7;
905 if (len < 8) {
906 error = EMSGSIZE;
907 goto bad;
908 }
909
910 /*
911 * if the interface will not calculate checksums on
912 * fragmented packets, then do it here.
913 */
914 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
915 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) {
916 in_delayed_cksum(m);
917 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
918 }
919
920 {
921 int mhlen, firstlen = len;
922 struct mbuf **mnext = &m->m_nextpkt;
923 int nfrags = 1;
924
925 /*
926 * Loop through length of segment after first fragment,
927 * make new header and copy data of each part and link onto chain.
928 */
929 m0 = m;
930 mhlen = sizeof (struct ip);
931 for (off = hlen + len; off < (u_short)ip->ip_len; off += len) {
932 MGETHDR(m, M_DONTWAIT, MT_HEADER);
933 if (m == 0) {
934 error = ENOBUFS;
935 ipstat.ips_odropped++;
936 goto sendorfree;
937 }
938 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
939 m->m_data += max_linkhdr;
940 mhip = mtod(m, struct ip *);
941 *mhip = *ip;
942 if (hlen > sizeof (struct ip)) {
943 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
944 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
945 }
946 m->m_len = mhlen;
947 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
948 if (off + len >= (u_short)ip->ip_len)
949 len = (u_short)ip->ip_len - off;
950 else
951 mhip->ip_off |= IP_MF;
952 mhip->ip_len = htons((u_short)(len + mhlen));
953 m->m_next = m_copy(m0, off, len);
954 if (m->m_next == 0) {
955 (void) m_free(m);
956 error = ENOBUFS; /* ??? */
957 ipstat.ips_odropped++;
958 goto sendorfree;
959 }
960 m->m_pkthdr.len = mhlen + len;
961 m->m_pkthdr.rcvif = (struct ifnet *)0;
962 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
963 mhip->ip_off = htons(mhip->ip_off);
964 mhip->ip_sum = 0;
965 if (sw_csum & CSUM_DELAY_IP) {
966 if (mhip->ip_vhl == IP_VHL_BORING) {
967 mhip->ip_sum = in_cksum_hdr(mhip);
968 } else {
969 mhip->ip_sum = in_cksum(m, mhlen);
970 }
971 }
972 *mnext = m;
973 mnext = &m->m_nextpkt;
974 nfrags++;
975 }
976 ipstat.ips_ofragments += nfrags;
977
978 /* set first/last markers for fragment chain */
979 m->m_flags |= M_LASTFRAG;
980 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
981 m0->m_pkthdr.csum_data = nfrags;
982
983 /*
984 * Update first fragment by trimming what's been copied out
985 * and updating header, then send each fragment (in order).
986 */
987 m = m0;
988 m_adj(m, hlen + firstlen - (u_short)ip->ip_len);
989 m->m_pkthdr.len = hlen + firstlen;
990 ip->ip_len = htons((u_short)m->m_pkthdr.len);
991 ip->ip_off |= IP_MF;
992 ip->ip_off = htons(ip->ip_off);
993 ip->ip_sum = 0;
994 if (sw_csum & CSUM_DELAY_IP) {
995 if (ip->ip_vhl == IP_VHL_BORING) {
996 ip->ip_sum = in_cksum_hdr(ip);
997 } else {
998 ip->ip_sum = in_cksum(m, hlen);
999 }
1000 }
1001sendorfree:
1002 for (m = m0; m; m = m0) {
1003 m0 = m->m_nextpkt;
1004 m->m_nextpkt = 0;
1005#ifdef IPSEC
1006 /* clean ipsec history once it goes out of the node */
1007 ipsec_delaux(m);
1008#endif
1009 if (error == 0) {
1010 /* Record statistics for this interface address. */
1011 if (ia != NULL) {
1012 ia->ia_ifa.if_opackets++;
1013 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1014 }
1015
1016 error = (*ifp->if_output)(ifp, m,
1017 (struct sockaddr *)dst, ro->ro_rt);
1018 } else
1019 m_freem(m);
1020 }
1021
1022 if (error == 0)
1023 ipstat.ips_fragmented++;
1024 }
1025done:
1026#ifdef IPSEC
1027 if (ro == &iproute && ro->ro_rt) {
1028 RTFREE(ro->ro_rt);
1029 ro->ro_rt = NULL;
1030 }
1031 if (sp != NULL) {
1032 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1033 printf("DP ip_output call free SP:%p\n", sp));
1034 key_freesp(sp);
1035 }
1036#endif /* IPSEC */
1037 return (error);
1038bad:
1039 m_freem(m);
1040 goto done;
1041}
1042
1043void
1044in_delayed_cksum(struct mbuf *m)
1045{
1046 struct ip *ip;
1047 u_short csum, offset;
1048
1049 ip = mtod(m, struct ip *);
1050 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1051 csum = in_cksum_skip(m, ip->ip_len, offset);
1052 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1053 csum = 0xffff;
1054 offset += m->m_pkthdr.csum_data; /* checksum offset */
1055
1056 if (offset + sizeof(u_short) > m->m_len) {
1057 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1058 m->m_len, offset, ip->ip_p);
1059 /*
1060 * XXX
1061 * this shouldn't happen, but if it does, the
1062 * correct behavior may be to insert the checksum
1063 * in the existing chain instead of rearranging it.
