2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
33 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
34 * $FreeBSD: src/sys/netinet/ip_output.c,v 1.99.2.37 2003/04/15 06:44:45 silby Exp $
35 * $DragonFly: src/sys/netinet/ip_output.c,v 1.10 2004/02/14 21:12:39 dillon Exp $
42 #include "opt_ipdivert.h"
43 #include "opt_ipfilter.h"
44 #include "opt_ipsec.h"
45 #include "opt_pfil_hooks.h"
46 #include "opt_random_ip_id.h"
47 #include "opt_mbuf_stress_test.h"
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/kernel.h>
52 #include <sys/malloc.h>
54 #include <sys/protosw.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
58 #include <sys/sysctl.h>
59 #include <sys/in_cksum.h>
65 #include <net/route.h>
67 #include <netinet/in.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/ip.h>
70 #include <netinet/in_pcb.h>
71 #include <netinet/in_var.h>
72 #include <netinet/ip_var.h>
74 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
77 #include <netinet6/ipsec.h>
78 #include <netproto/key/key.h>
80 #include <netproto/key/key_debug.h>
82 #define KEYDEBUG(lev,arg)
87 #include <netipsec/ipsec.h>
88 #include <netipsec/xform.h>
89 #include <netipsec/key.h>
92 #include <net/ipfw/ip_fw.h>
93 #include <net/dummynet/ip_dummynet.h>
95 #define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\
96 x, (ntohl(a.s_addr)>>24)&0xFF,\
97 (ntohl(a.s_addr)>>16)&0xFF,\
98 (ntohl(a.s_addr)>>8)&0xFF,\
99 (ntohl(a.s_addr))&0xFF, y);
103 #ifdef MBUF_STRESS_TEST
104 int mbuf_frag_size = 0;
105 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
106 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
109 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
110 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
111 static void ip_mloopback
112 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
113 static int ip_getmoptions
114 (struct sockopt *, struct ip_moptions *);
115 static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
116 static int ip_setmoptions
117 (struct sockopt *, struct ip_moptions **);
119 int ip_optcopy(struct ip *, struct ip *);
120 extern int (*fr_checkp) (struct ip *, int, struct ifnet *, int, struct mbuf **);
123 extern struct protosw inetsw[];
126 * IP output. The packet in mbuf chain m contains a skeletal IP
127 * header (with len, off, ttl, proto, tos, src, dst).
128 * The mbuf chain containing the packet will be freed.
129 * The mbuf opt, if present, will not be freed.
132 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
133 int flags, struct ip_moptions *imo, struct inpcb *inp)
136 struct ifnet *ifp = NULL; /* keep compiler happy */
138 int hlen = sizeof (struct ip);
139 int len, off, error = 0;
140 struct sockaddr_in *dst = NULL; /* keep compiler happy */
141 struct in_ifaddr *ia = NULL;
142 int isbroadcast, sw_csum;
143 struct in_addr pkt_dst;
145 struct route iproute;
146 struct secpolicy *sp = NULL;
147 struct socket *so = inp ? inp->inp_socket : NULL;
150 struct route iproute;
152 struct secpolicy *sp = NULL;
153 struct tdb_ident *tdbi;
155 #endif /* FAST_IPSEC */
156 struct ip_fw_args args;
157 int src_was_INADDR_ANY = 0; /* as the name says... */
161 args.next_hop = NULL;
162 args.divert_rule = 0; /* divert cookie */
164 /* Grab info from MT_TAG mbufs prepended to the chain. */
165 for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) {
166 switch(m0->_m_tag_id) {
168 printf("ip_output: unrecognised MT_TAG tag %d\n",
172 case PACKET_TAG_DUMMYNET:
174 * the packet was already tagged, so part of the
175 * processing was already done, and we need to go down.
176 * Get parameters from the header.
178 args.rule = ((struct dn_pkt *)m0)->rule;
180 ro = & ( ((struct dn_pkt *)m0)->ro ) ;
182 dst = ((struct dn_pkt *)m0)->dn_dst ;
183 ifp = ((struct dn_pkt *)m0)->ifp ;
184 flags = ((struct dn_pkt *)m0)->flags ;
187 case PACKET_TAG_DIVERT:
188 args.divert_rule = (int)m0->m_data & 0xffff;
191 case PACKET_TAG_IPFORWARD:
192 args.next_hop = (struct sockaddr_in *)m0->m_data;
198 KASSERT(!m || (m->m_flags & M_PKTHDR) != 0, ("ip_output: no HDR"));
200 KASSERT(ro != NULL, ("ip_output: no route, proto %d",
201 mtod(m, struct ip *)->ip_p));
204 if (args.rule != NULL) { /* dummynet already saw us */
205 ip = mtod(m, struct ip *);
206 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
208 ia = ifatoia(ro->ro_rt->rt_ifa);
214 m = ip_insertoptions(m, opt, &len);
218 ip = mtod(m, struct ip *);
219 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
224 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
225 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
228 ip->ip_id = ip_randomid();
230 ip->ip_id = htons(ip_id++);
232 ipstat.ips_localout++;
234 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
240 bzero(ro, sizeof (*ro));
242 #endif /* FAST_IPSEC */
243 dst = (struct sockaddr_in *)&ro->ro_dst;
245 * If there is a cached route,
246 * check that it is to the same destination
247 * and is still up. If not, free it and try again.
248 * The address family should also be checked in case of sharing the
251 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
252 dst->sin_family != AF_INET ||
253 dst->sin_addr.s_addr != pkt_dst.s_addr)) {
255 ro->ro_rt = (struct rtentry *)0;
257 if (ro->ro_rt == 0) {
258 bzero(dst, sizeof(*dst));
259 dst->sin_family = AF_INET;
260 dst->sin_len = sizeof(*dst);
261 dst->sin_addr = pkt_dst;
264 * If routing to interface only,
265 * short circuit routing lookup.
