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 1.15 2004/06/24 08:15:17 dillon Exp $
42 #include "opt_ipdivert.h"
43 #include "opt_ipfilter.h"
44 #include "opt_ipsec.h"
45 #include "opt_random_ip_id.h"
46 #include "opt_mbuf_stress_test.h"
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/kernel.h>
51 #include <sys/malloc.h>
53 #include <sys/protosw.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
57 #include <sys/sysctl.h>
58 #include <sys/in_cksum.h>
61 #include <net/netisr.h>
63 #include <net/route.h>
65 #include <netinet/in.h>
66 #include <netinet/in_systm.h>
67 #include <netinet/ip.h>
68 #include <netinet/in_pcb.h>
69 #include <netinet/in_var.h>
70 #include <netinet/ip_var.h>
72 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
75 #include <netinet6/ipsec.h>
76 #include <netproto/key/key.h>
78 #include <netproto/key/key_debug.h>
80 #define KEYDEBUG(lev,arg)
85 #include <netipsec/ipsec.h>
86 #include <netipsec/xform.h>
87 #include <netipsec/key.h>
90 #include <net/ipfw/ip_fw.h>
91 #include <net/dummynet/ip_dummynet.h>
93 #define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\
94 x, (ntohl(a.s_addr)>>24)&0xFF,\
95 (ntohl(a.s_addr)>>16)&0xFF,\
96 (ntohl(a.s_addr)>>8)&0xFF,\
97 (ntohl(a.s_addr))&0xFF, y);
101 #ifdef MBUF_STRESS_TEST
102 int mbuf_frag_size = 0;
103 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
104 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
107 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
108 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
109 static void ip_mloopback
110 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
111 static int ip_getmoptions
112 (struct sockopt *, struct ip_moptions *);
113 static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
114 static int ip_setmoptions
115 (struct sockopt *, struct ip_moptions **);
117 int ip_optcopy(struct ip *, struct ip *);
118 extern int (*fr_checkp) (struct ip *, int, struct ifnet *, int, struct mbuf **);
121 extern struct protosw inetsw[];
124 * IP output. The packet in mbuf chain m contains a skeletal IP
125 * header (with len, off, ttl, proto, tos, src, dst).
126 * The mbuf chain containing the packet will be freed.
127 * The mbuf opt, if present, will not be freed.
130 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
131 int flags, struct ip_moptions *imo, struct inpcb *inp)
134 struct ifnet *ifp = NULL; /* keep compiler happy */
136 int hlen = sizeof (struct ip);
137 int len, off, error = 0;
138 struct sockaddr_in *dst = NULL; /* keep compiler happy */
139 struct in_ifaddr *ia = NULL;
140 int isbroadcast, sw_csum;
141 struct in_addr pkt_dst;
143 struct route iproute;
144 struct secpolicy *sp = NULL;
145 struct socket *so = inp ? inp->inp_socket : NULL;
148 struct route iproute;
150 struct secpolicy *sp = NULL;
151 struct tdb_ident *tdbi;
153 #endif /* FAST_IPSEC */
154 struct ip_fw_args args;
155 int src_was_INADDR_ANY = 0; /* as the name says... */
159 args.next_hop = NULL;
160 args.divert_rule = 0; /* divert cookie */
162 /* Grab info from MT_TAG mbufs prepended to the chain. */
163 for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) {
164 switch(m0->_m_tag_id) {
166 printf("ip_output: unrecognised MT_TAG tag %d\n",
170 case PACKET_TAG_DUMMYNET:
172 * the packet was already tagged, so part of the
173 * processing was already done, and we need to go down.
174 * Get parameters from the header.
176 args.rule = ((struct dn_pkt *)m0)->rule;
178 ro = & ( ((struct dn_pkt *)m0)->ro ) ;
180 dst = ((struct dn_pkt *)m0)->dn_dst ;
181 ifp = ((struct dn_pkt *)m0)->ifp ;
182 flags = ((struct dn_pkt *)m0)->flags ;
185 case PACKET_TAG_DIVERT:
186 args.divert_rule = (int)m0->m_data & 0xffff;
189 case PACKET_TAG_IPFORWARD:
190 args.next_hop = (struct sockaddr_in *)m0->m_data;
196 KASSERT(!m || (m->m_flags & M_PKTHDR) != 0, ("ip_output: no HDR"));
198 KASSERT(ro != NULL, ("ip_output: no route, proto %d",
199 mtod(m, struct ip *)->ip_p));
202 if (args.rule != NULL) { /* dummynet already saw us */
203 ip = mtod(m, struct ip *);
204 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
206 ia = ifatoia(ro->ro_rt->rt_ifa);
212 m = ip_insertoptions(m, opt, &len);
216 ip = mtod(m, struct ip *);
217 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
222 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
223 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
226 ip->ip_id = ip_randomid();
228 ip->ip_id = htons(ip_id++);
230 ipstat.ips_localout++;
232 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
238 bzero(ro, sizeof (*ro));
240 #endif /* FAST_IPSEC */
241 dst = (struct sockaddr_in *)&ro->ro_dst;
243 * If there is a cached route,
244 * check that it is to the same destination
245 * and is still up. If not, free it and try again.