1064 */
1065 m = m_pullup(m, offset + sizeof(u_short));
1066 }
1067 *(u_short *)(m->m_data + offset) = csum;
1068}
1069
1070/*
1071 * Insert IP options into preformed packet.
1072 * Adjust IP destination as required for IP source routing,
1073 * as indicated by a non-zero in_addr at the start of the options.
1074 *
1075 * XXX This routine assumes that the packet has no options in place.
1076 */
1077static struct mbuf *
1078ip_insertoptions(m, opt, phlen)
1079 register struct mbuf *m;
1080 struct mbuf *opt;
1081 int *phlen;
1082{
1083 register struct ipoption *p = mtod(opt, struct ipoption *);
1084 struct mbuf *n;
1085 register struct ip *ip = mtod(m, struct ip *);
1086 unsigned optlen;
1087
1088 optlen = opt->m_len - sizeof(p->ipopt_dst);
1089 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET)
1090 return (m); /* XXX should fail */
1091 if (p->ipopt_dst.s_addr)
1092 ip->ip_dst = p->ipopt_dst;
1093 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1094 MGETHDR(n, M_DONTWAIT, MT_HEADER);
1095 if (n == 0)
1096 return (m);
1097 n->m_pkthdr.rcvif = (struct ifnet *)0;
1098 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1099 m->m_len -= sizeof(struct ip);
1100 m->m_data += sizeof(struct ip);
1101 n->m_next = m;
1102 m = n;
1103 m->m_len = optlen + sizeof(struct ip);
1104 m->m_data += max_linkhdr;
1105 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip));
1106 } else {
1107 m->m_data -= optlen;
1108 m->m_len += optlen;
1109 m->m_pkthdr.len += optlen;
1110 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1111 }
1112 ip = mtod(m, struct ip *);
1113 bcopy(p->ipopt_list, ip + 1, optlen);
1114 *phlen = sizeof(struct ip) + optlen;
1115 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1116 ip->ip_len += optlen;
1117 return (m);
1118}
1119
1120/*
1121 * Copy options from ip to jp,
1122 * omitting those not copied during fragmentation.
1123 */
1124int
1125ip_optcopy(ip, jp)
1126 struct ip *ip, *jp;
1127{
1128 register u_char *cp, *dp;
1129 int opt, optlen, cnt;
1130
1131 cp = (u_char *)(ip + 1);
1132 dp = (u_char *)(jp + 1);
1133 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1134 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1135 opt = cp[0];
1136 if (opt == IPOPT_EOL)
1137 break;
1138 if (opt == IPOPT_NOP) {
1139 /* Preserve for IP mcast tunnel's LSRR alignment. */
1140 *dp++ = IPOPT_NOP;
1141 optlen = 1;
1142 continue;
1143 }
1144
1145 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp),
1146 ("ip_optcopy: malformed ipv4 option"));
1147 optlen = cp[IPOPT_OLEN];
1148 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt,
1149 ("ip_optcopy: malformed ipv4 option"));
1150
1151 /* bogus lengths should have been caught by ip_dooptions */
1152 if (optlen > cnt)
1153 optlen = cnt;
1154 if (IPOPT_COPIED(opt)) {
1155 bcopy(cp, dp, optlen);
1156 dp += optlen;
1157 }
1158 }
1159 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1160 *dp++ = IPOPT_EOL;
1161 return (optlen);
1162}
1163
1164/*
1165 * IP socket option processing.