267 if (flags & IP_ROUTETOIF) {
268 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
269 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
270 ipstat.ips_noroute++;
276 isbroadcast = in_broadcast(dst->sin_addr, ifp);
277 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
278 imo != NULL && imo->imo_multicast_ifp != NULL) {
280 * Bypass the normal routing lookup for multicast
281 * packets if the interface is specified.
283 ifp = imo->imo_multicast_ifp;
285 isbroadcast = 0; /* fool gcc */
288 * If this is the case, we probably don't want to allocate
289 * a protocol-cloned route since we didn't get one from the
290 * ULP. This lets TCP do its thing, while not burdening
291 * forwarding or ICMP with the overhead of cloning a route.
292 * Of course, we still want to do any cloning requested by
293 * the link layer, as this is probably required in all cases
294 * for correct operation (as it is for ARP).
297 rtalloc_ign(ro, RTF_PRCLONING);
298 if (ro->ro_rt == 0) {
299 ipstat.ips_noroute++;
300 error = EHOSTUNREACH;
303 ia = ifatoia(ro->ro_rt->rt_ifa);
304 ifp = ro->ro_rt->rt_ifp;
306 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
307 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
308 if (ro->ro_rt->rt_flags & RTF_HOST)
309 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
311 isbroadcast = in_broadcast(dst->sin_addr, ifp);
313 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
314 struct in_multi *inm;
316 m->m_flags |= M_MCAST;
318 * IP destination address is multicast. Make sure "dst"
319 * still points to the address in "ro". (It may have been
320 * changed to point to a gateway address, above.)
322 dst = (struct sockaddr_in *)&ro->ro_dst;
324 * See if the caller provided any multicast options
327 ip->ip_ttl = imo->imo_multicast_ttl;
328 if (imo->imo_multicast_vif != -1)
331 ip_mcast_src(imo->imo_multicast_vif) :
334 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
336 * Confirm that the outgoing interface supports multicast.
338 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
339 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
340 ipstat.ips_noroute++;
346 * If source address not specified yet, use address
347 * of outgoing interface.
349 if (ip->ip_src.s_addr == INADDR_ANY) {
350 /* Interface may have no addresses. */
352 ip->ip_src = IA_SIN(ia)->sin_addr;
355 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
358 * delayed checksums are not currently
359 * compatible with IP multicast routing
361 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
363 m->m_pkthdr.csum_flags &=
367 IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
369 (imo == NULL || imo->imo_multicast_loop)) {
371 * If we belong to the destination multicast group
372 * on the outgoing interface, and the caller did not
373 * forbid loopback, loop back a copy.
375 ip_mloopback(ifp, m, dst, hlen);
379 * If we are acting as a multicast router, perform
380 * multicast forwarding as if the packet had just
381 * arrived on the interface to which we are about
382 * to send. The multicast forwarding function
383 * recursively calls this function, using the
384 * IP_FORWARDING flag to prevent infinite recursion.
386 * Multicasts that are looped back by ip_mloopback(),
387 * above, will be forwarded by the ip_input() routine,
390 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
392 * If rsvp daemon is not running, do not
393 * set ip_moptions. This ensures that the packet
394 * is multicast and not just sent down one link
395 * as prescribed by rsvpd.
400 ip_mforward(ip, ifp, m, imo) != 0) {
408 * Multicasts with a time-to-live of zero may be looped-
409 * back, above, but must not be transmitted on a network.
410 * Also, multicasts addressed to the loopback interface
411 * are not sent -- the above call to ip_mloopback() will
412 * loop back a copy if this host actually belongs to the
413 * destination group on the loopback interface.
415 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
424 * If the source address is not specified yet, use the address
425 * of the outoing interface. In case, keep note we did that, so
426 * if the the firewall changes the next-hop causing the output
427 * interface to change, we can fix that.
429 if (ip->ip_src.s_addr == INADDR_ANY) {
430 /* Interface may have no addresses. */
432 ip->ip_src = IA_SIN(ia)->sin_addr;
433 src_was_INADDR_ANY = 1;
438 * Verify that we have any chance at all of being able to queue
439 * the packet or packet fragments
441 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
442 ifp->if_snd.ifq_maxlen) {
444 ipstat.ips_odropped++;
449 * Look for broadcast address and
450 * verify user is allowed to send
454 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
455 error = EADDRNOTAVAIL;
458 if ((flags & IP_ALLOWBROADCAST) == 0) {
462 /* don't allow broadcast messages to be fragmented */
463 if (ip->ip_len > ifp->if_mtu) {
467 m->m_flags |= M_BCAST;
469 m->m_flags &= ~M_BCAST;
474 /* get SP for this packet */
476 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
478 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
481 ipsecstat.out_inval++;
488 switch (sp->policy) {
489 case IPSEC_POLICY_DISCARD:
491 * This packet is just discarded.
493 ipsecstat.out_polvio++;
496 case IPSEC_POLICY_BYPASS:
497 case IPSEC_POLICY_NONE:
498 /* no need to do IPsec. */
501 case IPSEC_POLICY_IPSEC:
502 if (sp->req == NULL) {
503 /* acquire a policy */
504 error = key_spdacquire(sp);
509 case IPSEC_POLICY_ENTRUST:
511 printf("ip_output: Invalid policy found. %d\n", sp->policy);
514 struct ipsec_output_state state;
515 bzero(&state, sizeof(state));
517 if (flags & IP_ROUTETOIF) {
519 bzero(&iproute, sizeof(iproute));
522 state.dst = (struct sockaddr *)dst;
528 * delayed checksums are not currently compatible with IPsec
530 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
532 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
535 ip->ip_len = htons(ip->ip_len);
536 ip->ip_off = htons(ip->ip_off);
538 error = ipsec4_output(&state, sp, flags);
541 if (flags & IP_ROUTETOIF) {
543 * if we have tunnel mode SA, we may need to ignore
546 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
547 flags &= ~IP_ROUTETOIF;
552 dst = (struct sockaddr_in *)state.dst;
554 /* mbuf is already reclaimed in ipsec4_output. */
564 printf("ip4_output (ipsec): error code %d\n", error);
567 /* don't show these error codes to the user */
575 /* be sure to update variables that are affected by ipsec4_output() */
576 ip = mtod(m, struct ip *);
578 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
580 hlen = ip->ip_hl << 2;
582 if (ro->ro_rt == NULL) {
583 if ((flags & IP_ROUTETOIF) == 0) {
585 "can't update route after IPsec processing\n");
586 error = EHOSTUNREACH; /*XXX*/
590 ia = ifatoia(ro->ro_rt->rt_ifa);
591 ifp = ro->ro_rt->rt_ifp;
594 /* make it flipped, again. */
595 ip->ip_len = ntohs(ip->ip_len);
596 ip->ip_off = ntohs(ip->ip_off);
601 * Check the security policy (SP) for the packet and, if
602 * required, do IPsec-related processing. There are two
603 * cases here; the first time a packet is sent through
604 * it will be untagged and handled by ipsec4_checkpolicy.
