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. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
30 * $FreeBSD: src/sys/netinet/ip_output.c,v 1.99.2.37 2003/04/15 06:44:45 silby Exp $
31 * $DragonFly: src/sys/netinet/ip_output.c,v 1.28 2005/04/18 14:26:57 joerg Exp $
38 #include "opt_ipdivert.h"
39 #include "opt_ipfilter.h"
40 #include "opt_ipsec.h"
41 #include "opt_random_ip_id.h"
42 #include "opt_mbuf_stress_test.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/malloc.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
53 #include <sys/sysctl.h>
54 #include <sys/in_cksum.h>
57 #include <net/netisr.h>
59 #include <net/route.h>
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/ip.h>
64 #include <netinet/in_pcb.h>
65 #include <netinet/in_var.h>
66 #include <netinet/ip_var.h>
68 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
71 #include <netinet6/ipsec.h>
72 #include <netproto/key/key.h>
74 #include <netproto/key/key_debug.h>
76 #define KEYDEBUG(lev,arg)
81 #include <netproto/ipsec/ipsec.h>
82 #include <netproto/ipsec/xform.h>
83 #include <netproto/ipsec/key.h>
86 #include <net/ipfw/ip_fw.h>
87 #include <net/dummynet/ip_dummynet.h>
89 #define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\
90 x, (ntohl(a.s_addr)>>24)&0xFF,\
91 (ntohl(a.s_addr)>>16)&0xFF,\
92 (ntohl(a.s_addr)>>8)&0xFF,\
93 (ntohl(a.s_addr))&0xFF, y);
97 #ifdef MBUF_STRESS_TEST
98 int mbuf_frag_size = 0;
99 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
100 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
103 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
104 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
105 static void ip_mloopback
106 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
107 static int ip_getmoptions
108 (struct sockopt *, struct ip_moptions *);
109 static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
110 static int ip_setmoptions
111 (struct sockopt *, struct ip_moptions **);
113 int ip_optcopy(struct ip *, struct ip *);
116 extern struct protosw inetsw[];
119 * IP output. The packet in mbuf chain m contains a skeletal IP
120 * header (with len, off, ttl, proto, tos, src, dst).
121 * The mbuf chain containing the packet will be freed.
122 * The mbuf opt, if present, will not be freed.
125 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
126 int flags, struct ip_moptions *imo, struct inpcb *inp)
129 struct ifnet *ifp = NULL; /* keep compiler happy */
131 int hlen = sizeof(struct ip);
132 int len, off, error = 0;
133 struct sockaddr_in *dst = NULL; /* keep compiler happy */
134 struct in_ifaddr *ia = NULL;
135 int isbroadcast, sw_csum;
136 struct in_addr pkt_dst;
137 struct route iproute;
139 struct secpolicy *sp = NULL;
140 struct socket *so = inp ? inp->inp_socket : NULL;
144 struct secpolicy *sp = NULL;
145 struct tdb_ident *tdbi;
147 #endif /* FAST_IPSEC */
148 struct ip_fw_args args;
149 int src_was_INADDR_ANY = 0; /* as the name says... */
153 args.next_hop = NULL;
155 /* Grab info from MT_TAG mbufs prepended to the chain. */
156 while (m0 != NULL && m0->m_type == MT_TAG) {
157 switch(m0->_m_tag_id) {
158 case PACKET_TAG_DUMMYNET:
160 * the packet was already tagged, so part of the
161 * processing was already done, and we need to go down.
162 * Get parameters from the header.
164 args.rule = ((struct dn_pkt *)m0)->rule;
166 ro = &((struct dn_pkt *)m0)->ro;
168 dst = ((struct dn_pkt *)m0)->dn_dst ;
169 ifp = ((struct dn_pkt *)m0)->ifp ;
170 flags = ((struct dn_pkt *)m0)->flags ;
172 case PACKET_TAG_IPFORWARD:
173 args.next_hop = (struct sockaddr_in *)m0->m_data;
176 printf("ip_output: unrecognised MT_TAG tag %d\n",
183 KASSERT(m != NULL && (m->m_flags & M_PKTHDR), ("ip_output: no HDR"));
187 bzero(ro, sizeof *ro);
190 if (args.rule != NULL) { /* dummynet already saw us */
191 ip = mtod(m, struct ip *);
192 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
194 ia = ifatoia(ro->ro_rt->rt_ifa);
200 m = ip_insertoptions(m, opt, &len);
204 ip = mtod(m, struct ip *);
205 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
210 if (!(flags & (IP_FORWARDING|IP_RAWOUTPUT))) {
211 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
214 ip->ip_id = ip_randomid();
216 ip->ip_id = htons(ip_id++);
218 ipstat.ips_localout++;
220 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
223 dst = (struct sockaddr_in *)&ro->ro_dst;
225 * If there is a cached route,
226 * check that it is to the same destination
227 * and is still up. If not, free it and try again.
228 * The address family should also be checked in case of sharing the
232 (!(ro->ro_rt->rt_flags & RTF_UP) ||
233 dst->sin_family != AF_INET ||
234 dst->sin_addr.s_addr != pkt_dst.s_addr)) {
236 ro->ro_rt = (struct rtentry *)NULL;
238 if (ro->ro_rt == NULL) {
239 bzero(dst, sizeof *dst);
240 dst->sin_family = AF_INET;
241 dst->sin_len = sizeof *dst;
242 dst->sin_addr = pkt_dst;
245 * If routing to interface only,
246 * short circuit routing lookup.
248 if (flags & IP_ROUTETOIF) {
249 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
250 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) {
251 ipstat.ips_noroute++;
257 isbroadcast = in_broadcast(dst->sin_addr, ifp);
258 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
259 imo != NULL && imo->imo_multicast_ifp != NULL) {
261 * Bypass the normal routing lookup for multicast
262 * packets if the interface is specified.
264 ifp = imo->imo_multicast_ifp;
266 isbroadcast = 0; /* fool gcc */
269 * If this is the case, we probably don't want to allocate
270 * a protocol-cloned route since we didn't get one from the
271 * ULP. This lets TCP do its thing, while not burdening
272 * forwarding or ICMP with the overhead of cloning a route.