1166 */
1167int
1168ip_ctloutput(so, sopt)
1169 struct socket *so;
1170 struct sockopt *sopt;
1171{
1172 struct inpcb *inp = sotoinpcb(so);
1173 int error, optval;
1174
1175 error = optval = 0;
1176 if (sopt->sopt_level != IPPROTO_IP) {
1177 return (EINVAL);
1178 }
1179
1180 switch (sopt->sopt_dir) {
1181 case SOPT_SET:
1182 switch (sopt->sopt_name) {
1183 case IP_OPTIONS:
1184#ifdef notyet
1185 case IP_RETOPTS:
1186#endif
1187 {
1188 struct mbuf *m;
1189 if (sopt->sopt_valsize > MLEN) {
1190 error = EMSGSIZE;
1191 break;
1192 }
1193 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER);
1194 if (m == 0) {
1195 error = ENOBUFS;
1196 break;
1197 }
1198 m->m_len = sopt->sopt_valsize;
1199 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1200 m->m_len);
1201
1202 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1203 m));
1204 }
1205
1206 case IP_TOS:
1207 case IP_TTL:
1208 case IP_RECVOPTS:
1209 case IP_RECVRETOPTS:
1210 case IP_RECVDSTADDR:
1211 case IP_RECVIF:
1212 case IP_FAITH:
1213 error = sooptcopyin(sopt, &optval, sizeof optval,
1214 sizeof optval);
1215 if (error)
1216 break;
1217
1218 switch (sopt->sopt_name) {
1219 case IP_TOS:
1220 inp->inp_ip_tos = optval;
1221 break;
1222
1223 case IP_TTL:
1224 inp->inp_ip_ttl = optval;
1225 break;
1226#define OPTSET(bit) \
1227 if (optval) \
1228 inp->inp_flags |= bit; \
1229 else \
1230 inp->inp_flags &= ~bit;
1231
1232 case IP_RECVOPTS:
1233 OPTSET(INP_RECVOPTS);
1234 break;
1235
1236 case IP_RECVRETOPTS:
1237 OPTSET(INP_RECVRETOPTS);
1238 break;
1239
1240 case IP_RECVDSTADDR:
1241 OPTSET(INP_RECVDSTADDR);
1242 break;
1243
1244 case IP_RECVIF:
1245 OPTSET(INP_RECVIF);
1246 break;
1247
1248 case IP_FAITH:
1249 OPTSET(INP_FAITH);
1250 break;
1251 }
1252 break;
1253#undef OPTSET
1254
1255 case IP_MULTICAST_IF:
1256 case IP_MULTICAST_VIF:
1257 case IP_MULTICAST_TTL:
1258 case IP_MULTICAST_LOOP:
1259 case IP_ADD_MEMBERSHIP:
1260 case IP_DROP_MEMBERSHIP:
1261 error = ip_setmoptions(sopt, &inp->inp_moptions);
1262 break;
1263
1264 case IP_PORTRANGE:
1265 error = sooptcopyin(sopt, &optval, sizeof optval,
1266 sizeof optval);
1267 if (error)
1268 break;
1269
1270 switch (optval) {
1271 case IP_PORTRANGE_DEFAULT:
1272 inp->inp_flags &= ~(INP_LOWPORT);
1273 inp->inp_flags &= ~(INP_HIGHPORT);
1274 break;
1275
1276 case IP_PORTRANGE_HIGH:
1277 inp->inp_flags &= ~(INP_LOWPORT);
1278 inp->inp_flags |= INP_HIGHPORT;
1279 break;
1280
1281 case IP_PORTRANGE_LOW:
1282 inp->inp_flags &= ~(INP_HIGHPORT);
1283 inp->inp_flags |= INP_LOWPORT;
1284 break;
1285
1286 default:
1287 error = EINVAL;
1288 break;
1289 }
1290 break;
1291
1292#ifdef IPSEC
1293 case IP_IPSEC_POLICY:
1294 {
1295 caddr_t req;
1296 size_t len = 0;
1297 int priv;
1298 struct mbuf *m;
1299 int optname;
1300
1301 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1302 break;
1303 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1304 break;
1305 priv = (sopt->sopt_td != NULL &&
1306 suser(sopt->sopt_td) != 0) ? 0 : 1;
1307 req = mtod(m, caddr_t);