246 * The address family should also be checked in case of sharing the
249 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
250 dst->sin_family != AF_INET ||
251 dst->sin_addr.s_addr != pkt_dst.s_addr)) {
253 ro->ro_rt = (struct rtentry *)0;
255 if (ro->ro_rt == 0) {
256 bzero(dst, sizeof(*dst));
257 dst->sin_family = AF_INET;
258 dst->sin_len = sizeof(*dst);
259 dst->sin_addr = pkt_dst;
262 * If routing to interface only,
263 * short circuit routing lookup.
265 if (flags & IP_ROUTETOIF) {
266 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
267 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
268 ipstat.ips_noroute++;
274 isbroadcast = in_broadcast(dst->sin_addr, ifp);
275 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
276 imo != NULL && imo->imo_multicast_ifp != NULL) {
278 * Bypass the normal routing lookup for multicast
279 * packets if the interface is specified.
281 ifp = imo->imo_multicast_ifp;
283 isbroadcast = 0; /* fool gcc */
286 * If this is the case, we probably don't want to allocate
287 * a protocol-cloned route since we didn't get one from the
288 * ULP. This lets TCP do its thing, while not burdening
289 * forwarding or ICMP with the overhead of cloning a route.
290 * Of course, we still want to do any cloning requested by
291 * the link layer, as this is probably required in all cases
292 * for correct operation (as it is for ARP).
295 rtalloc_ign(ro, RTF_PRCLONING);
296 if (ro->ro_rt == 0) {
297 ipstat.ips_noroute++;
298 error = EHOSTUNREACH;
301 ia = ifatoia(ro->ro_rt->rt_ifa);
302 ifp = ro->ro_rt->rt_ifp;
304 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
305 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
306 if (ro->ro_rt->rt_flags & RTF_HOST)
307 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
309 isbroadcast = in_broadcast(dst->sin_addr, ifp);
311 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
312 struct in_multi *inm;
314 m->m_flags |= M_MCAST;
316 * IP destination address is multicast. Make sure "dst"
317 * still points to the address in "ro". (It may have been
318 * changed to point to a gateway address, above.)
320 dst = (struct sockaddr_in *)&ro->ro_dst;
322 * See if the caller provided any multicast options
325 ip->ip_ttl = imo->imo_multicast_ttl;
326 if (imo->imo_multicast_vif != -1)
329 ip_mcast_src(imo->imo_multicast_vif) :
332 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
334 * Confirm that the outgoing interface supports multicast.
336 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
337 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
338 ipstat.ips_noroute++;
344 * If source address not specified yet, use address
345 * of outgoing interface.
347 if (ip->ip_src.s_addr == INADDR_ANY) {
348 /* Interface may have no addresses. */
350 ip->ip_src = IA_SIN(ia)->sin_addr;
353 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
356 * delayed checksums are not currently
357 * compatible with IP multicast routing
359 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
361 m->m_pkthdr.csum_flags &=
365 IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
367 (imo == NULL || imo->imo_multicast_loop)) {
369 * If we belong to the destination multicast group
370 * on the outgoing interface, and the caller did not
371 * forbid loopback, loop back a copy.
373 ip_mloopback(ifp, m, dst, hlen);
377 * If we are acting as a multicast router, perform
378 * multicast forwarding as if the packet had just
379 * arrived on the interface to which we are about
380 * to send. The multicast forwarding function
381 * recursively calls this function, using the
382 * IP_FORWARDING flag to prevent infinite recursion.
384 * Multicasts that are looped back by ip_mloopback(),
385 * above, will be forwarded by the ip_input() routine,
388 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
390 * If rsvp daemon is not running, do not
391 * set ip_moptions. This ensures that the packet
392 * is multicast and not just sent down one link
393 * as prescribed by rsvpd.
398 ip_mforward(ip, ifp, m, imo) != 0) {
406 * Multicasts with a time-to-live of zero may be looped-
407 * back, above, but must not be transmitted on a network.
408 * Also, multicasts addressed to the loopback interface
409 * are not sent -- the above call to ip_mloopback() will
410 * loop back a copy if this host actually belongs to the
411 * destination group on the loopback interface.
413 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
422 * If the source address is not specified yet, use the address
423 * of the outoing interface. In case, keep note we did that, so
424 * if the the firewall changes the next-hop causing the output
425 * interface to change, we can fix that.
427 if (ip->ip_src.s_addr == INADDR_ANY) {
428 /* Interface may have no addresses. */
430 ip->ip_src = IA_SIN(ia)->sin_addr;
431 src_was_INADDR_ANY = 1;
436 * Verify that we have any chance at all of being able to queue
437 * the packet or packet fragments
439 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
440 ifp->if_snd.ifq_maxlen) {
442 ipstat.ips_odropped++;
447 * Look for broadcast address and
448 * verify user is allowed to send
452 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
453 error = EADDRNOTAVAIL;
456 if ((flags & IP_ALLOWBROADCAST) == 0) {
460 /* don't allow broadcast messages to be fragmented */
461 if (ip->ip_len > ifp->if_mtu) {
465 m->m_flags |= M_BCAST;
467 m->m_flags &= ~M_BCAST;
472 /* get SP for this packet */
474 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
476 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
479 ipsecstat.out_inval++;
486 switch (sp->policy) {
487 case IPSEC_POLICY_DISCARD:
489 * This packet is just discarded.