605 * If the packet is resubmitted to ip_output (e.g. after
606 * AH, ESP, etc. processing), there will be a tag to bypass
607 * the lookup and related policy checking.
609 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
612 tdbi = (struct tdb_ident *)(mtag + 1);
613 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
615 error = -EINVAL; /* force silent drop */
616 m_tag_delete(m, mtag);
618 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
622 * There are four return cases:
623 * sp != NULL apply IPsec policy
624 * sp == NULL, error == 0 no IPsec handling needed
625 * sp == NULL, error == -EINVAL discard packet w/o error
626 * sp == NULL, error != 0 discard packet, report error
629 /* Loop detection, check if ipsec processing already done */
630 KASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
631 for (mtag = m_tag_first(m); mtag != NULL;
632 mtag = m_tag_next(m, mtag)) {
633 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
635 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
636 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
639 * Check if policy has an SA associated with it.
640 * This can happen when an SP has yet to acquire
641 * an SA; e.g. on first reference. If it occurs,
642 * then we let ipsec4_process_packet do its thing.
644 if (sp->req->sav == NULL)
646 tdbi = (struct tdb_ident *)(mtag + 1);
647 if (tdbi->spi == sp->req->sav->spi &&
648 tdbi->proto == sp->req->sav->sah->saidx.proto &&
649 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
650 sizeof (union sockaddr_union)) == 0) {
652 * No IPsec processing is needed, free
655 * NB: null pointer to avoid free at
658 KEY_FREESP(&sp), sp = NULL;
665 * Do delayed checksums now because we send before
666 * this is done in the normal processing path.
668 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
670 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
673 ip->ip_len = htons(ip->ip_len);
674 ip->ip_off = htons(ip->ip_off);
676 /* NB: callee frees mbuf */
677 error = ipsec4_process_packet(m, sp->req, flags, 0);
679 * Preserve KAME behaviour: ENOENT can be returned
680 * when an SA acquire is in progress. Don't propagate
681 * this to user-level; it confuses applications.
683 * XXX this will go away when the SADB is redone.
694 * Hack: -EINVAL is used to signal that a packet
695 * should be silently discarded. This is typically
696 * because we asked key management for an SA and
697 * it was delayed (e.g. kicked up to IKE).
699 if (error == -EINVAL)
703 /* No IPsec processing for this packet. */
707 * If deferred crypto processing is needed, check that
708 * the interface supports it.
710 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
711 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) {
712 /* notify IPsec to do its own crypto */
713 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
714 error = EHOSTUNREACH;
720 #endif /* FAST_IPSEC */
723 * - Xlate: translate packet's addr/port (NAT).
724 * - Firewall: deny/allow/etc.
725 * - Wrap: fake packet's addr/port <unimpl.>
726 * - Encapsulate: put it in another IP and send out. <unimp.>
730 * Run through list of hooks for output packets.
732 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT);
733 if (error != 0 || m == NULL)
735 ip = mtod(m, struct ip *);
736 #endif /* PFIL_HOOKS */
739 * Check with the firewall...
740 * but not if we are already being fwd'd from a firewall.
742 if (fw_enable && IPFW_LOADED && !args.next_hop) {
743 struct sockaddr_in *old = dst;
748 off = ip_fw_chk_ptr(&args);
753 * On return we must do the following:
754 * m == NULL -> drop the pkt (old interface, deprecated)
755 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface)
756 * 1<=off<= 0xffff -> DIVERT
757 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe
758 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet
759 * dst != old -> IPFIREWALL_FORWARD
760 * off==0, dst==old -> accept
761 * If some of the above modules are not compiled in, then
762 * we should't have to check the corresponding condition
763 * (because the ipfw control socket should not accept
764 * unsupported rules), but better play safe and drop
765 * packets in case of doubt.
767 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
773 ip = mtod(m, struct ip *);
774 if (off == 0 && dst == old) /* common case */
776 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) {
778 * pass the pkt to dummynet. Need to include
779 * pipe number, m, ifp, ro, dst because these are
780 * not recomputed in the next pass.
781 * All other parameters have been already used and
782 * so they are not needed anymore.
783 * XXX note: if the ifp or ro entry are deleted
784 * while a pkt is in dummynet, we are in trouble!
790 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT,
795 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) {
796 struct mbuf *clone = NULL;
798 /* Clone packet if we're doing a 'tee' */
799 if ((off & IP_FW_PORT_TEE_FLAG) != 0)
800 clone = m_dup(m, M_DONTWAIT);
804 * delayed checksums are not currently compatible
805 * with divert sockets.
807 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
809 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
812 /* Restore packet header fields to original values */
813 ip->ip_len = htons(ip->ip_len);
814 ip->ip_off = htons(ip->ip_off);
816 /* Deliver packet to divert input routine */
817 divert_packet(m, 0, off & 0xffff, args.divert_rule);
819 /* If 'tee', continue with original packet */
822 ip = mtod(m, struct ip *);
829 /* IPFIREWALL_FORWARD */
831 * Check dst to make sure it is directly reachable on the
832 * interface we previously thought it was.