273 * Of course, we still want to do any cloning requested by
274 * the link layer, as this is probably required in all cases
275 * for correct operation (as it is for ARP).
277 if (ro->ro_rt == NULL)
278 rtalloc_ign(ro, RTF_PRCLONING);
279 if (ro->ro_rt == NULL) {
280 ipstat.ips_noroute++;
281 error = EHOSTUNREACH;
284 ia = ifatoia(ro->ro_rt->rt_ifa);
285 ifp = ro->ro_rt->rt_ifp;
287 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
288 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
289 if (ro->ro_rt->rt_flags & RTF_HOST)
290 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
292 isbroadcast = in_broadcast(dst->sin_addr, ifp);
294 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
295 struct in_multi *inm;
297 m->m_flags |= M_MCAST;
299 * IP destination address is multicast. Make sure "dst"
300 * still points to the address in "ro". (It may have been
301 * changed to point to a gateway address, above.)
303 dst = (struct sockaddr_in *)&ro->ro_dst;
305 * See if the caller provided any multicast options
308 ip->ip_ttl = imo->imo_multicast_ttl;
309 if (imo->imo_multicast_vif != -1)
312 ip_mcast_src(imo->imo_multicast_vif) :
315 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
317 * Confirm that the outgoing interface supports multicast.
319 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
320 if (!(ifp->if_flags & IFF_MULTICAST)) {
321 ipstat.ips_noroute++;
327 * If source address not specified yet, use address
328 * of outgoing interface.
330 if (ip->ip_src.s_addr == INADDR_ANY) {
331 /* Interface may have no addresses. */
333 ip->ip_src = IA_SIN(ia)->sin_addr;
336 IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
338 (imo == NULL || imo->imo_multicast_loop)) {
340 * If we belong to the destination multicast group
341 * on the outgoing interface, and the caller did not
342 * forbid loopback, loop back a copy.
344 ip_mloopback(ifp, m, dst, hlen);
348 * If we are acting as a multicast router, perform
349 * multicast forwarding as if the packet had just
350 * arrived on the interface to which we are about
351 * to send. The multicast forwarding function
352 * recursively calls this function, using the
353 * IP_FORWARDING flag to prevent infinite recursion.
355 * Multicasts that are looped back by ip_mloopback(),
356 * above, will be forwarded by the ip_input() routine,
359 if (ip_mrouter && !(flags & IP_FORWARDING)) {
361 * If rsvp daemon is not running, do not
362 * set ip_moptions. This ensures that the packet
363 * is multicast and not just sent down one link
364 * as prescribed by rsvpd.
369 ip_mforward(ip, ifp, m, imo) != 0) {
377 * Multicasts with a time-to-live of zero may be looped-
378 * back, above, but must not be transmitted on a network.
379 * Also, multicasts addressed to the loopback interface
380 * are not sent -- the above call to ip_mloopback() will
381 * loop back a copy if this host actually belongs to the
382 * destination group on the loopback interface.
384 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
393 * If the source address is not specified yet, use the address
394 * of the outoing interface. In case, keep note we did that, so
395 * if the the firewall changes the next-hop causing the output
396 * interface to change, we can fix that.
398 if (ip->ip_src.s_addr == INADDR_ANY) {
399 /* Interface may have no addresses. */
401 ip->ip_src = IA_SIN(ia)->sin_addr;
402 src_was_INADDR_ANY = 1;
408 * Disable packet drop hack.
409 * Packetdrop should be done by queueing.
413 * Verify that we have any chance at all of being able to queue
414 * the packet or packet fragments
416 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
417 ifp->if_snd.ifq_maxlen) {
419 ipstat.ips_odropped++;
425 * Look for broadcast address and
426 * verify user is allowed to send
430 if (!(ifp->if_flags & IFF_BROADCAST)) {
431 error = EADDRNOTAVAIL;
434 if (!(flags & IP_ALLOWBROADCAST)) {
438 /* don't allow broadcast messages to be fragmented */
439 if (ip->ip_len > ifp->if_mtu) {
443 m->m_flags |= M_BCAST;
445 m->m_flags &= ~M_BCAST;
450 /* get SP for this packet */
452 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
454 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
457 ipsecstat.out_inval++;
464 switch (sp->policy) {
465 case IPSEC_POLICY_DISCARD:
467 * This packet is just discarded.
469 ipsecstat.out_polvio++;
472 case IPSEC_POLICY_BYPASS:
473 case IPSEC_POLICY_NONE:
474 /* no need to do IPsec. */
477 case IPSEC_POLICY_IPSEC:
478 if (sp->req == NULL) {
479 /* acquire a policy */
480 error = key_spdacquire(sp);
485 case IPSEC_POLICY_ENTRUST:
487 printf("ip_output: Invalid policy found. %d\n", sp->policy);
490 struct ipsec_output_state state;
491 bzero(&state, sizeof state);
493 if (flags & IP_ROUTETOIF) {
495 bzero(&iproute, sizeof iproute);
498 state.dst = (struct sockaddr *)dst;
504 * delayed checksums are not currently compatible with IPsec
506 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
508 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
511 ip->ip_len = htons(ip->ip_len);
512 ip->ip_off = htons(ip->ip_off);
514 error = ipsec4_output(&state, sp, flags);
517 if (flags & IP_ROUTETOIF) {
519 * if we have tunnel mode SA, we may need to ignore
522 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
523 flags &= ~IP_ROUTETOIF;
528 dst = (struct sockaddr_in *)state.dst;
530 /* mbuf is already reclaimed in ipsec4_output. */
540 printf("ip4_output (ipsec): error code %d\n", error);
543 /* don't show these error codes to the user */
551 /* be sure to update variables that are affected by ipsec4_output() */
552 ip = mtod(m, struct ip *);
554 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
556 hlen = ip->ip_hl << 2;
558 if (ro->ro_rt == NULL) {
559 if (!(flags & IP_ROUTETOIF)) {
561 "can't update route after IPsec processing\n");
562 error = EHOSTUNREACH; /*XXX*/
566 ia = ifatoia(ro->ro_rt->rt_ifa);
567 ifp = ro->ro_rt->rt_ifp;
570 /* make it flipped, again. */
571 ip->ip_len = ntohs(ip->ip_len);
572 ip->ip_off = ntohs(ip->ip_off);
577 * Check the security policy (SP) for the packet and, if
578 * required, do IPsec-related processing. There are two
579 * cases here; the first time a packet is sent through
580 * it will be untagged and handled by ipsec4_checkpolicy.