491 ipsecstat.out_polvio++;
494 case IPSEC_POLICY_BYPASS:
495 case IPSEC_POLICY_NONE:
496 /* no need to do IPsec. */
499 case IPSEC_POLICY_IPSEC:
500 if (sp->req == NULL) {
501 /* acquire a policy */
502 error = key_spdacquire(sp);
507 case IPSEC_POLICY_ENTRUST:
509 printf("ip_output: Invalid policy found. %d\n", sp->policy);
512 struct ipsec_output_state state;
513 bzero(&state, sizeof(state));
515 if (flags & IP_ROUTETOIF) {
517 bzero(&iproute, sizeof(iproute));
520 state.dst = (struct sockaddr *)dst;
526 * delayed checksums are not currently compatible with IPsec
528 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
530 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
533 ip->ip_len = htons(ip->ip_len);
534 ip->ip_off = htons(ip->ip_off);
536 error = ipsec4_output(&state, sp, flags);
539 if (flags & IP_ROUTETOIF) {
541 * if we have tunnel mode SA, we may need to ignore
544 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
545 flags &= ~IP_ROUTETOIF;
550 dst = (struct sockaddr_in *)state.dst;
552 /* mbuf is already reclaimed in ipsec4_output. */
562 printf("ip4_output (ipsec): error code %d\n", error);
565 /* don't show these error codes to the user */
573 /* be sure to update variables that are affected by ipsec4_output() */
574 ip = mtod(m, struct ip *);
576 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
578 hlen = ip->ip_hl << 2;
580 if (ro->ro_rt == NULL) {
581 if ((flags & IP_ROUTETOIF) == 0) {
583 "can't update route after IPsec processing\n");
584 error = EHOSTUNREACH; /*XXX*/
588 ia = ifatoia(ro->ro_rt->rt_ifa);
589 ifp = ro->ro_rt->rt_ifp;
592 /* make it flipped, again. */
593 ip->ip_len = ntohs(ip->ip_len);
594 ip->ip_off = ntohs(ip->ip_off);
599 * Check the security policy (SP) for the packet and, if
600 * required, do IPsec-related processing. There are two
601 * cases here; the first time a packet is sent through
602 * it will be untagged and handled by ipsec4_checkpolicy.
603 * If the packet is resubmitted to ip_output (e.g. after
604 * AH, ESP, etc. processing), there will be a tag to bypass
605 * the lookup and related policy checking.
607 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
610 tdbi = (struct tdb_ident *)(mtag + 1);
611 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
613 error = -EINVAL; /* force silent drop */
614 m_tag_delete(m, mtag);
616 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
620 * There are four return cases:
621 * sp != NULL apply IPsec policy
622 * sp == NULL, error == 0 no IPsec handling needed
623 * sp == NULL, error == -EINVAL discard packet w/o error
624 * sp == NULL, error != 0 discard packet, report error
627 /* Loop detection, check if ipsec processing already done */
628 KASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
629 for (mtag = m_tag_first(m); mtag != NULL;
630 mtag = m_tag_next(m, mtag)) {
631 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
633 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
634 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
637 * Check if policy has an SA associated with it.
638 * This can happen when an SP has yet to acquire
639 * an SA; e.g. on first reference. If it occurs,
640 * then we let ipsec4_process_packet do its thing.
642 if (sp->req->sav == NULL)
644 tdbi = (struct tdb_ident *)(mtag + 1);
645 if (tdbi->spi == sp->req->sav->spi &&
646 tdbi->proto == sp->req->sav->sah->saidx.proto &&
647 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
648 sizeof (union sockaddr_union)) == 0) {
650 * No IPsec processing is needed, free
653 * NB: null pointer to avoid free at
656 KEY_FREESP(&sp), sp = NULL;
663 * Do delayed checksums now because we send before
664 * this is done in the normal processing path.
666 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
668 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
671 ip->ip_len = htons(ip->ip_len);
672 ip->ip_off = htons(ip->ip_off);
674 /* NB: callee frees mbuf */
675 error = ipsec4_process_packet(m, sp->req, flags, 0);
677 * Preserve KAME behaviour: ENOENT can be returned
678 * when an SA acquire is in progress. Don't propagate
679 * this to user-level; it confuses applications.
681 * XXX this will go away when the SADB is redone.
692 * Hack: -EINVAL is used to signal that a packet
693 * should be silently discarded. This is typically
694 * because we asked key management for an SA and
695 * it was delayed (e.g. kicked up to IKE).
697 if (error == -EINVAL)
701 /* No IPsec processing for this packet. */
705 * If deferred crypto processing is needed, check that
706 * the interface supports it.
708 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
709 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) {
710 /* notify IPsec to do its own crypto */
711 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
712 error = EHOSTUNREACH;
718 #endif /* FAST_IPSEC */
721 * - Xlate: translate packet's addr/port (NAT).
722 * - Firewall: deny/allow/etc.
723 * - Wrap: fake packet's addr/port <unimpl.>
724 * - Encapsulate: put it in another IP and send out. <unimp.>
728 * Run through list of hooks for output packets.
730 if (pfil_has_hooks(&inet_pfil_hook)) {
731 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT);
732 if (error != 0 || m == NULL)
734 ip = mtod(m, struct ip *);
738 * Check with the firewall...
739 * but not if we are already being fwd'd from a firewall.