833 * If it isn't (which may be likely in some situations) we have
834 * to re-route it (ie, find a route for the next-hop and the
835 * associated interface) and set them here. This is nested
836 * forwarding which in most cases is undesirable, except where
837 * such control is nigh impossible. So we do it here.
840 if (off == 0 && old != dst) { /* FORWARD, dst has changed */
843 * XXX To improve readability, this block should be
844 * changed into a function call as below:
846 error = ip_ipforward(&m, &dst, &ifp);
849 if (m == NULL) /* ip_input consumed the mbuf */
852 struct in_ifaddr *ia;
855 * XXX sro_fwd below is static, and a pointer
856 * to it gets passed to routines downstream.
857 * This could have surprisingly bad results in
858 * practice, because its content is overwritten
859 * by subsequent packets.
861 /* There must be a better way to do this next line... */
862 static struct route sro_fwd;
863 struct route *ro_fwd = &sro_fwd;
866 print_ip("IPFIREWALL_FORWARD: New dst ip: ",
867 dst->sin_addr, "\n");
871 * We need to figure out if we have been forwarded
872 * to a local socket. If so, then we should somehow
873 * "loop back" to ip_input, and get directed to the
874 * PCB as if we had received this packet. This is
875 * because it may be dificult to identify the packets
876 * you want to forward until they are being output
877 * and have selected an interface. (e.g. locally
878 * initiated packets) If we used the loopback inteface,
879 * we would not be able to control what happens
880 * as the packet runs through ip_input() as
881 * it is done through a ISR.
884 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) {
886 * If the addr to forward to is one
887 * of ours, we pretend to
888 * be the destination for this packet.
890 if (IA_SIN(ia)->sin_addr.s_addr ==
891 dst->sin_addr.s_addr)
894 if (ia) { /* tell ip_input "dont filter" */
897 tag.mh_type = MT_TAG;
898 tag.mh_flags = PACKET_TAG_IPFORWARD;
899 tag.mh_data = (caddr_t)args.next_hop;
902 if (m->m_pkthdr.rcvif == NULL)
903 m->m_pkthdr.rcvif = ifunit("lo0");
904 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
905 m->m_pkthdr.csum_flags |=
906 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
907 m0->m_pkthdr.csum_data = 0xffff;
909 m->m_pkthdr.csum_flags |=
910 CSUM_IP_CHECKED | CSUM_IP_VALID;
911 ip->ip_len = htons(ip->ip_len);
912 ip->ip_off = htons(ip->ip_off);
913 ip_input((struct mbuf *)&tag);
916 /* Some of the logic for this was
919 * This rewrites the cached route in a local PCB.
920 * Is this what we want to do?
922 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst));
925 rtalloc_ign(ro_fwd, RTF_PRCLONING);
927 if (ro_fwd->ro_rt == 0) {
928 ipstat.ips_noroute++;
929 error = EHOSTUNREACH;
933 ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
934 ifp = ro_fwd->ro_rt->rt_ifp;
935 ro_fwd->ro_rt->rt_use++;
936 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
937 dst = (struct sockaddr_in *)
938 ro_fwd->ro_rt->rt_gateway;
939 if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
941 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
943 isbroadcast = in_broadcast(dst->sin_addr, ifp);
946 ro->ro_rt = ro_fwd->ro_rt;
947 dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
949 #endif /* ... block to be put into a function */
951 * If we added a default src ip earlier,
952 * which would have been gotten from the-then
953 * interface, do it again, from the new one.
955 if (src_was_INADDR_ANY)
956 ip->ip_src = IA_SIN(ia)->sin_addr;
961 * if we get here, none of the above matches, and
962 * we have to drop the pkt
965 error = EACCES; /* not sure this is the right error msg */
970 /* 127/8 must not appear on wire - RFC1122. */
971 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
972 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
973 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
974 ipstat.ips_badaddr++;
975 error = EADDRNOTAVAIL;
980 m->m_pkthdr.csum_flags |= CSUM_IP;
981 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
982 if (sw_csum & CSUM_DELAY_DATA) {
984 sw_csum &= ~CSUM_DELAY_DATA;
986 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
989 * If small enough for interface, or the interface will take
990 * care of the fragmentation for us, can just send directly.
992 if (ip->ip_len <= ifp->if_mtu || ifp->if_hwassist & CSUM_FRAGMENT) {
993 ip->ip_len = htons(ip->ip_len);
994 ip->ip_off = htons(ip->ip_off);
996 if (sw_csum & CSUM_DELAY_IP) {
997 if (ip->ip_vhl == IP_VHL_BORING) {
998 ip->ip_sum = in_cksum_hdr(ip);
1000 ip->ip_sum = in_cksum(m, hlen);
1004 /* Record statistics for this interface address. */
1005 if (!(flags & IP_FORWARDING) && ia) {
1006 ia->ia_ifa.if_opackets++;
1007 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1011 /* clean ipsec history once it goes out of the node */
1015 #ifdef MBUF_STRESS_TEST
1016 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) {
1017 struct mbuf *m1, *m2;
1020 tmp = length = m->m_pkthdr.len;
1022 while ((length -= mbuf_frag_size) >= 1) {
1023 m1 = m_split(m, length, M_DONTWAIT);
1026 m1->m_flags &= ~M_PKTHDR;
1028 while (m2->m_next != NULL)
1032 m->m_pkthdr.len = tmp;
1035 error = (*ifp->if_output)(ifp, m,
1036 (struct sockaddr *)dst, ro->ro_rt);
1040 if (ip->ip_off & IP_DF) {
1043 * This case can happen if the user changed the MTU
1044 * of an interface after enabling IP on it. Because
1045 * most netifs don't keep track of routes pointing to
1046 * them, there is no way for one to update all its
1047 * routes when the MTU is changed.
1049 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
1050 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
1051 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
1052 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
1054 ipstat.ips_cantfrag++;
1059 * Too large for interface; fragment if possible. If successful,
1060 * on return, m will point to a list of packets to be sent.