581 * If the packet is resubmitted to ip_output (e.g. after
582 * AH, ESP, etc. processing), there will be a tag to bypass
583 * the lookup and related policy checking.
585 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
588 tdbi = (struct tdb_ident *)(mtag + 1);
589 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
591 error = -EINVAL; /* force silent drop */
592 m_tag_delete(m, mtag);
594 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
598 * There are four return cases:
599 * sp != NULL apply IPsec policy
600 * sp == NULL, error == 0 no IPsec handling needed
601 * sp == NULL, error == -EINVAL discard packet w/o error
602 * sp == NULL, error != 0 discard packet, report error
605 /* Loop detection, check if ipsec processing already done */
606 KASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
607 for (mtag = m_tag_first(m); mtag != NULL;
608 mtag = m_tag_next(m, mtag)) {
609 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
611 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
612 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
615 * Check if policy has an SA associated with it.
616 * This can happen when an SP has yet to acquire
617 * an SA; e.g. on first reference. If it occurs,
618 * then we let ipsec4_process_packet do its thing.
620 if (sp->req->sav == NULL)
622 tdbi = (struct tdb_ident *)(mtag + 1);
623 if (tdbi->spi == sp->req->sav->spi &&
624 tdbi->proto == sp->req->sav->sah->saidx.proto &&
625 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
626 sizeof(union sockaddr_union)) == 0) {
628 * No IPsec processing is needed, free
631 * NB: null pointer to avoid free at
634 KEY_FREESP(&sp), sp = NULL;
641 * Do delayed checksums now because we send before
642 * this is done in the normal processing path.
644 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
646 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
649 ip->ip_len = htons(ip->ip_len);
650 ip->ip_off = htons(ip->ip_off);
652 /* NB: callee frees mbuf */
653 error = ipsec4_process_packet(m, sp->req, flags, 0);
655 * Preserve KAME behaviour: ENOENT can be returned
656 * when an SA acquire is in progress. Don't propagate
657 * this to user-level; it confuses applications.
659 * XXX this will go away when the SADB is redone.
670 * Hack: -EINVAL is used to signal that a packet
671 * should be silently discarded. This is typically
672 * because we asked key management for an SA and
673 * it was delayed (e.g. kicked up to IKE).
675 if (error == -EINVAL)
679 /* No IPsec processing for this packet. */
683 * If deferred crypto processing is needed, check that
684 * the interface supports it.
686 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
687 if (mtag != NULL && !(ifp->if_capenable & IFCAP_IPSEC)) {
688 /* notify IPsec to do its own crypto */
689 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
690 error = EHOSTUNREACH;
696 #endif /* FAST_IPSEC */
699 * - Xlate: translate packet's addr/port (NAT).
700 * - Firewall: deny/allow/etc.
701 * - Wrap: fake packet's addr/port <unimpl.>
702 * - Encapsulate: put it in another IP and send out. <unimp.>
706 * Run through list of hooks for output packets.
708 if (pfil_has_hooks(&inet_pfil_hook)) {
709 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT);
710 if (error != 0 || m == NULL)
712 ip = mtod(m, struct ip *);
716 * Check with the firewall...
717 * but not if we are already being fwd'd from a firewall.
719 if (fw_enable && IPFW_LOADED && !args.next_hop) {
720 struct sockaddr_in *old = dst;
725 off = ip_fw_chk_ptr(&args);
730 * On return we must do the following:
731 * m == NULL -> drop the pkt (old interface, deprecated)
732 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface)
733 * 1<=off<= 0xffff -> DIVERT
734 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe
735 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet
736 * dst != old -> IPFIREWALL_FORWARD
737 * off==0, dst==old -> accept
738 * If some of the above modules are not compiled in, then
739 * we should't have to check the corresponding condition
740 * (because the ipfw control socket should not accept
741 * unsupported rules), but better play safe and drop
742 * packets in case of doubt.
744 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
750 ip = mtod(m, struct ip *);
751 if (off == 0 && dst == old) /* common case */
753 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG)) {
755 * pass the pkt to dummynet. Need to include
756 * pipe number, m, ifp, ro, dst because these are
757 * not recomputed in the next pass.
758 * All other parameters have been already used and
759 * so they are not needed anymore.
760 * XXX note: if the ifp or ro entry are deleted
761 * while a pkt is in dummynet, we are in trouble!
767 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT,
772 if (off != 0 && !(off & IP_FW_PORT_DYNT_FLAG)) {
773 struct mbuf *clone = NULL;
775 /* Clone packet if we're doing a 'tee' */
776 if ((off & IP_FW_PORT_TEE_FLAG))
777 clone = m_dup(m, MB_DONTWAIT);
781 * delayed checksums are not currently compatible
782 * with divert sockets.
784 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
786 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
789 /* Restore packet header fields to original values */
790 ip->ip_len = htons(ip->ip_len);
791 ip->ip_off = htons(ip->ip_off);
793 /* Deliver packet to divert input routine */
794 divert_packet(m, 0, off & 0xffff);
796 /* If 'tee', continue with original packet */
799 ip = mtod(m, struct ip *);
806 /* IPFIREWALL_FORWARD */
808 * Check dst to make sure it is directly reachable on the
809 * interface we previously thought it was.
810 * If it isn't (which may be likely in some situations) we have
811 * to re-route it (ie, find a route for the next-hop and the
812 * associated interface) and set them here. This is nested
813 * forwarding which in most cases is undesirable, except where
814 * such control is nigh impossible. So we do it here.