741 if (fw_enable && IPFW_LOADED && !args.next_hop) {
742 struct sockaddr_in *old = dst;
747 off = ip_fw_chk_ptr(&args);
752 * On return we must do the following:
753 * m == NULL -> drop the pkt (old interface, deprecated)
754 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface)
755 * 1<=off<= 0xffff -> DIVERT
756 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe
757 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet
758 * dst != old -> IPFIREWALL_FORWARD
759 * off==0, dst==old -> accept
760 * If some of the above modules are not compiled in, then
761 * we should't have to check the corresponding condition
762 * (because the ipfw control socket should not accept
763 * unsupported rules), but better play safe and drop
764 * packets in case of doubt.
766 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
772 ip = mtod(m, struct ip *);
773 if (off == 0 && dst == old) /* common case */
775 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) {
777 * pass the pkt to dummynet. Need to include
778 * pipe number, m, ifp, ro, dst because these are
779 * not recomputed in the next pass.
780 * All other parameters have been already used and
781 * so they are not needed anymore.
782 * XXX note: if the ifp or ro entry are deleted
783 * while a pkt is in dummynet, we are in trouble!
789 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT,
794 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) {
795 struct mbuf *clone = NULL;
797 /* Clone packet if we're doing a 'tee' */
798 if ((off & IP_FW_PORT_TEE_FLAG) != 0)
799 clone = m_dup(m, MB_DONTWAIT);
803 * delayed checksums are not currently compatible
804 * with divert sockets.
806 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
808 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
811 /* Restore packet header fields to original values */
812 ip->ip_len = htons(ip->ip_len);
813 ip->ip_off = htons(ip->ip_off);
815 /* Deliver packet to divert input routine */
816 divert_packet(m, 0, off & 0xffff, args.divert_rule);
818 /* If 'tee', continue with original packet */
821 ip = mtod(m, struct ip *);
828 /* IPFIREWALL_FORWARD */
830 * Check dst to make sure it is directly reachable on the
831 * interface we previously thought it was.
832 * If it isn't (which may be likely in some situations) we have
833 * to re-route it (ie, find a route for the next-hop and the
834 * associated interface) and set them here. This is nested
835 * forwarding which in most cases is undesirable, except where
836 * such control is nigh impossible. So we do it here.
839 if (off == 0 && old != dst) { /* FORWARD, dst has changed */
842 * XXX To improve readability, this block should be
843 * changed into a function call as below:
845 error = ip_ipforward(&m, &dst, &ifp);
848 if (m == NULL) /* ip_input consumed the mbuf */
851 struct in_ifaddr *ia;
854 * XXX sro_fwd below is static, and a pointer
855 * to it gets passed to routines downstream.
856 * This could have surprisingly bad results in
857 * practice, because its content is overwritten
858 * by subsequent packets.
860 /* There must be a better way to do this next line... */
861 static struct route sro_fwd;
862 struct route *ro_fwd = &sro_fwd;
865 print_ip("IPFIREWALL_FORWARD: New dst ip: ",
866 dst->sin_addr, "\n");
870 * We need to figure out if we have been forwarded
871 * to a local socket. If so, then we should somehow
872 * "loop back" to ip_input, and get directed to the
873 * PCB as if we had received this packet. This is
874 * because it may be dificult to identify the packets
875 * you want to forward until they are being output
876 * and have selected an interface. (e.g. locally
877 * initiated packets) If we used the loopback inteface,
878 * we would not be able to control what happens
879 * as the packet runs through ip_input() as
880 * it is done through a ISR.
883 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) {
885 * If the addr to forward to is one
886 * of ours, we pretend to
887 * be the destination for this packet.
889 if (IA_SIN(ia)->sin_addr.s_addr ==
890 dst->sin_addr.s_addr)
893 if (ia) { /* tell ip_input "dont filter" */
896 tag.mh_type = MT_TAG;
897 tag.mh_flags = PACKET_TAG_IPFORWARD;
898 tag.mh_data = (caddr_t)args.next_hop;
901 if (m->m_pkthdr.rcvif == NULL)
902 m->m_pkthdr.rcvif = ifunit("lo0");
903 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
904 m->m_pkthdr.csum_flags |=
905 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
906 m0->m_pkthdr.csum_data = 0xffff;
908 m->m_pkthdr.csum_flags |=
909 CSUM_IP_CHECKED | CSUM_IP_VALID;
910 ip->ip_len = htons(ip->ip_len);
911 ip->ip_off = htons(ip->ip_off);
912 ip_input((struct mbuf *)&tag);
915 /* Some of the logic for this was
918 * This rewrites the cached route in a local PCB.
919 * Is this what we want to do?
921 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst));
924 rtalloc_ign(ro_fwd, RTF_PRCLONING);
926 if (ro_fwd->ro_rt == 0) {
927 ipstat.ips_noroute++;
928 error = EHOSTUNREACH;
932 ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
933 ifp = ro_fwd->ro_rt->rt_ifp;
934 ro_fwd->ro_rt->rt_use++;
935 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
936 dst = (struct sockaddr_in *)
937 ro_fwd->ro_rt->rt_gateway;
938 if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
940 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
942 isbroadcast = in_broadcast(dst->sin_addr, ifp);
945 ro->ro_rt = ro_fwd->ro_rt;
946 dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
948 #endif /* ... block to be put into a function */
950 * If we added a default src ip earlier,
951 * which would have been gotten from the-then
952 * interface, do it again, from the new one.