1062 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
1069 /* clean ipsec history once it goes out of the node */
1073 /* Record statistics for this interface address. */
1075 ia->ia_ifa.if_opackets++;
1076 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1079 error = (*ifp->if_output)(ifp, m,
1080 (struct sockaddr *)dst, ro->ro_rt);
1086 ipstat.ips_fragmented++;
1090 if (ro == &iproute && ro->ro_rt) {
1095 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1096 printf("DP ip_output call free SP:%p\n", sp));
1101 if (ro == &iproute && ro->ro_rt) {
1115 * Create a chain of fragments which fit the given mtu. m_frag points to the
1116 * mbuf to be fragmented; on return it points to the chain with the fragments.
1117 * Return 0 if no error. If error, m_frag may contain a partially built
1118 * chain of fragments that should be freed by the caller.
1120 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
1121 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
1124 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
1125 u_long if_hwassist_flags, int sw_csum)
1128 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1129 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
1131 struct mbuf *m0 = *m_frag; /* the original packet */
1133 struct mbuf **mnext;
1136 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
1137 ipstat.ips_cantfrag++;
1142 * Must be able to put at least 8 bytes per fragment.
1148 * If the interface will not calculate checksums on
1149 * fragmented packets, then do it here.
1151 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
1152 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
1153 in_delayed_cksum(m0);
1154 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1157 if (len > PAGE_SIZE) {
1159 * Fragment large datagrams such that each segment
1160 * contains a multiple of PAGE_SIZE amount of data,
1161 * plus headers. This enables a receiver to perform
1162 * page-flipping zero-copy optimizations.
1164 * XXX When does this help given that sender and receiver
1165 * could have different page sizes, and also mtu could
1166 * be less than the receiver's page size ?
1171 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
1175 * firstlen (off - hlen) must be aligned on an
1179 goto smart_frag_failure;
1180 off = ((off - hlen) & ~7) + hlen;
1181 newlen = (~PAGE_MASK) & mtu;
1182 if ((newlen + sizeof (struct ip)) > mtu) {
1183 /* we failed, go back the default */
1194 firstlen = off - hlen;
1195 mnext = &m0->m_nextpkt; /* pointer to next packet */
1198 * Loop through length of segment after first fragment,
1199 * make new header and copy data of each part and link onto chain.
1200 * Here, m0 is the original packet, m is the fragment being created.
1201 * The fragments are linked off the m_nextpkt of the original
1202 * packet, which after processing serves as the first fragment.
1204 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
1205 struct ip *mhip; /* ip header on the fragment */
1207 int mhlen = sizeof (struct ip);
1209 MGETHDR(m, M_DONTWAIT, MT_HEADER);
1212 ipstat.ips_odropped++;
1215 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
1217 * In the first mbuf, leave room for the link header, then
1218 * copy the original IP header including options. The payload
1219 * goes into an additional mbuf chain returned by m_copy().
1221 m->m_data += max_linkhdr;
1222 mhip = mtod(m, struct ip *);
1224 if (hlen > sizeof (struct ip)) {
1225 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1226 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
1229 /* XXX do we need to add ip->ip_off below ? */
1230 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
1231 if (off + len >= ip->ip_len) { /* last fragment */
1232 len = ip->ip_len - off;
1233 m->m_flags |= M_LASTFRAG;
1235 mhip->ip_off |= IP_MF;
1236 mhip->ip_len = htons((u_short)(len + mhlen));
1237 m->m_next = m_copy(m0, off, len);
1238 if (m->m_next == 0) { /* copy failed */
1240 error = ENOBUFS; /* ??? */
1241 ipstat.ips_odropped++;
1244 m->m_pkthdr.len = mhlen + len;
1245 m->m_pkthdr.rcvif = (struct ifnet *)0;
1246 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1247 mhip->ip_off = htons(mhip->ip_off);
1249 if (sw_csum & CSUM_DELAY_IP)
1250 mhip->ip_sum = in_cksum(m, mhlen);
1252 mnext = &m->m_nextpkt;
1254 ipstat.ips_ofragments += nfrags;
1256 /* set first marker for fragment chain */
1257 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1258 m0->m_pkthdr.csum_data = nfrags;
1261 * Update first fragment by trimming what's been copied out
1262 * and updating header.
1264 m_adj(m0, hlen + firstlen - ip->ip_len);
1265 m0->m_pkthdr.len = hlen + firstlen;
1266 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1267 ip->ip_off |= IP_MF;
1268 ip->ip_off = htons(ip->ip_off);
1270 if (sw_csum & CSUM_DELAY_IP)
1271 ip->ip_sum = in_cksum(m0, hlen);
1279 in_delayed_cksum(struct mbuf *m)
1282 u_short csum, offset;
1284 ip = mtod(m, struct ip *);
1285 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1286 csum = in_cksum_skip(m, ip->ip_len, offset);
1287 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1289 offset += m->m_pkthdr.csum_data; /* checksum offset */
1291 if (offset + sizeof(u_short) > m->m_len) {
1292 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1293 m->m_len, offset, ip->ip_p);
1296 * this shouldn't happen, but if it does, the
1297 * correct behavior may be to insert the checksum
1298 * in the existing chain instead of rearranging it.
1300 m = m_pullup(m, offset + sizeof(u_short));
1302 *(u_short *)(m->m_data + offset) = csum;
1306 * Insert IP options into preformed packet.
1307 * Adjust IP destination as required for IP source routing,
1308 * as indicated by a non-zero in_addr at the start of the options.
1310 * XXX This routine assumes that the packet has no options in place.
1312 static struct mbuf *
1313 ip_insertoptions(m, opt, phlen)
1318 struct ipoption *p = mtod(opt, struct ipoption *);
1320 struct ip *ip = mtod(m, struct ip *);
1323 optlen = opt->m_len - sizeof(p->ipopt_dst);
1324 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
1326 return (m); /* XXX should fail */
1328 if (p->ipopt_dst.s_addr)
1329 ip->ip_dst = p->ipopt_dst;
1330 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1331 MGETHDR(n, M_DONTWAIT, MT_HEADER);
1336 n->m_pkthdr.rcvif = (struct ifnet *)0;
1337 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1338 m->m_len -= sizeof(struct ip);
1339 m->m_data += sizeof(struct ip);
1342 m->m_len = optlen + sizeof(struct ip);
1343 m->m_data += max_linkhdr;
1344 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip));
1346 m->m_data -= optlen;
1348 m->m_pkthdr.len += optlen;
1349 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1351 ip = mtod(m, struct ip *);
1352 bcopy(p->ipopt_list, ip + 1, optlen);
1353 *phlen = sizeof(struct ip) + optlen;
1354 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1355 ip->ip_len += optlen;
1360 * Copy options from ip to jp,
1361 * omitting those not copied during fragmentation.