817 if (off == 0 && old != dst) { /* FORWARD, dst has changed */
820 * XXX To improve readability, this block should be
821 * changed into a function call as below:
823 error = ip_ipforward(&m, &dst, &ifp);
826 if (m == NULL) /* ip_input consumed the mbuf */
829 struct in_ifaddr *ia;
832 * XXX sro_fwd below is static, and a pointer
833 * to it gets passed to routines downstream.
834 * This could have surprisingly bad results in
835 * practice, because its content is overwritten
836 * by subsequent packets.
838 /* There must be a better way to do this next line... */
839 static struct route sro_fwd;
840 struct route *ro_fwd = &sro_fwd;
843 print_ip("IPFIREWALL_FORWARD: New dst ip: ",
844 dst->sin_addr, "\n");
848 * We need to figure out if we have been forwarded
849 * to a local socket. If so, then we should somehow
850 * "loop back" to ip_input, and get directed to the
851 * PCB as if we had received this packet. This is
852 * because it may be dificult to identify the packets
853 * you want to forward until they are being output
854 * and have selected an interface. (e.g. locally
855 * initiated packets) If we used the loopback inteface,
856 * we would not be able to control what happens
857 * as the packet runs through ip_input() as
858 * it is done through a ISR.
860 LIST_FOREACH(ia, INADDR_HASH(dst->sin_addr.s_addr),
863 * If the addr to forward to is one
864 * of ours, we pretend to
865 * be the destination for this packet.
867 if (IA_SIN(ia)->sin_addr.s_addr ==
868 dst->sin_addr.s_addr)
871 if (ia != NULL) { /* tell ip_input "dont filter" */
874 tag.mh_type = MT_TAG;
875 tag.mh_flags = PACKET_TAG_IPFORWARD;
876 tag.mh_data = (caddr_t)args.next_hop;
879 if (m->m_pkthdr.rcvif == NULL)
880 m->m_pkthdr.rcvif = ifunit("lo0");
881 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
882 m->m_pkthdr.csum_flags |=
883 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
884 m0->m_pkthdr.csum_data = 0xffff;
886 m->m_pkthdr.csum_flags |=
887 CSUM_IP_CHECKED | CSUM_IP_VALID;
888 ip->ip_len = htons(ip->ip_len);
889 ip->ip_off = htons(ip->ip_off);
890 ip_input((struct mbuf *)&tag);
893 /* Some of the logic for this was nicked from above.
895 * This rewrites the cached route in a local PCB.
896 * Is this what we want to do?
898 bcopy(dst, &ro_fwd->ro_dst, sizeof *dst);
899 ro_fwd->ro_rt = NULL;
901 rtalloc_ign(ro_fwd, RTF_PRCLONING);
902 if (ro_fwd->ro_rt == NULL) {
903 ipstat.ips_noroute++;
904 error = EHOSTUNREACH;
908 ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
909 ifp = ro_fwd->ro_rt->rt_ifp;
910 ro_fwd->ro_rt->rt_use++;
911 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
912 dst = (struct sockaddr_in *)
913 ro_fwd->ro_rt->rt_gateway;
914 if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
916 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
918 isbroadcast = in_broadcast(dst->sin_addr, ifp);
919 if (ro->ro_rt != NULL)
921 ro->ro_rt = ro_fwd->ro_rt;
922 dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
924 #endif /* ... block to be put into a function */
926 * If we added a default src ip earlier,
927 * which would have been gotten from the-then
928 * interface, do it again, from the new one.
930 if (src_was_INADDR_ANY)
931 ip->ip_src = IA_SIN(ia)->sin_addr;
936 * if we get here, none of the above matches, and
937 * we have to drop the pkt
940 error = EACCES; /* not sure this is the right error msg */
945 /* 127/8 must not appear on wire - RFC1122. */
946 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
947 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
948 if (!(ifp->if_flags & IFF_LOOPBACK)) {
949 ipstat.ips_badaddr++;
950 error = EADDRNOTAVAIL;
955 m->m_pkthdr.csum_flags |= CSUM_IP;
956 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
957 if (sw_csum & CSUM_DELAY_DATA) {
959 sw_csum &= ~CSUM_DELAY_DATA;
961 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
964 * If small enough for interface, or the interface will take
965 * care of the fragmentation for us, can just send directly.
967 if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
968 !(ip->ip_off & IP_DF))) {
969 ip->ip_len = htons(ip->ip_len);
970 ip->ip_off = htons(ip->ip_off);
972 if (sw_csum & CSUM_DELAY_IP) {
973 if (ip->ip_vhl == IP_VHL_BORING) {
974 ip->ip_sum = in_cksum_hdr(ip);
976 ip->ip_sum = in_cksum(m, hlen);
980 /* Record statistics for this interface address. */
981 if (!(flags & IP_FORWARDING) && ia) {
982 ia->ia_ifa.if_opackets++;
983 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
987 /* clean ipsec history once it goes out of the node */
991 #ifdef MBUF_STRESS_TEST
992 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) {
993 struct mbuf *m1, *m2;
996 tmp = length = m->m_pkthdr.len;
998 while ((length -= mbuf_frag_size) >= 1) {
999 m1 = m_split(m, length, MB_DONTWAIT);
1002 m1->m_flags &= ~M_PKTHDR;
1004 while (m2->m_next != NULL)
1008 m->m_pkthdr.len = tmp;
1011 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst,
1016 if (ip->ip_off & IP_DF) {
1019 * This case can happen if the user changed the MTU
1020 * of an interface after enabling IP on it. Because
1021 * most netifs don't keep track of routes pointing to
1022 * them, there is no way for one to update all its
1023 * routes when the MTU is changed.
1025 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
1026 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
1027 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
1028 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
1030 ipstat.ips_cantfrag++;
1035 * Too large for interface; fragment if possible. If successful,
1036 * on return, m will point to a list of packets to be sent.