954 if (src_was_INADDR_ANY)
955 ip->ip_src = IA_SIN(ia)->sin_addr;
960 * if we get here, none of the above matches, and
961 * we have to drop the pkt
964 error = EACCES; /* not sure this is the right error msg */
969 /* 127/8 must not appear on wire - RFC1122. */
970 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
971 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
972 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
973 ipstat.ips_badaddr++;
974 error = EADDRNOTAVAIL;
979 m->m_pkthdr.csum_flags |= CSUM_IP;
980 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
981 if (sw_csum & CSUM_DELAY_DATA) {
983 sw_csum &= ~CSUM_DELAY_DATA;
985 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
988 * If small enough for interface, or the interface will take
989 * care of the fragmentation for us, can just send directly.
991 if (ip->ip_len <= ifp->if_mtu || ifp->if_hwassist & CSUM_FRAGMENT) {
992 ip->ip_len = htons(ip->ip_len);
993 ip->ip_off = htons(ip->ip_off);
995 if (sw_csum & CSUM_DELAY_IP) {
996 if (ip->ip_vhl == IP_VHL_BORING) {
997 ip->ip_sum = in_cksum_hdr(ip);
999 ip->ip_sum = in_cksum(m, hlen);
1003 /* Record statistics for this interface address. */
1004 if (!(flags & IP_FORWARDING) && ia) {
1005 ia->ia_ifa.if_opackets++;
1006 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1010 /* clean ipsec history once it goes out of the node */
1014 #ifdef MBUF_STRESS_TEST
1015 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) {
1016 struct mbuf *m1, *m2;
1019 tmp = length = m->m_pkthdr.len;
1021 while ((length -= mbuf_frag_size) >= 1) {
1022 m1 = m_split(m, length, MB_DONTWAIT);
1025 m1->m_flags &= ~M_PKTHDR;
1027 while (m2->m_next != NULL)
1031 m->m_pkthdr.len = tmp;
1034 error = (*ifp->if_output)(ifp, m,
1035 (struct sockaddr *)dst, ro->ro_rt);
1039 if (ip->ip_off & IP_DF) {
1042 * This case can happen if the user changed the MTU
1043 * of an interface after enabling IP on it. Because
1044 * most netifs don't keep track of routes pointing to
1045 * them, there is no way for one to update all its
1046 * routes when the MTU is changed.
1048 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
1049 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
1050 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
1051 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
1053 ipstat.ips_cantfrag++;
1058 * Too large for interface; fragment if possible. If successful,
1059 * on return, m will point to a list of packets to be sent.
1061 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
1068 /* clean ipsec history once it goes out of the node */
1072 /* Record statistics for this interface address. */
1074 ia->ia_ifa.if_opackets++;
1075 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1078 error = (*ifp->if_output)(ifp, m,
1079 (struct sockaddr *)dst, ro->ro_rt);
1085 ipstat.ips_fragmented++;
1089 if (ro == &iproute && ro->ro_rt) {
1094 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1095 printf("DP ip_output call free SP:%p\n", sp));
1100 if (ro == &iproute && ro->ro_rt) {
1114 * Create a chain of fragments which fit the given mtu. m_frag points to the
1115 * mbuf to be fragmented; on return it points to the chain with the fragments.
1116 * Return 0 if no error. If error, m_frag may contain a partially built
1117 * chain of fragments that should be freed by the caller.
1119 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
1120 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
1123 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
1124 u_long if_hwassist_flags, int sw_csum)
1127 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1128 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
1130 struct mbuf *m0 = *m_frag; /* the original packet */
1132 struct mbuf **mnext;
1135 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
1136 ipstat.ips_cantfrag++;
1141 * Must be able to put at least 8 bytes per fragment.
1147 * If the interface will not calculate checksums on
1148 * fragmented packets, then do it here.
1150 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
1151 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
1152 in_delayed_cksum(m0);
1153 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1156 if (len > PAGE_SIZE) {
1158 * Fragment large datagrams such that each segment
1159 * contains a multiple of PAGE_SIZE amount of data,
1160 * plus headers. This enables a receiver to perform
1161 * page-flipping zero-copy optimizations.
1163 * XXX When does this help given that sender and receiver
1164 * could have different page sizes, and also mtu could
1165 * be less than the receiver's page size ?
1170 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
1174 * firstlen (off - hlen) must be aligned on an
1178 goto smart_frag_failure;
1179 off = ((off - hlen) & ~7) + hlen;
1180 newlen = (~PAGE_MASK) & mtu;
1181 if ((newlen + sizeof (struct ip)) > mtu) {
1182 /* we failed, go back the default */
1193 firstlen = off - hlen;
1194 mnext = &m0->m_nextpkt; /* pointer to next packet */
1197 * Loop through length of segment after first fragment,
1198 * make new header and copy data of each part and link onto chain.
1199 * Here, m0 is the original packet, m is the fragment being created.
1200 * The fragments are linked off the m_nextpkt of the original
1201 * packet, which after processing serves as the first fragment.
1203 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
1204 struct ip *mhip; /* ip header on the fragment */
1206 int mhlen = sizeof (struct ip);
1208 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1211 ipstat.ips_odropped++;
1214 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
1216 * In the first mbuf, leave room for the link header, then
1217 * copy the original IP header including options. The payload
1218 * goes into an additional mbuf chain returned by m_copy().