1368 int opt, optlen, cnt;
1370 cp = (u_char *)(ip + 1);
1371 dp = (u_char *)(jp + 1);
1372 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1373 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1375 if (opt == IPOPT_EOL)
1377 if (opt == IPOPT_NOP) {
1378 /* Preserve for IP mcast tunnel's LSRR alignment. */
1384 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp),
1385 ("ip_optcopy: malformed ipv4 option"));
1386 optlen = cp[IPOPT_OLEN];
1387 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt,
1388 ("ip_optcopy: malformed ipv4 option"));
1390 /* bogus lengths should have been caught by ip_dooptions */
1393 if (IPOPT_COPIED(opt)) {
1394 bcopy(cp, dp, optlen);
1398 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1404 * IP socket option processing.
1407 ip_ctloutput(so, sopt)
1409 struct sockopt *sopt;
1411 struct inpcb *inp = sotoinpcb(so);
1415 if (sopt->sopt_level != IPPROTO_IP) {
1419 switch (sopt->sopt_dir) {
1421 switch (sopt->sopt_name) {
1428 if (sopt->sopt_valsize > MLEN) {
1432 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_HEADER);
1437 m->m_len = sopt->sopt_valsize;
1438 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1441 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1448 case IP_RECVRETOPTS:
1449 case IP_RECVDSTADDR:
1452 error = sooptcopyin(sopt, &optval, sizeof optval,
1457 switch (sopt->sopt_name) {
1459 inp->inp_ip_tos = optval;
1463 inp->inp_ip_ttl = optval;
1465 #define OPTSET(bit) \
1467 inp->inp_flags |= bit; \
1469 inp->inp_flags &= ~bit;
1472 OPTSET(INP_RECVOPTS);
1475 case IP_RECVRETOPTS:
1476 OPTSET(INP_RECVRETOPTS);
1479 case IP_RECVDSTADDR:
1480 OPTSET(INP_RECVDSTADDR);
1494 case IP_MULTICAST_IF:
1495 case IP_MULTICAST_VIF:
1496 case IP_MULTICAST_TTL:
1497 case IP_MULTICAST_LOOP:
1498 case IP_ADD_MEMBERSHIP:
1499 case IP_DROP_MEMBERSHIP:
1500 error = ip_setmoptions(sopt, &inp->inp_moptions);
1504 error = sooptcopyin(sopt, &optval, sizeof optval,
1510 case IP_PORTRANGE_DEFAULT:
1511 inp->inp_flags &= ~(INP_LOWPORT);
1512 inp->inp_flags &= ~(INP_HIGHPORT);
1515 case IP_PORTRANGE_HIGH:
1516 inp->inp_flags &= ~(INP_LOWPORT);
1517 inp->inp_flags |= INP_HIGHPORT;
1520 case IP_PORTRANGE_LOW:
1521 inp->inp_flags &= ~(INP_HIGHPORT);
1522 inp->inp_flags |= INP_LOWPORT;
1531 #if defined(IPSEC) || defined(FAST_IPSEC)
1532 case IP_IPSEC_POLICY:
1540 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1542 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1544 priv = (sopt->sopt_td != NULL &&
1545 suser(sopt->sopt_td) != 0) ? 0 : 1;
1546 req = mtod(m, caddr_t);
1548 optname = sopt->sopt_name;
1549 error = ipsec4_set_policy(inp, optname, req, len, priv);
1556 error = ENOPROTOOPT;
1562 switch (sopt->sopt_name) {
1565 if (inp->inp_options)
1566 error = sooptcopyout(sopt,
1567 mtod(inp->inp_options,
1569 inp->inp_options->m_len);
1571 sopt->sopt_valsize = 0;
1577 case IP_RECVRETOPTS:
1578 case IP_RECVDSTADDR:
1582 switch (sopt->sopt_name) {
1585 optval = inp->inp_ip_tos;
1589 optval = inp->inp_ip_ttl;
1592 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1595 optval = OPTBIT(INP_RECVOPTS);
1598 case IP_RECVRETOPTS:
1599 optval = OPTBIT(INP_RECVRETOPTS);
1602 case IP_RECVDSTADDR:
1603 optval = OPTBIT(INP_RECVDSTADDR);
1607 optval = OPTBIT(INP_RECVIF);
1611 if (inp->inp_flags & INP_HIGHPORT)
1612 optval = IP_PORTRANGE_HIGH;
1613 else if (inp->inp_flags & INP_LOWPORT)
1614 optval = IP_PORTRANGE_LOW;
1620 optval = OPTBIT(INP_FAITH);
1623 error = sooptcopyout(sopt, &optval, sizeof optval);
1626 case IP_MULTICAST_IF:
1627 case IP_MULTICAST_VIF:
1628 case IP_MULTICAST_TTL:
1629 case IP_MULTICAST_LOOP:
1630 case IP_ADD_MEMBERSHIP:
1631 case IP_DROP_MEMBERSHIP:
1632 error = ip_getmoptions(sopt, inp->inp_moptions);
1635 #if defined(IPSEC) || defined(FAST_IPSEC)
1636 case IP_IPSEC_POLICY:
1638 struct mbuf *m = NULL;
1643 req = mtod(m, caddr_t);
1646 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m);
1648 error = soopt_mcopyout(sopt, m); /* XXX */
1656 error = ENOPROTOOPT;
1665 * Set up IP options in pcb for insertion in output packets.
1666 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1667 * with destination address if source routed.