1038 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
1043 m->m_nextpkt = NULL;
1045 /* clean ipsec history once it goes out of the node */
1049 /* Record statistics for this interface address. */
1051 ia->ia_ifa.if_opackets++;
1052 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1054 error = (*ifp->if_output)(ifp, m,
1055 (struct sockaddr *)dst,
1062 ipstat.ips_fragmented++;
1065 if (ro == &iproute && ro->ro_rt != NULL) {
1071 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1072 printf("DP ip_output call free SP:%p\n", sp));
1087 * Create a chain of fragments which fit the given mtu. m_frag points to the
1088 * mbuf to be fragmented; on return it points to the chain with the fragments.
1089 * Return 0 if no error. If error, m_frag may contain a partially built
1090 * chain of fragments that should be freed by the caller.
1092 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
1093 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
1096 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
1097 u_long if_hwassist_flags, int sw_csum)
1100 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1101 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
1103 struct mbuf *m0 = *m_frag; /* the original packet */
1105 struct mbuf **mnext;
1108 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
1109 ipstat.ips_cantfrag++;
1114 * Must be able to put at least 8 bytes per fragment.
1120 * If the interface will not calculate checksums on
1121 * fragmented packets, then do it here.
1123 if ((m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) &&
1124 !(if_hwassist_flags & CSUM_IP_FRAGS)) {
1125 in_delayed_cksum(m0);
1126 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1129 if (len > PAGE_SIZE) {
1131 * Fragment large datagrams such that each segment
1132 * contains a multiple of PAGE_SIZE amount of data,
1133 * plus headers. This enables a receiver to perform
1134 * page-flipping zero-copy optimizations.
1136 * XXX When does this help given that sender and receiver
1137 * could have different page sizes, and also mtu could
1138 * be less than the receiver's page size ?
1143 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
1147 * firstlen (off - hlen) must be aligned on an
1151 goto smart_frag_failure;
1152 off = ((off - hlen) & ~7) + hlen;
1153 newlen = (~PAGE_MASK) & mtu;
1154 if ((newlen + sizeof(struct ip)) > mtu) {
1155 /* we failed, go back the default */
1166 firstlen = off - hlen;
1167 mnext = &m0->m_nextpkt; /* pointer to next packet */
1170 * Loop through length of segment after first fragment,
1171 * make new header and copy data of each part and link onto chain.
1172 * Here, m0 is the original packet, m is the fragment being created.
1173 * The fragments are linked off the m_nextpkt of the original
1174 * packet, which after processing serves as the first fragment.
1176 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
1177 struct ip *mhip; /* ip header on the fragment */
1179 int mhlen = sizeof(struct ip);
1181 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1184 ipstat.ips_odropped++;
1187 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
1189 * In the first mbuf, leave room for the link header, then
1190 * copy the original IP header including options. The payload
1191 * goes into an additional mbuf chain returned by m_copy().
1193 m->m_data += max_linkhdr;
1194 mhip = mtod(m, struct ip *);
1196 if (hlen > sizeof(struct ip)) {
1197 mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip);
1198 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
1201 /* XXX do we need to add ip->ip_off below ? */
1202 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
1203 if (off + len >= ip->ip_len) { /* last fragment */
1204 len = ip->ip_len - off;
1205 m->m_flags |= M_LASTFRAG;
1207 mhip->ip_off |= IP_MF;
1208 mhip->ip_len = htons((u_short)(len + mhlen));
1209 m->m_next = m_copy(m0, off, len);
1210 if (m->m_next == NULL) { /* copy failed */
1212 error = ENOBUFS; /* ??? */
1213 ipstat.ips_odropped++;
1216 m->m_pkthdr.len = mhlen + len;
1217 m->m_pkthdr.rcvif = (struct ifnet *)NULL;
1218 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1219 mhip->ip_off = htons(mhip->ip_off);
1221 if (sw_csum & CSUM_DELAY_IP)
1222 mhip->ip_sum = in_cksum(m, mhlen);
1224 mnext = &m->m_nextpkt;
1226 ipstat.ips_ofragments += nfrags;
1228 /* set first marker for fragment chain */
1229 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1230 m0->m_pkthdr.csum_data = nfrags;
1233 * Update first fragment by trimming what's been copied out
1234 * and updating header.
1236 m_adj(m0, hlen + firstlen - ip->ip_len);
1237 m0->m_pkthdr.len = hlen + firstlen;
1238 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1239 ip->ip_off |= IP_MF;
1240 ip->ip_off = htons(ip->ip_off);
1242 if (sw_csum & CSUM_DELAY_IP)
1243 ip->ip_sum = in_cksum(m0, hlen);
1251 in_delayed_cksum(struct mbuf *m)
1254 u_short csum, offset;
1256 ip = mtod(m, struct ip *);
1257 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1258 csum = in_cksum_skip(m, ip->ip_len, offset);
1259 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1261 offset += m->m_pkthdr.csum_data; /* checksum offset */
1263 if (offset + sizeof(u_short) > m->m_len) {
1264 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1265 m->m_len, offset, ip->ip_p);
1268 * this shouldn't happen, but if it does, the
1269 * correct behavior may be to insert the checksum
1270 * in the existing chain instead of rearranging it.
1272 m = m_pullup(m, offset + sizeof(u_short));
1274 *(u_short *)(m->m_data + offset) = csum;
1278 * Insert IP options into preformed packet.
1279 * Adjust IP destination as required for IP source routing,
1280 * as indicated by a non-zero in_addr at the start of the options.
1282 * XXX This routine assumes that the packet has no options in place.
1284 static struct mbuf *
1285 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
1287 struct ipoption *p = mtod(opt, struct ipoption *);
1289 struct ip *ip = mtod(m, struct ip *);
1292 optlen = opt->m_len - sizeof p->ipopt_dst;
1293 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
1295 return (m); /* XXX should fail */
1297 if (p->ipopt_dst.s_addr)
1298 ip->ip_dst = p->ipopt_dst;
1299 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1300 MGETHDR(n, MB_DONTWAIT, MT_HEADER);
1305 n->m_pkthdr.rcvif = (struct ifnet *)NULL;
1306 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1307 m->m_len -= sizeof(struct ip);
1308 m->m_data += sizeof(struct ip);
1311 m->m_len = optlen + sizeof(struct ip);
1312 m->m_data += max_linkhdr;
1313 memcpy(mtod(m, void *), ip, sizeof(struct ip));
1315 m->m_data -= optlen;
1317 m->m_pkthdr.len += optlen;
1318 ovbcopy(ip, mtod(m, caddr_t), sizeof(struct ip));
1320 ip = mtod(m, struct ip *);
1321 bcopy(p->ipopt_list, ip + 1, optlen);
1322 *phlen = sizeof(struct ip) + optlen;
1323 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1324 ip->ip_len += optlen;
1329 * Copy options from ip to jp,
1330 * omitting those not copied during fragmentation.