1220 m->m_data += max_linkhdr;
1221 mhip = mtod(m, struct ip *);
1223 if (hlen > sizeof (struct ip)) {
1224 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1225 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
1228 /* XXX do we need to add ip->ip_off below ? */
1229 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
1230 if (off + len >= ip->ip_len) { /* last fragment */
1231 len = ip->ip_len - off;
1232 m->m_flags |= M_LASTFRAG;
1234 mhip->ip_off |= IP_MF;
1235 mhip->ip_len = htons((u_short)(len + mhlen));
1236 m->m_next = m_copy(m0, off, len);
1237 if (m->m_next == 0) { /* copy failed */
1239 error = ENOBUFS; /* ??? */
1240 ipstat.ips_odropped++;
1243 m->m_pkthdr.len = mhlen + len;
1244 m->m_pkthdr.rcvif = (struct ifnet *)0;
1245 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1246 mhip->ip_off = htons(mhip->ip_off);
1248 if (sw_csum & CSUM_DELAY_IP)
1249 mhip->ip_sum = in_cksum(m, mhlen);
1251 mnext = &m->m_nextpkt;
1253 ipstat.ips_ofragments += nfrags;
1255 /* set first marker for fragment chain */
1256 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1257 m0->m_pkthdr.csum_data = nfrags;
1260 * Update first fragment by trimming what's been copied out
1261 * and updating header.
1263 m_adj(m0, hlen + firstlen - ip->ip_len);
1264 m0->m_pkthdr.len = hlen + firstlen;
1265 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1266 ip->ip_off |= IP_MF;
1267 ip->ip_off = htons(ip->ip_off);
1269 if (sw_csum & CSUM_DELAY_IP)
1270 ip->ip_sum = in_cksum(m0, hlen);
1278 in_delayed_cksum(struct mbuf *m)
1281 u_short csum, offset;
1283 ip = mtod(m, struct ip *);
1284 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1285 csum = in_cksum_skip(m, ip->ip_len, offset);
1286 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1288 offset += m->m_pkthdr.csum_data; /* checksum offset */
1290 if (offset + sizeof(u_short) > m->m_len) {
1291 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1292 m->m_len, offset, ip->ip_p);
1295 * this shouldn't happen, but if it does, the
1296 * correct behavior may be to insert the checksum
1297 * in the existing chain instead of rearranging it.
1299 m = m_pullup(m, offset + sizeof(u_short));
1301 *(u_short *)(m->m_data + offset) = csum;
1305 * Insert IP options into preformed packet.
1306 * Adjust IP destination as required for IP source routing,
1307 * as indicated by a non-zero in_addr at the start of the options.
1309 * XXX This routine assumes that the packet has no options in place.
1311 static struct mbuf *
1312 ip_insertoptions(m, opt, phlen)
1317 struct ipoption *p = mtod(opt, struct ipoption *);
1319 struct ip *ip = mtod(m, struct ip *);
1322 optlen = opt->m_len - sizeof(p->ipopt_dst);
1323 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
1325 return (m); /* XXX should fail */
1327 if (p->ipopt_dst.s_addr)
1328 ip->ip_dst = p->ipopt_dst;
1329 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1330 MGETHDR(n, MB_DONTWAIT, MT_HEADER);
1335 n->m_pkthdr.rcvif = (struct ifnet *)0;
1336 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1337 m->m_len -= sizeof(struct ip);
1338 m->m_data += sizeof(struct ip);
1341 m->m_len = optlen + sizeof(struct ip);
1342 m->m_data += max_linkhdr;
1343 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip));
1345 m->m_data -= optlen;
1347 m->m_pkthdr.len += optlen;
1348 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1350 ip = mtod(m, struct ip *);
1351 bcopy(p->ipopt_list, ip + 1, optlen);
1352 *phlen = sizeof(struct ip) + optlen;
1353 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1354 ip->ip_len += optlen;
1359 * Copy options from ip to jp,
1360 * omitting those not copied during fragmentation.
1367 int opt, optlen, cnt;
1369 cp = (u_char *)(ip + 1);
1370 dp = (u_char *)(jp + 1);
1371 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1372 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1374 if (opt == IPOPT_EOL)
1376 if (opt == IPOPT_NOP) {
1377 /* Preserve for IP mcast tunnel's LSRR alignment. */
1383 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp),
1384 ("ip_optcopy: malformed ipv4 option"));
1385 optlen = cp[IPOPT_OLEN];
1386 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt,
1387 ("ip_optcopy: malformed ipv4 option"));
1389 /* bogus lengths should have been caught by ip_dooptions */
1392 if (IPOPT_COPIED(opt)) {
1393 bcopy(cp, dp, optlen);
1397 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1403 * IP socket option processing.