1308 len = m->m_len;
1309 optname = sopt->sopt_name;
1310 error = ipsec4_set_policy(inp, optname, req, len, priv);
1311 m_freem(m);
1312 break;
1313 }
1314#endif /*IPSEC*/
1315
1316 default:
1317 error = ENOPROTOOPT;
1318 break;
1319 }
1320 break;
1321
1322 case SOPT_GET:
1323 switch (sopt->sopt_name) {
1324 case IP_OPTIONS:
1325 case IP_RETOPTS:
1326 if (inp->inp_options)
1327 error = sooptcopyout(sopt,
1328 mtod(inp->inp_options,
1329 char *),
1330 inp->inp_options->m_len);
1331 else
1332 sopt->sopt_valsize = 0;
1333 break;
1334
1335 case IP_TOS:
1336 case IP_TTL:
1337 case IP_RECVOPTS:
1338 case IP_RECVRETOPTS:
1339 case IP_RECVDSTADDR:
1340 case IP_RECVIF:
1341 case IP_PORTRANGE:
1342 case IP_FAITH:
1343 switch (sopt->sopt_name) {
1344
1345 case IP_TOS:
1346 optval = inp->inp_ip_tos;
1347 break;
1348
1349 case IP_TTL:
1350 optval = inp->inp_ip_ttl;
1351 break;
1352
1353#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1354
1355 case IP_RECVOPTS:
1356 optval = OPTBIT(INP_RECVOPTS);
1357 break;
1358
1359 case IP_RECVRETOPTS:
1360 optval = OPTBIT(INP_RECVRETOPTS);
1361 break;
1362
1363 case IP_RECVDSTADDR:
1364 optval = OPTBIT(INP_RECVDSTADDR);
1365 break;
1366
1367 case IP_RECVIF:
1368 optval = OPTBIT(INP_RECVIF);
1369 break;
1370
1371 case IP_PORTRANGE:
1372 if (inp->inp_flags & INP_HIGHPORT)
1373 optval = IP_PORTRANGE_HIGH;
1374 else if (inp->inp_flags & INP_LOWPORT)
1375 optval = IP_PORTRANGE_LOW;
1376 else
1377 optval = 0;
1378 break;
1379
1380 case IP_FAITH:
1381 optval = OPTBIT(INP_FAITH);
1382 break;
1383 }
1384 error = sooptcopyout(sopt, &optval, sizeof optval);
1385 break;
1386
1387 case IP_MULTICAST_IF:
1388 case IP_MULTICAST_VIF:
1389 case IP_MULTICAST_TTL:
1390 case IP_MULTICAST_LOOP:
1391 case IP_ADD_MEMBERSHIP:
1392 case IP_DROP_MEMBERSHIP:
1393 error = ip_getmoptions(sopt, inp->inp_moptions);
1394 break;
1395
1396#ifdef IPSEC
1397 case IP_IPSEC_POLICY:
1398 {
1399 struct mbuf *m = NULL;
1400 caddr_t req = NULL;
1401 size_t len = 0;
1402
1403 if (m != 0) {
1404 req = mtod(m, caddr_t);
1405 len = m->m_len;
1406 }
1407 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m);
1408 if (error == 0)
1409 error = soopt_mcopyout(sopt, m); /* XXX */
1410 if (error == 0)
1411 m_freem(m);
1412 break;
1413 }
1414#endif /*IPSEC*/
1415
1416 default:
1417 error = ENOPROTOOPT;
1418 break;
1419 }
1420 break;
1421 }
1422 return (error);
1423}
1424
1425/*
1426 * Set up IP options in pcb for insertion in output packets.
1427 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1428 * with destination address if source routed.
1429 */
1430static int
1431ip_pcbopts(optname, pcbopt, m)
1432 int optname;
1433 struct mbuf **pcbopt;
1434 register struct mbuf *m;
1435{
1436 register int cnt, optlen;
1437 register u_char *cp;
1438 u_char opt;
1439
1440 /* turn off any old options */
1441 if (*pcbopt)
1442 (void)m_free(*pcbopt);
1443 *pcbopt = 0;
1444 if (m == (struct mbuf *)0 || m->m_len == 0) {
1445 /*
1446 * Only turning off any previous options.