1333 ip_optcopy(struct ip *ip, struct ip *jp)
1336 int opt, optlen, cnt;
1338 cp = (u_char *)(ip + 1);
1339 dp = (u_char *)(jp + 1);
1340 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1341 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1343 if (opt == IPOPT_EOL)
1345 if (opt == IPOPT_NOP) {
1346 /* Preserve for IP mcast tunnel's LSRR alignment. */
1352 KASSERT(cnt >= IPOPT_OLEN + sizeof *cp,
1353 ("ip_optcopy: malformed ipv4 option"));
1354 optlen = cp[IPOPT_OLEN];
1355 KASSERT(optlen >= IPOPT_OLEN + sizeof *cp && optlen <= cnt,
1356 ("ip_optcopy: malformed ipv4 option"));
1358 /* bogus lengths should have been caught by ip_dooptions */
1361 if (IPOPT_COPIED(opt)) {
1362 bcopy(cp, dp, optlen);
1366 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1372 * IP socket option processing.
1375 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1377 struct inpcb *inp = so->so_pcb;
1381 if (sopt->sopt_level != IPPROTO_IP) {
1385 switch (sopt->sopt_dir) {
1387 switch (sopt->sopt_name) {
1394 if (sopt->sopt_valsize > MLEN) {
1398 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1403 m->m_len = sopt->sopt_valsize;
1404 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1407 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1414 case IP_RECVRETOPTS:
1415 case IP_RECVDSTADDR:
1418 error = sooptcopyin(sopt, &optval, sizeof optval,
1423 switch (sopt->sopt_name) {
1425 inp->inp_ip_tos = optval;
1429 inp->inp_ip_ttl = optval;
1431 #define OPTSET(bit) \
1433 inp->inp_flags |= bit; \
1435 inp->inp_flags &= ~bit;
1438 OPTSET(INP_RECVOPTS);
1441 case IP_RECVRETOPTS:
1442 OPTSET(INP_RECVRETOPTS);
1445 case IP_RECVDSTADDR:
1446 OPTSET(INP_RECVDSTADDR);
1460 case IP_MULTICAST_IF:
1461 case IP_MULTICAST_VIF:
1462 case IP_MULTICAST_TTL:
1463 case IP_MULTICAST_LOOP:
1464 case IP_ADD_MEMBERSHIP:
1465 case IP_DROP_MEMBERSHIP:
1466 error = ip_setmoptions(sopt, &inp->inp_moptions);
1470 error = sooptcopyin(sopt, &optval, sizeof optval,
1476 case IP_PORTRANGE_DEFAULT:
1477 inp->inp_flags &= ~(INP_LOWPORT);
1478 inp->inp_flags &= ~(INP_HIGHPORT);
1481 case IP_PORTRANGE_HIGH:
1482 inp->inp_flags &= ~(INP_LOWPORT);
1483 inp->inp_flags |= INP_HIGHPORT;
1486 case IP_PORTRANGE_LOW:
1487 inp->inp_flags &= ~(INP_HIGHPORT);
1488 inp->inp_flags |= INP_LOWPORT;
1497 #if defined(IPSEC) || defined(FAST_IPSEC)
1498 case IP_IPSEC_POLICY:
1506 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1508 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1510 priv = (sopt->sopt_td != NULL &&
1511 suser(sopt->sopt_td) != 0) ? 0 : 1;
1512 req = mtod(m, caddr_t);
1514 optname = sopt->sopt_name;
1515 error = ipsec4_set_policy(inp, optname, req, len, priv);
1522 error = ENOPROTOOPT;
1528 switch (sopt->sopt_name) {
1531 if (inp->inp_options)
1532 error = sooptcopyout(sopt,
1533 mtod(inp->inp_options,
1535 inp->inp_options->m_len);
1537 sopt->sopt_valsize = 0;
1543 case IP_RECVRETOPTS:
1544 case IP_RECVDSTADDR:
1548 switch (sopt->sopt_name) {
1551 optval = inp->inp_ip_tos;
1555 optval = inp->inp_ip_ttl;
1558 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1561 optval = OPTBIT(INP_RECVOPTS);
1564 case IP_RECVRETOPTS:
1565 optval = OPTBIT(INP_RECVRETOPTS);
1568 case IP_RECVDSTADDR:
1569 optval = OPTBIT(INP_RECVDSTADDR);
1573 optval = OPTBIT(INP_RECVIF);
1577 if (inp->inp_flags & INP_HIGHPORT)
1578 optval = IP_PORTRANGE_HIGH;
1579 else if (inp->inp_flags & INP_LOWPORT)
1580 optval = IP_PORTRANGE_LOW;
1586 optval = OPTBIT(INP_FAITH);
1589 error = sooptcopyout(sopt, &optval, sizeof optval);
1592 case IP_MULTICAST_IF:
1593 case IP_MULTICAST_VIF:
1594 case IP_MULTICAST_TTL:
1595 case IP_MULTICAST_LOOP:
1596 case IP_ADD_MEMBERSHIP:
1597 case IP_DROP_MEMBERSHIP:
1598 error = ip_getmoptions(sopt, inp->inp_moptions);
1601 #if defined(IPSEC) || defined(FAST_IPSEC)
1602 case IP_IPSEC_POLICY:
1604 struct mbuf *m = NULL;
1609 req = mtod(m, caddr_t);
1612 error = ipsec4_get_policy(so->so_pcb, req, len, &m);
1614 error = soopt_mcopyout(sopt, m); /* XXX */
1622 error = ENOPROTOOPT;
1631 * Set up IP options in pcb for insertion in output packets.