1406 ip_ctloutput(so, sopt)
1408 struct sockopt *sopt;
1410 struct inpcb *inp = sotoinpcb(so);
1414 if (sopt->sopt_level != IPPROTO_IP) {
1418 switch (sopt->sopt_dir) {
1420 switch (sopt->sopt_name) {
1427 if (sopt->sopt_valsize > MLEN) {
1431 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1436 m->m_len = sopt->sopt_valsize;
1437 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1440 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1447 case IP_RECVRETOPTS:
1448 case IP_RECVDSTADDR:
1451 error = sooptcopyin(sopt, &optval, sizeof optval,
1456 switch (sopt->sopt_name) {
1458 inp->inp_ip_tos = optval;
1462 inp->inp_ip_ttl = optval;
1464 #define OPTSET(bit) \
1466 inp->inp_flags |= bit; \
1468 inp->inp_flags &= ~bit;
1471 OPTSET(INP_RECVOPTS);
1474 case IP_RECVRETOPTS:
1475 OPTSET(INP_RECVRETOPTS);
1478 case IP_RECVDSTADDR:
1479 OPTSET(INP_RECVDSTADDR);
1493 case IP_MULTICAST_IF:
1494 case IP_MULTICAST_VIF:
1495 case IP_MULTICAST_TTL:
1496 case IP_MULTICAST_LOOP:
1497 case IP_ADD_MEMBERSHIP:
1498 case IP_DROP_MEMBERSHIP:
1499 error = ip_setmoptions(sopt, &inp->inp_moptions);
1503 error = sooptcopyin(sopt, &optval, sizeof optval,
1509 case IP_PORTRANGE_DEFAULT:
1510 inp->inp_flags &= ~(INP_LOWPORT);
1511 inp->inp_flags &= ~(INP_HIGHPORT);
1514 case IP_PORTRANGE_HIGH:
1515 inp->inp_flags &= ~(INP_LOWPORT);
1516 inp->inp_flags |= INP_HIGHPORT;
1519 case IP_PORTRANGE_LOW:
1520 inp->inp_flags &= ~(INP_HIGHPORT);
1521 inp->inp_flags |= INP_LOWPORT;
1530 #if defined(IPSEC) || defined(FAST_IPSEC)
1531 case IP_IPSEC_POLICY:
1539 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1541 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1543 priv = (sopt->sopt_td != NULL &&
1544 suser(sopt->sopt_td) != 0) ? 0 : 1;
1545 req = mtod(m, caddr_t);
1547 optname = sopt->sopt_name;
1548 error = ipsec4_set_policy(inp, optname, req, len, priv);
1555 error = ENOPROTOOPT;
1561 switch (sopt->sopt_name) {
1564 if (inp->inp_options)
1565 error = sooptcopyout(sopt,
1566 mtod(inp->inp_options,
1568 inp->inp_options->m_len);
1570 sopt->sopt_valsize = 0;
1576 case IP_RECVRETOPTS:
1577 case IP_RECVDSTADDR:
1581 switch (sopt->sopt_name) {
1584 optval = inp->inp_ip_tos;
1588 optval = inp->inp_ip_ttl;
1591 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1594 optval = OPTBIT(INP_RECVOPTS);
1597 case IP_RECVRETOPTS:
1598 optval = OPTBIT(INP_RECVRETOPTS);
1601 case IP_RECVDSTADDR:
1602 optval = OPTBIT(INP_RECVDSTADDR);
1606 optval = OPTBIT(INP_RECVIF);
1610 if (inp->inp_flags & INP_HIGHPORT)
1611 optval = IP_PORTRANGE_HIGH;
1612 else if (inp->inp_flags & INP_LOWPORT)
1613 optval = IP_PORTRANGE_LOW;
1619 optval = OPTBIT(INP_FAITH);
1622 error = sooptcopyout(sopt, &optval, sizeof optval);
1625 case IP_MULTICAST_IF:
1626 case IP_MULTICAST_VIF:
1627 case IP_MULTICAST_TTL:
1628 case IP_MULTICAST_LOOP:
1629 case IP_ADD_MEMBERSHIP:
1630 case IP_DROP_MEMBERSHIP:
1631 error = ip_getmoptions(sopt, inp->inp_moptions);
1634 #if defined(IPSEC) || defined(FAST_IPSEC)
1635 case IP_IPSEC_POLICY:
1637 struct mbuf *m = NULL;
1642 req = mtod(m, caddr_t);
1645 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m);
1647 error = soopt_mcopyout(sopt, m); /* XXX */
1655 error = ENOPROTOOPT;
1664 * Set up IP options in pcb for insertion in output packets.
1665 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1666 * with destination address if source routed.
1669 ip_pcbopts(optname, pcbopt, m)
1671 struct mbuf **pcbopt;
1678 /* turn off any old options */
1680 (void)m_free(*pcbopt);
1682 if (m == (struct mbuf *)0 || m->m_len == 0) {
1684 * Only turning off any previous options.
1691 if (m->m_len % sizeof(int32_t))
1694 * IP first-hop destination address will be stored before
1695 * actual options; move other options back
1696 * and clear it when none present.
1698 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1701 m->m_len += sizeof(struct in_addr);
1702 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1703 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
1704 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1706 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1707 opt = cp[IPOPT_OPTVAL];
1708 if (opt == IPOPT_EOL)
1710 if (opt == IPOPT_NOP)
1713 if (cnt < IPOPT_OLEN + sizeof(*cp))
1715 optlen = cp[IPOPT_OLEN];
1716 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1727 * user process specifies route as:
1729 * D must be our final destination (but we can't
1730 * check that since we may not have connected yet).
1731 * A is first hop destination, which doesn't appear in
1732 * actual IP option, but is stored before the options.
1734 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1736 m->m_len -= sizeof(struct in_addr);
1737 cnt -= sizeof(struct in_addr);
1738 optlen -= sizeof(struct in_addr);
1739 cp[IPOPT_OLEN] = optlen;
1741 * Move first hop before start of options.
1743 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1744 sizeof(struct in_addr));
1746 * Then copy rest of options back
1747 * to close up the deleted entry.