1447 */
1448 if (m)
1449 (void)m_free(m);
1450 return (0);
1451 }
1452
1453 if (m->m_len % sizeof(int32_t))
1454 goto bad;
1455 /*
1456 * IP first-hop destination address will be stored before
1457 * actual options; move other options back
1458 * and clear it when none present.
1459 */
1460 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1461 goto bad;
1462 cnt = m->m_len;
1463 m->m_len += sizeof(struct in_addr);
1464 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1465 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
1466 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1467
1468 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1469 opt = cp[IPOPT_OPTVAL];
1470 if (opt == IPOPT_EOL)
1471 break;
1472 if (opt == IPOPT_NOP)
1473 optlen = 1;
1474 else {
1475 if (cnt < IPOPT_OLEN + sizeof(*cp))
1476 goto bad;
1477 optlen = cp[IPOPT_OLEN];
1478 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1479 goto bad;
1480 }
1481 switch (opt) {
1482
1483 default:
1484 break;
1485
1486 case IPOPT_LSRR:
1487 case IPOPT_SSRR:
1488 /*
1489 * user process specifies route as:
1490 * ->A->B->C->D
1491 * D must be our final destination (but we can't
1492 * check that since we may not have connected yet).
1493 * A is first hop destination, which doesn't appear in
1494 * actual IP option, but is stored before the options.
1495 */
1496 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1497 goto bad;
1498 m->m_len -= sizeof(struct in_addr);
1499 cnt -= sizeof(struct in_addr);
1500 optlen -= sizeof(struct in_addr);
1501 cp[IPOPT_OLEN] = optlen;
1502 /*
1503 * Move first hop before start of options.
1504 */
1505 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1506 sizeof(struct in_addr));
1507 /*
1508 * Then copy rest of options back
1509 * to close up the deleted entry.
1510 */
1511 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1512 sizeof(struct in_addr)),
1513 (caddr_t)&cp[IPOPT_OFFSET+1],
1514 (unsigned)cnt + sizeof(struct in_addr));
1515 break;
1516 }
1517 }
1518 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1519 goto bad;
1520 *pcbopt = m;
1521 return (0);
1522
1523bad:
1524 (void)m_free(m);
1525 return (EINVAL);
1526}
1527
1528/*
1529 * XXX
1530 * The whole multicast option thing needs to be re-thought.
1531 * Several of these options are equally applicable to non-multicast
1532 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1533 * standard option (IP_TTL).
1534 */
1535
1536/*
1537 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1538 */
1539static struct ifnet *
1540ip_multicast_if(a, ifindexp)
1541 struct in_addr *a;
1542 int *ifindexp;
1543{
1544 int ifindex;
1545 struct ifnet *ifp;
1546
1547 if (ifindexp)
1548 *ifindexp = 0;
1549 if (ntohl(a->s_addr) >> 24 == 0) {
1550 ifindex = ntohl(a->s_addr) & 0xffffff;
1551 if (ifindex < 0 || if_index < ifindex)
1552 return NULL;
1553 ifp = ifnet_byindex(ifindex);
1554 if (ifindexp)
1555 *ifindexp = ifindex;
1556 } else {
1557 INADDR_TO_IFP(*a, ifp);
1558 }
1559 return ifp;
1560}
1561
1562/*
1563 * Set the IP multicast options in response to user setsockopt().
1564 */
1565static int
1566ip_setmoptions(sopt, imop)
1567 struct sockopt *sopt;
1568 struct ip_moptions **imop;
1569{
1570 int error = 0;
1571 int i;
1572 struct in_addr addr;
1573 struct ip_mreq mreq;
1574 struct ifnet *ifp;
1575 struct ip_moptions *imo = *imop;
1576 struct route ro;
1577 struct sockaddr_in *dst;
1578 int ifindex;
1579 int s;
1580
1581 if (imo == NULL) {
1582 /*
1583 * No multicast option buffer attached to the pcb;
1584 * allocate one and initialize to default values.