1632 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1633 * with destination address if source routed.
1636 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
1642 /* turn off any old options */
1646 if (m == NULL || m->m_len == 0) {
1648 * Only turning off any previous options.
1655 if (m->m_len % sizeof(int32_t))
1658 * IP first-hop destination address will be stored before
1659 * actual options; move other options back
1660 * and clear it when none present.
1662 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1665 m->m_len += sizeof(struct in_addr);
1666 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1667 ovbcopy(mtod(m, caddr_t), cp, cnt);
1668 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1670 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1671 opt = cp[IPOPT_OPTVAL];
1672 if (opt == IPOPT_EOL)
1674 if (opt == IPOPT_NOP)
1677 if (cnt < IPOPT_OLEN + sizeof *cp)
1679 optlen = cp[IPOPT_OLEN];
1680 if (optlen < IPOPT_OLEN + sizeof *cp || optlen > cnt)
1691 * user process specifies route as:
1693 * D must be our final destination (but we can't
1694 * check that since we may not have connected yet).
1695 * A is first hop destination, which doesn't appear in
1696 * actual IP option, but is stored before the options.
1698 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1700 m->m_len -= sizeof(struct in_addr);
1701 cnt -= sizeof(struct in_addr);
1702 optlen -= sizeof(struct in_addr);
1703 cp[IPOPT_OLEN] = optlen;
1705 * Move first hop before start of options.
1707 bcopy(&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1708 sizeof(struct in_addr));
1710 * Then copy rest of options back
1711 * to close up the deleted entry.
1713 ovbcopy(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr),
1714 &cp[IPOPT_OFFSET+1],
1715 cnt - (IPOPT_MINOFF - 1));
1719 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1731 * The whole multicast option thing needs to be re-thought.
1732 * Several of these options are equally applicable to non-multicast
1733 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1734 * standard option (IP_TTL).
1738 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1740 static struct ifnet *
1741 ip_multicast_if(struct in_addr *a, int *ifindexp)
1748 if (ntohl(a->s_addr) >> 24 == 0) {
1749 ifindex = ntohl(a->s_addr) & 0xffffff;
1750 if (ifindex < 0 || if_index < ifindex)
1752 ifp = ifindex2ifnet[ifindex];
1754 *ifindexp = ifindex;
1756 INADDR_TO_IFP(*a, ifp);
1762 * Set the IP multicast options in response to user setsockopt().
1765 ip_setmoptions(struct sockopt *sopt, struct ip_moptions **imop)
1769 struct in_addr addr;
1770 struct ip_mreq mreq;
1772 struct ip_moptions *imo = *imop;
1778 * No multicast option buffer attached to the pcb;
1779 * allocate one and initialize to default values.
1781 imo = malloc(sizeof *imo, M_IPMOPTS, M_WAITOK);
1786 imo->imo_multicast_ifp = NULL;
1787 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1788 imo->imo_multicast_vif = -1;
1789 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1790 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1791 imo->imo_num_memberships = 0;
1794 switch (sopt->sopt_name) {
1795 /* store an index number for the vif you wanna use in the send */
1796 case IP_MULTICAST_VIF:
1797 if (legal_vif_num == 0) {
1801 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
1804 if (!legal_vif_num(i) && (i != -1)) {
1808 imo->imo_multicast_vif = i;
1811 case IP_MULTICAST_IF:
1813 * Select the interface for outgoing multicast packets.
1815 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr);
1819 * INADDR_ANY is used to remove a previous selection.
1820 * When no interface is selected, a default one is
1821 * chosen every time a multicast packet is sent.
1823 if (addr.s_addr == INADDR_ANY) {
1824 imo->imo_multicast_ifp = NULL;
1828 * The selected interface is identified by its local
1829 * IP address. Find the interface and confirm that
1830 * it supports multicasting.
1833 ifp = ip_multicast_if(&addr, &ifindex);
1834 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1836 error = EADDRNOTAVAIL;
1839 imo->imo_multicast_ifp = ifp;
1841 imo->imo_multicast_addr = addr;
1843 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1847 case IP_MULTICAST_TTL:
1849 * Set the IP time-to-live for outgoing multicast packets.
1850 * The original multicast API required a char argument,
1851 * which is inconsistent with the rest of the socket API.
1852 * We allow either a char or an int.
1854 if (sopt->sopt_valsize == 1) {
1856 error = sooptcopyin(sopt, &ttl, 1, 1);
1859 imo->imo_multicast_ttl = ttl;
1862 error = sooptcopyin(sopt, &ttl, sizeof ttl, sizeof ttl);
1868 imo->imo_multicast_ttl = ttl;
1872 case IP_MULTICAST_LOOP:
1874 * Set the loopback flag for outgoing multicast packets.
1875 * Must be zero or one. The original multicast API required a
1876 * char argument, which is inconsistent with the rest
1877 * of the socket API. We allow either a char or an int.
1879 if (sopt->sopt_valsize == 1) {
1882 error = sooptcopyin(sopt, &loop, 1, 1);
1885 imo->imo_multicast_loop = !!loop;
1889 error = sooptcopyin(sopt, &loop, sizeof loop,
1893 imo->imo_multicast_loop = !!loop;
1897 case IP_ADD_MEMBERSHIP:
1899 * Add a multicast group membership.
1900 * Group must be a valid IP multicast address.
1902 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1906 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1912 * If no interface address was provided, use the interface of
1913 * the route to the given multicast address.
1915 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1916 struct sockaddr_in dst;
1919 bzero(&dst, sizeof(struct sockaddr_in));
1920 dst.sin_len = sizeof(struct sockaddr_in);
1921 dst.sin_family = AF_INET;
1922 dst.sin_addr = mreq.imr_multiaddr;
1923 rt = rtlookup((struct sockaddr *)&dst);
1925 error = EADDRNOTAVAIL;
1932 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1936 * See if we found an interface, and confirm that it
1937 * supports multicast.