1670 ip_pcbopts(optname, pcbopt, m)
1672 struct mbuf **pcbopt;
1679 /* turn off any old options */
1681 (void)m_free(*pcbopt);
1683 if (m == (struct mbuf *)0 || m->m_len == 0) {
1685 * Only turning off any previous options.
1692 if (m->m_len % sizeof(int32_t))
1695 * IP first-hop destination address will be stored before
1696 * actual options; move other options back
1697 * and clear it when none present.
1699 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1702 m->m_len += sizeof(struct in_addr);
1703 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1704 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
1705 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1707 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1708 opt = cp[IPOPT_OPTVAL];
1709 if (opt == IPOPT_EOL)
1711 if (opt == IPOPT_NOP)
1714 if (cnt < IPOPT_OLEN + sizeof(*cp))
1716 optlen = cp[IPOPT_OLEN];
1717 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1728 * user process specifies route as:
1730 * D must be our final destination (but we can't
1731 * check that since we may not have connected yet).
1732 * A is first hop destination, which doesn't appear in
1733 * actual IP option, but is stored before the options.
1735 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1737 m->m_len -= sizeof(struct in_addr);
1738 cnt -= sizeof(struct in_addr);
1739 optlen -= sizeof(struct in_addr);
1740 cp[IPOPT_OLEN] = optlen;
1742 * Move first hop before start of options.
1744 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1745 sizeof(struct in_addr));
1747 * Then copy rest of options back
1748 * to close up the deleted entry.
1750 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1751 sizeof(struct in_addr)),
1752 (caddr_t)&cp[IPOPT_OFFSET+1],
1753 (unsigned)cnt + sizeof(struct in_addr));
1757 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1769 * The whole multicast option thing needs to be re-thought.
1770 * Several of these options are equally applicable to non-multicast
1771 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1772 * standard option (IP_TTL).
1776 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1778 static struct ifnet *
1779 ip_multicast_if(a, ifindexp)
1788 if (ntohl(a->s_addr) >> 24 == 0) {
1789 ifindex = ntohl(a->s_addr) & 0xffffff;
1790 if (ifindex < 0 || if_index < ifindex)
1792 ifp = ifindex2ifnet[ifindex];
1794 *ifindexp = ifindex;
1796 INADDR_TO_IFP(*a, ifp);
1802 * Set the IP multicast options in response to user setsockopt().
1805 ip_setmoptions(sopt, imop)
1806 struct sockopt *sopt;
1807 struct ip_moptions **imop;
1811 struct in_addr addr;
1812 struct ip_mreq mreq;
1814 struct ip_moptions *imo = *imop;
1816 struct sockaddr_in *dst;
1822 * No multicast option buffer attached to the pcb;
1823 * allocate one and initialize to default values.
1825 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
1831 imo->imo_multicast_ifp = NULL;
1832 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1833 imo->imo_multicast_vif = -1;
1834 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1835 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1836 imo->imo_num_memberships = 0;
1839 switch (sopt->sopt_name) {
1840 /* store an index number for the vif you wanna use in the send */
1841 case IP_MULTICAST_VIF:
1842 if (legal_vif_num == 0) {
1846 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
1849 if (!legal_vif_num(i) && (i != -1)) {
1853 imo->imo_multicast_vif = i;
1856 case IP_MULTICAST_IF:
1858 * Select the interface for outgoing multicast packets.
1860 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr);
1864 * INADDR_ANY is used to remove a previous selection.
1865 * When no interface is selected, a default one is
1866 * chosen every time a multicast packet is sent.
1868 if (addr.s_addr == INADDR_ANY) {
1869 imo->imo_multicast_ifp = NULL;
1873 * The selected interface is identified by its local
1874 * IP address. Find the interface and confirm that
1875 * it supports multicasting.
1878 ifp = ip_multicast_if(&addr, &ifindex);
1879 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1881 error = EADDRNOTAVAIL;
1884 imo->imo_multicast_ifp = ifp;
1886 imo->imo_multicast_addr = addr;
1888 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1892 case IP_MULTICAST_TTL:
1894 * Set the IP time-to-live for outgoing multicast packets.
1895 * The original multicast API required a char argument,
1896 * which is inconsistent with the rest of the socket API.
1897 * We allow either a char or an int.
1899 if (sopt->sopt_valsize == 1) {
1901 error = sooptcopyin(sopt, &ttl, 1, 1);
1904 imo->imo_multicast_ttl = ttl;
1907 error = sooptcopyin(sopt, &ttl, sizeof ttl,
1914 imo->imo_multicast_ttl = ttl;
1918 case IP_MULTICAST_LOOP:
1920 * Set the loopback flag for outgoing multicast packets.
1921 * Must be zero or one. The original multicast API required a
1922 * char argument, which is inconsistent with the rest
1923 * of the socket API. We allow either a char or an int.
1925 if (sopt->sopt_valsize == 1) {
1927 error = sooptcopyin(sopt, &loop, 1, 1);
1930 imo->imo_multicast_loop = !!loop;
1933 error = sooptcopyin(sopt, &loop, sizeof loop,
1937 imo->imo_multicast_loop = !!loop;
1941 case IP_ADD_MEMBERSHIP:
1943 * Add a multicast group membership.
1944 * Group must be a valid IP multicast address.
1946 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1950 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1956 * If no interface address was provided, use the interface of
1957 * the route to the given multicast address.
1959 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1960 bzero((caddr_t)&ro, sizeof(ro));
1961 dst = (struct sockaddr_in *)&ro.ro_dst;
1962 dst->sin_len = sizeof(*dst);
1963 dst->sin_family = AF_INET;
1964 dst->sin_addr = mreq.imr_multiaddr;
1966 if (ro.ro_rt == NULL) {
1967 error = EADDRNOTAVAIL;
1971 ifp = ro.ro_rt->rt_ifp;
1975 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1979 * See if we found an interface, and confirm that it
1980 * supports multicast.
1982 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1983 error = EADDRNOTAVAIL;
1988 * See if the membership already exists or if all the
1989 * membership slots are full.