1585 */
1586 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
1587 M_WAITOK);
1588
1589 if (imo == NULL)
1590 return (ENOBUFS);
1591 *imop = imo;
1592 imo->imo_multicast_ifp = NULL;
1593 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1594 imo->imo_multicast_vif = -1;
1595 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1596 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1597 imo->imo_num_memberships = 0;
1598 }
1599
1600 switch (sopt->sopt_name) {
1601 /* store an index number for the vif you wanna use in the send */
1602 case IP_MULTICAST_VIF:
1603 if (legal_vif_num == 0) {
1604 error = EOPNOTSUPP;
1605 break;
1606 }
1607 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
1608 if (error)
1609 break;
1610 if (!legal_vif_num(i) && (i != -1)) {
1611 error = EINVAL;
1612 break;
1613 }
1614 imo->imo_multicast_vif = i;
1615 break;
1616
1617 case IP_MULTICAST_IF:
1618 /*
1619 * Select the interface for outgoing multicast packets.
1620 */
1621 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr);
1622 if (error)
1623 break;
1624 /*
1625 * INADDR_ANY is used to remove a previous selection.
1626 * When no interface is selected, a default one is
1627 * chosen every time a multicast packet is sent.
1628 */
1629 if (addr.s_addr == INADDR_ANY) {
1630 imo->imo_multicast_ifp = NULL;
1631 break;
1632 }
1633 /*
1634 * The selected interface is identified by its local
1635 * IP address. Find the interface and confirm that
1636 * it supports multicasting.
1637 */
1638 s = splimp();
1639 ifp = ip_multicast_if(&addr, &ifindex);
1640 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1641 splx(s);
1642 error = EADDRNOTAVAIL;
1643 break;
1644 }
1645 imo->imo_multicast_ifp = ifp;
1646 if (ifindex)
1647 imo->imo_multicast_addr = addr;
1648 else
1649 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1650 splx(s);
1651 break;
1652
1653 case IP_MULTICAST_TTL:
1654 /*
1655 * Set the IP time-to-live for outgoing multicast packets.
1656 * The original multicast API required a char argument,
1657 * which is inconsistent with the rest of the socket API.
1658 * We allow either a char or an int.
1659 */
1660 if (sopt->sopt_valsize == 1) {
1661 u_char ttl;
1662 error = sooptcopyin(sopt, &ttl, 1, 1);
1663 if (error)
1664 break;
1665 imo->imo_multicast_ttl = ttl;
1666 } else {
1667 u_int ttl;
1668 error = sooptcopyin(sopt, &ttl, sizeof ttl,
1669 sizeof ttl);
1670 if (error)
1671 break;
1672 if (ttl > 255)
1673 error = EINVAL;
1674 else
1675 imo->imo_multicast_ttl = ttl;
1676 }
1677 break;
1678
1679 case IP_MULTICAST_LOOP:
1680 /*
1681 * Set the loopback flag for outgoing multicast packets.
1682 * Must be zero or one. The original multicast API required a
1683 * char argument, which is inconsistent with the rest
1684 * of the socket API. We allow either a char or an int.
1685 */
1686 if (sopt->sopt_valsize == 1) {
1687 u_char loop;
1688 error = sooptcopyin(sopt, &loop, 1, 1);
1689 if (error)
1690 break;
1691 imo->imo_multicast_loop = !!loop;
1692 } else {
1693 u_int loop;
1694 error = sooptcopyin(sopt, &loop, sizeof loop,
1695 sizeof loop);
1696 if (error)
1697 break;
1698 imo->imo_multicast_loop = !!loop;
1699 }
1700 break;
1701
1702 case IP_ADD_MEMBERSHIP:
1703 /*
1704 * Add a multicast group membership.
1705 * Group must be a valid IP multicast address.
1706 */
1707 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1708 if (error)
1709 break;
1710
1711 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1712 error = EINVAL;
1713 break;
1714 }
1715 s = splimp();
1716 /*
1717 * If no interface address was provided, use the interface of
1718 * the route to the given multicast address.
1719 */
1720 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1721 bzero((caddr_t)&ro, sizeof(ro));
1722 dst = (struct sockaddr_in *)&ro.ro_dst;
1723 dst->sin_len = sizeof(*dst);
1724 dst->sin_family = AF_INET;
1725 dst->sin_addr = mreq.imr_multiaddr;
1726 rtalloc(&ro);
1727 if (ro.ro_rt == NULL) {
1728 error = EADDRNOTAVAIL;
1729 splx(s);
1730 break;
1731 }
1732 ifp = ro.ro_rt->rt_ifp;
1733 rtfree(ro.ro_rt);
1734 }
1735 else {
1736 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1737 }
1738
1739 /*
1740 * See if we found an interface, and confirm that it
1741 * supports multicast.