1939 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1940 error = EADDRNOTAVAIL;
1945 * See if the membership already exists or if all the
1946 * membership slots are full.
1948 for (i = 0; i < imo->imo_num_memberships; ++i) {
1949 if (imo->imo_membership[i]->inm_ifp == ifp &&
1950 imo->imo_membership[i]->inm_addr.s_addr
1951 == mreq.imr_multiaddr.s_addr)
1954 if (i < imo->imo_num_memberships) {
1959 if (i == IP_MAX_MEMBERSHIPS) {
1960 error = ETOOMANYREFS;
1965 * Everything looks good; add a new record to the multicast
1966 * address list for the given interface.
1968 if ((imo->imo_membership[i] =
1969 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
1974 ++imo->imo_num_memberships;
1978 case IP_DROP_MEMBERSHIP:
1980 * Drop a multicast group membership.
1981 * Group must be a valid IP multicast address.
1983 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1987 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1994 * If an interface address was specified, get a pointer
1995 * to its ifnet structure.
1997 if (mreq.imr_interface.s_addr == INADDR_ANY)
2000 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
2002 error = EADDRNOTAVAIL;
2008 * Find the membership in the membership array.
2010 for (i = 0; i < imo->imo_num_memberships; ++i) {
2012 imo->imo_membership[i]->inm_ifp == ifp) &&
2013 imo->imo_membership[i]->inm_addr.s_addr ==
2014 mreq.imr_multiaddr.s_addr)
2017 if (i == imo->imo_num_memberships) {
2018 error = EADDRNOTAVAIL;
2023 * Give up the multicast address record to which the
2024 * membership points.
2026 in_delmulti(imo->imo_membership[i]);
2028 * Remove the gap in the membership array.
2030 for (++i; i < imo->imo_num_memberships; ++i)
2031 imo->imo_membership[i-1] = imo->imo_membership[i];
2032 --imo->imo_num_memberships;
2042 * If all options have default values, no need to keep the mbuf.
2044 if (imo->imo_multicast_ifp == NULL &&
2045 imo->imo_multicast_vif == -1 &&
2046 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2047 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2048 imo->imo_num_memberships == 0) {
2049 free(*imop, M_IPMOPTS);
2057 * Return the IP multicast options in response to user getsockopt().
2060 ip_getmoptions(struct sockopt *sopt, struct ip_moptions *imo)
2062 struct in_addr addr;
2063 struct in_ifaddr *ia;
2068 switch (sopt->sopt_name) {
2069 case IP_MULTICAST_VIF:
2071 optval = imo->imo_multicast_vif;
2074 error = sooptcopyout(sopt, &optval, sizeof optval);
2077 case IP_MULTICAST_IF:
2078 if (imo == NULL || imo->imo_multicast_ifp == NULL)
2079 addr.s_addr = INADDR_ANY;
2080 else if (imo->imo_multicast_addr.s_addr) {
2081 /* return the value user has set */
2082 addr = imo->imo_multicast_addr;
2084 IFP_TO_IA(imo->imo_multicast_ifp, ia);
2085 addr.s_addr = (ia == NULL) ? INADDR_ANY
2086 : IA_SIN(ia)->sin_addr.s_addr;
2088 error = sooptcopyout(sopt, &addr, sizeof addr);
2091 case IP_MULTICAST_TTL:
2093 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
2095 optval = coptval = imo->imo_multicast_ttl;
2096 if (sopt->sopt_valsize == 1)
2097 error = sooptcopyout(sopt, &coptval, 1);
2099 error = sooptcopyout(sopt, &optval, sizeof optval);
2102 case IP_MULTICAST_LOOP:
2104 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
2106 optval = coptval = imo->imo_multicast_loop;
2107 if (sopt->sopt_valsize == 1)
2108 error = sooptcopyout(sopt, &coptval, 1);
2110 error = sooptcopyout(sopt, &optval, sizeof optval);
2114 error = ENOPROTOOPT;
2121 * Discard the IP multicast options.
2124 ip_freemoptions(struct ip_moptions *imo)
2129 for (i = 0; i < imo->imo_num_memberships; ++i)
2130 in_delmulti(imo->imo_membership[i]);
2131 free(imo, M_IPMOPTS);
2136 * Routine called from ip_output() to loop back a copy of an IP multicast
2137 * packet to the input queue of a specified interface. Note that this
2138 * calls the output routine of the loopback "driver", but with an interface
2139 * pointer that might NOT be a loopback interface -- evil, but easier than
2140 * replicating that code here.
2143 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
2149 copym = m_copypacket(m, MB_DONTWAIT);
2150 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2151 copym = m_pullup(copym, hlen);
2152 if (copym != NULL) {
2154 * if the checksum hasn't been computed, mark it as valid
2156 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2157 in_delayed_cksum(copym);
2158 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2159 copym->m_pkthdr.csum_flags |=
2160 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2161 copym->m_pkthdr.csum_data = 0xffff;
2164 * We don't bother to fragment if the IP length is greater
2165 * than the interface's MTU. Can this possibly matter?
2167 ip = mtod(copym, struct ip *);
2168 ip->ip_len = htons(ip->ip_len);
2169 ip->ip_off = htons(ip->ip_off);
2171 if (ip->ip_vhl == IP_VHL_BORING) {
2172 ip->ip_sum = in_cksum_hdr(ip);
2174 ip->ip_sum = in_cksum(copym, hlen);
2178 * It's not clear whether there are any lingering
2179 * reentrancy problems in other areas which might
2180 * be exposed by using ip_input directly (in
2181 * particular, everything which modifies the packet
2182 * in-place). Yet another option is using the
2183 * protosw directly to deliver the looped back
2184 * packet. For the moment, we'll err on the side
2185 * of safety by using if_simloop().
2188 if (dst->sin_family != AF_INET) {
2189 printf("ip_mloopback: bad address family %d\n",
2191 dst->sin_family = AF_INET;
2196 copym->m_pkthdr.rcvif = ifp;
2199 if_simloop(ifp, copym, dst->sin_family, 0);