1991 for (i = 0; i < imo->imo_num_memberships; ++i) {
1992 if (imo->imo_membership[i]->inm_ifp == ifp &&
1993 imo->imo_membership[i]->inm_addr.s_addr
1994 == mreq.imr_multiaddr.s_addr)
1997 if (i < imo->imo_num_memberships) {
2002 if (i == IP_MAX_MEMBERSHIPS) {
2003 error = ETOOMANYREFS;
2008 * Everything looks good; add a new record to the multicast
2009 * address list for the given interface.
2011 if ((imo->imo_membership[i] =
2012 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
2017 ++imo->imo_num_memberships;
2021 case IP_DROP_MEMBERSHIP:
2023 * Drop a multicast group membership.
2024 * Group must be a valid IP multicast address.
2026 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
2030 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
2037 * If an interface address was specified, get a pointer
2038 * to its ifnet structure.
2040 if (mreq.imr_interface.s_addr == INADDR_ANY)
2043 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
2045 error = EADDRNOTAVAIL;
2051 * Find the membership in the membership array.
2053 for (i = 0; i < imo->imo_num_memberships; ++i) {
2055 imo->imo_membership[i]->inm_ifp == ifp) &&
2056 imo->imo_membership[i]->inm_addr.s_addr ==
2057 mreq.imr_multiaddr.s_addr)
2060 if (i == imo->imo_num_memberships) {
2061 error = EADDRNOTAVAIL;
2066 * Give up the multicast address record to which the
2067 * membership points.
2069 in_delmulti(imo->imo_membership[i]);
2071 * Remove the gap in the membership array.
2073 for (++i; i < imo->imo_num_memberships; ++i)
2074 imo->imo_membership[i-1] = imo->imo_membership[i];
2075 --imo->imo_num_memberships;
2085 * If all options have default values, no need to keep the mbuf.
2087 if (imo->imo_multicast_ifp == NULL &&
2088 imo->imo_multicast_vif == -1 &&
2089 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2090 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2091 imo->imo_num_memberships == 0) {
2092 free(*imop, M_IPMOPTS);
2100 * Return the IP multicast options in response to user getsockopt().
2103 ip_getmoptions(sopt, imo)
2104 struct sockopt *sopt;
2105 struct ip_moptions *imo;
2107 struct in_addr addr;
2108 struct in_ifaddr *ia;
2113 switch (sopt->sopt_name) {
2114 case IP_MULTICAST_VIF:
2116 optval = imo->imo_multicast_vif;
2119 error = sooptcopyout(sopt, &optval, sizeof optval);
2122 case IP_MULTICAST_IF:
2123 if (imo == NULL || imo->imo_multicast_ifp == NULL)
2124 addr.s_addr = INADDR_ANY;
2125 else if (imo->imo_multicast_addr.s_addr) {
2126 /* return the value user has set */
2127 addr = imo->imo_multicast_addr;
2129 IFP_TO_IA(imo->imo_multicast_ifp, ia);
2130 addr.s_addr = (ia == NULL) ? INADDR_ANY
2131 : IA_SIN(ia)->sin_addr.s_addr;
2133 error = sooptcopyout(sopt, &addr, sizeof addr);
2136 case IP_MULTICAST_TTL:
2138 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
2140 optval = coptval = imo->imo_multicast_ttl;
2141 if (sopt->sopt_valsize == 1)
2142 error = sooptcopyout(sopt, &coptval, 1);
2144 error = sooptcopyout(sopt, &optval, sizeof optval);
2147 case IP_MULTICAST_LOOP:
2149 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
2151 optval = coptval = imo->imo_multicast_loop;
2152 if (sopt->sopt_valsize == 1)
2153 error = sooptcopyout(sopt, &coptval, 1);
2155 error = sooptcopyout(sopt, &optval, sizeof optval);
2159 error = ENOPROTOOPT;
2166 * Discard the IP multicast options.
2169 ip_freemoptions(imo)
2170 struct ip_moptions *imo;
2175 for (i = 0; i < imo->imo_num_memberships; ++i)
2176 in_delmulti(imo->imo_membership[i]);
2177 free(imo, M_IPMOPTS);
2182 * Routine called from ip_output() to loop back a copy of an IP multicast
2183 * packet to the input queue of a specified interface. Note that this
2184 * calls the output routine of the loopback "driver", but with an interface
2185 * pointer that might NOT be a loopback interface -- evil, but easier than
2186 * replicating that code here.
2189 ip_mloopback(ifp, m, dst, hlen)
2192 struct sockaddr_in *dst;
2198 copym = m_copy(m, 0, M_COPYALL);
2199 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2200 copym = m_pullup(copym, hlen);
2201 if (copym != NULL) {
2203 * We don't bother to fragment if the IP length is greater
2204 * than the interface's MTU. Can this possibly matter?
2206 ip = mtod(copym, struct ip *);
2207 ip->ip_len = htons(ip->ip_len);
2208 ip->ip_off = htons(ip->ip_off);
2210 if (ip->ip_vhl == IP_VHL_BORING) {
2211 ip->ip_sum = in_cksum_hdr(ip);
2213 ip->ip_sum = in_cksum(copym, hlen);
2217 * It's not clear whether there are any lingering
2218 * reentrancy problems in other areas which might
2219 * be exposed by using ip_input directly (in
2220 * particular, everything which modifies the packet
2221 * in-place). Yet another option is using the
2222 * protosw directly to deliver the looped back
2223 * packet. For the moment, we'll err on the side
2224 * of safety by using if_simloop().
2227 if (dst->sin_family != AF_INET) {
2228 printf("ip_mloopback: bad address family %d\n",
2230 dst->sin_family = AF_INET;
2235 copym->m_pkthdr.rcvif = ifp;
2238 /* if the checksum hasn't been computed, mark it as valid */
2239 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2240 copym->m_pkthdr.csum_flags |=
2241 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2242 copym->m_pkthdr.csum_data = 0xffff;
2244 if_simloop(ifp, copym, dst->sin_family, 0);