1742 */
1743 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1744 error = EADDRNOTAVAIL;
1745 splx(s);
1746 break;
1747 }
1748 /*
1749 * See if the membership already exists or if all the
1750 * membership slots are full.
1751 */
1752 for (i = 0; i < imo->imo_num_memberships; ++i) {
1753 if (imo->imo_membership[i]->inm_ifp == ifp &&
1754 imo->imo_membership[i]->inm_addr.s_addr
1755 == mreq.imr_multiaddr.s_addr)
1756 break;
1757 }
1758 if (i < imo->imo_num_memberships) {
1759 error = EADDRINUSE;
1760 splx(s);
1761 break;
1762 }
1763 if (i == IP_MAX_MEMBERSHIPS) {
1764 error = ETOOMANYREFS;
1765 splx(s);
1766 break;
1767 }
1768 /*
1769 * Everything looks good; add a new record to the multicast
1770 * address list for the given interface.
1771 */
1772 if ((imo->imo_membership[i] =
1773 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
1774 error = ENOBUFS;
1775 splx(s);
1776 break;
1777 }
1778 ++imo->imo_num_memberships;
1779 splx(s);
1780 break;
1781
1782 case IP_DROP_MEMBERSHIP:
1783 /*
1784 * Drop a multicast group membership.
1785 * Group must be a valid IP multicast address.
1786 */
1787 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1788 if (error)
1789 break;
1790
1791 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1792 error = EINVAL;
1793 break;
1794 }
1795
1796 s = splimp();
1797 /*
1798 * If an interface address was specified, get a pointer
1799 * to its ifnet structure.
1800 */
1801 if (mreq.imr_interface.s_addr == INADDR_ANY)
1802 ifp = NULL;
1803 else {
1804 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1805 if (ifp == NULL) {
1806 error = EADDRNOTAVAIL;
1807 splx(s);
1808 break;
1809 }
1810 }
1811 /*
1812 * Find the membership in the membership array.
1813 */
1814 for (i = 0; i < imo->imo_num_memberships; ++i) {
1815 if ((ifp == NULL ||
1816 imo->imo_membership[i]->inm_ifp == ifp) &&
1817 imo->imo_membership[i]->inm_addr.s_addr ==
1818 mreq.imr_multiaddr.s_addr)
1819 break;
1820 }
1821 if (i == imo->imo_num_memberships) {
1822 error = EADDRNOTAVAIL;
1823 splx(s);
1824 break;
1825 }
1826 /*
1827 * Give up the multicast address record to which the
1828 * membership points.
1829 */
1830 in_delmulti(imo->imo_membership[i]);
1831 /*
1832 * Remove the gap in the membership array.
1833 */
1834 for (++i; i < imo->imo_num_memberships; ++i)
1835 imo->imo_membership[i-1] = imo->imo_membership[i];
1836 --imo->imo_num_memberships;
1837 splx(s);
1838 break;
1839
1840 default:
1841 error = EOPNOTSUPP;
1842 break;
1843 }
1844
1845 /*
1846 * If all options have default values, no need to keep the mbuf.
1847 */
1848 if (imo->imo_multicast_ifp == NULL &&
1849 imo->imo_multicast_vif == -1 &&
1850 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1851 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1852 imo->imo_num_memberships == 0) {
1853 free(*imop, M_IPMOPTS);
1854 *imop = NULL;
1855 }
1856
1857 return (error);
1858}
1859
1860/*
1861 * Return the IP multicast options in response to user getsockopt().
1862 */
1863static int
1864ip_getmoptions(sopt, imo)
1865 struct sockopt *sopt;
1866 register struct ip_moptions *imo;
1867{
1868 struct in_addr addr;
1869 struct in_ifaddr *ia;
1870 int error, optval;
1871 u_char coptval;
1872
1873 error = 0;
1874 switch (sopt->sopt_name) {
1875 case IP_MULTICAST_VIF:
1876 if (imo != NULL)
1877 optval = imo->imo_multicast_vif;
1878 else
1879 optval = -1;
1880 error = sooptcopyout(sopt, &optval, sizeof optval);
1881 break;
1882
1883 case IP_MULTICAST_IF:
1884 if (imo == NULL || imo->imo_multicast_ifp == NULL)
1885 addr.s_addr = INADDR_ANY;
1886 else if (imo->imo_multicast_addr.s_addr) {
1887 /* return the value user has set */
1888 addr = imo->imo_multicast_addr;
1889 } else {
1890 IFP_TO_IA(imo->imo_multicast_ifp, ia);
1891 addr.s_addr = (ia == NULL) ? INADDR_ANY
1892 : IA_SIN(ia)->sin_addr.s_addr;
1893 }
1894 error = sooptcopyout(sopt, &addr, sizeof addr);
1895 break;
1896
1897 case IP_MULTICAST_TTL:
1898 if (imo == 0)
1899 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
1900 else
1901 optval = coptval = imo->imo_multicast_ttl;
1902 if (sopt->sopt_valsize == 1)
1903 error = sooptcopyout(sopt, &coptval, 1);
1904 else
1905 error = sooptcopyout(sopt, &optval, sizeof optval);
1906 break;
1907
1908 case IP_MULTICAST_LOOP:
1909 if (imo == 0)
1910 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
1911 else
1912 optval = coptval = imo->imo_multicast_loop;
1913 if (sopt->sopt_valsize == 1)
1914 error = sooptcopyout(sopt, &coptval, 1);
1915 else
1916 error = sooptcopyout(sopt, &optval, sizeof optval);
1917 break;
1918
1919 default:
1920 error = ENOPROTOOPT;
1921 break;
1922 }
1923 return (error);
1924}
1925
1926/*
1927 * Discard the IP multicast options.
1928 */
1929void
1930ip_freemoptions(imo)
1931 register struct ip_moptions *imo;
1932{
1933 register int i;
1934
1935 if (imo != NULL) {
1936 for (i = 0; i < imo->imo_num_memberships; ++i)
1937 in_delmulti(imo->imo_membership[i]);
1938 free(imo, M_IPMOPTS);
1939 }
1940}
1941
1942/*
1943 * Routine called from ip_output() to loop back a copy of an IP multicast
1944 * packet to the input queue of a specified interface. Note that this
1945 * calls the output routine of the loopback "driver", but with an interface
1946 * pointer that might NOT be a loopback interface -- evil, but easier than
1947 * replicating that code here.
1948 */
1949static void
1950ip_mloopback(ifp, m, dst, hlen)
1951 struct ifnet *ifp;
1952 register struct mbuf *m;
1953 register struct sockaddr_in *dst;
1954 int hlen;
1955{
1956 register struct ip *ip;
1957 struct mbuf *copym;
1958
1959 copym = m_copy(m, 0, M_COPYALL);
1960 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1961 copym = m_pullup(copym, hlen);
1962 if (copym != NULL) {
1963 /*
1964 * We don't bother to fragment if the IP length is greater
1965 * than the interface's MTU. Can this possibly matter?
1966 */
1967 ip = mtod(copym, struct ip *);
1968 ip->ip_len = htons(ip->ip_len);
1969 ip->ip_off = htons(ip->ip_off);
1970 ip->ip_sum = 0;
1971 if (ip->ip_vhl == IP_VHL_BORING) {
1972 ip->ip_sum = in_cksum_hdr(ip);
1973 } else {
1974 ip->ip_sum = in_cksum(copym, hlen);
1975 }
1976 /*
1977 * NB:
1978 * It's not clear whether there are any lingering
1979 * reentrancy problems in other areas which might
1980 * be exposed by using ip_input directly (in
1981 * particular, everything which modifies the packet
1982 * in-place). Yet another option is using the
1983 * protosw directly to deliver the looped back
1984 * packet. For the moment, we'll err on the side
1985 * of safety by using if_simloop().
1986 */
1987#if 1 /* XXX */
1988 if (dst->sin_family != AF_INET) {
1989 printf("ip_mloopback: bad address family %d\n",
1990 dst->sin_family);
1991 dst->sin_family = AF_INET;
1992 }
1993#endif
1994
1995#ifdef notdef
1996 copym->m_pkthdr.rcvif = ifp;
1997 ip_input(copym);
1998#else
1999 /* if the checksum hasn't been computed, mark it as valid */
2000 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2001 copym->m_pkthdr.csum_flags |=
2002 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2003 copym->m_pkthdr.csum_data = 0xffff;
2004 }
2005 if_simloop(ifp, copym, dst->sin_family, 0);
2006#endif
2007 }
2008}