1749 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1750 sizeof(struct in_addr)),
1751 (caddr_t)&cp[IPOPT_OFFSET+1],
1752 (unsigned)cnt + sizeof(struct in_addr));
1756 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1768 * The whole multicast option thing needs to be re-thought.
1769 * Several of these options are equally applicable to non-multicast
1770 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1771 * standard option (IP_TTL).
1775 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1777 static struct ifnet *
1778 ip_multicast_if(a, ifindexp)
1787 if (ntohl(a->s_addr) >> 24 == 0) {
1788 ifindex = ntohl(a->s_addr) & 0xffffff;
1789 if (ifindex < 0 || if_index < ifindex)
1791 ifp = ifindex2ifnet[ifindex];
1793 *ifindexp = ifindex;
1795 INADDR_TO_IFP(*a, ifp);
1801 * Set the IP multicast options in response to user setsockopt().
1804 ip_setmoptions(sopt, imop)
1805 struct sockopt *sopt;
1806 struct ip_moptions **imop;
1810 struct in_addr addr;
1811 struct ip_mreq mreq;
1813 struct ip_moptions *imo = *imop;
1815 struct sockaddr_in *dst;
1821 * No multicast option buffer attached to the pcb;
1822 * allocate one and initialize to default values.
1824 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
1830 imo->imo_multicast_ifp = NULL;
1831 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1832 imo->imo_multicast_vif = -1;
1833 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1834 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1835 imo->imo_num_memberships = 0;
1838 switch (sopt->sopt_name) {
1839 /* store an index number for the vif you wanna use in the send */
1840 case IP_MULTICAST_VIF:
1841 if (legal_vif_num == 0) {
1845 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
1848 if (!legal_vif_num(i) && (i != -1)) {
1852 imo->imo_multicast_vif = i;
1855 case IP_MULTICAST_IF:
1857 * Select the interface for outgoing multicast packets.
1859 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr);
1863 * INADDR_ANY is used to remove a previous selection.
1864 * When no interface is selected, a default one is
1865 * chosen every time a multicast packet is sent.
1867 if (addr.s_addr == INADDR_ANY) {
1868 imo->imo_multicast_ifp = NULL;
1872 * The selected interface is identified by its local
1873 * IP address. Find the interface and confirm that
1874 * it supports multicasting.
1877 ifp = ip_multicast_if(&addr, &ifindex);
1878 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1880 error = EADDRNOTAVAIL;
1883 imo->imo_multicast_ifp = ifp;
1885 imo->imo_multicast_addr = addr;
1887 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1891 case IP_MULTICAST_TTL:
1893 * Set the IP time-to-live for outgoing multicast packets.
1894 * The original multicast API required a char argument,
1895 * which is inconsistent with the rest of the socket API.
1896 * We allow either a char or an int.
1898 if (sopt->sopt_valsize == 1) {
1900 error = sooptcopyin(sopt, &ttl, 1, 1);
1903 imo->imo_multicast_ttl = ttl;
1906 error = sooptcopyin(sopt, &ttl, sizeof ttl,
1913 imo->imo_multicast_ttl = ttl;
1917 case IP_MULTICAST_LOOP:
1919 * Set the loopback flag for outgoing multicast packets.
1920 * Must be zero or one. The original multicast API required a
1921 * char argument, which is inconsistent with the rest
1922 * of the socket API. We allow either a char or an int.
1924 if (sopt->sopt_valsize == 1) {
1926 error = sooptcopyin(sopt, &loop, 1, 1);
1929 imo->imo_multicast_loop = !!loop;
1932 error = sooptcopyin(sopt, &loop, sizeof loop,
1936 imo->imo_multicast_loop = !!loop;
1940 case IP_ADD_MEMBERSHIP:
1942 * Add a multicast group membership.
1943 * Group must be a valid IP multicast address.
1945 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1949 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1955 * If no interface address was provided, use the interface of
1956 * the route to the given multicast address.
1958 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1959 bzero((caddr_t)&ro, sizeof(ro));
1960 dst = (struct sockaddr_in *)&ro.ro_dst;
1961 dst->sin_len = sizeof(*dst);
1962 dst->sin_family = AF_INET;
1963 dst->sin_addr = mreq.imr_multiaddr;
1965 if (ro.ro_rt == NULL) {
1966 error = EADDRNOTAVAIL;
1970 ifp = ro.ro_rt->rt_ifp;
1974 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1978 * See if we found an interface, and confirm that it
1979 * supports multicast.
1981 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1982 error = EADDRNOTAVAIL;
1987 * See if the membership already exists or if all the
1988 * membership slots are full.
1990 for (i = 0; i < imo->imo_num_memberships; ++i) {
1991 if (imo->imo_membership[i]->inm_ifp == ifp &&
1992 imo->imo_membership[i]->inm_addr.s_addr
1993 == mreq.imr_multiaddr.s_addr)
1996 if (i < imo->imo_num_memberships) {
2001 if (i == IP_MAX_MEMBERSHIPS) {
2002 error = ETOOMANYREFS;
2007 * Everything looks good; add a new record to the multicast
2008 * address list for the given interface.
2010 if ((imo->imo_membership[i] =
2011 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
2016 ++imo->imo_num_memberships;
2020 case IP_DROP_MEMBERSHIP:
2022 * Drop a multicast group membership.
2023 * Group must be a valid IP multicast address.
2025 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
2029 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
2036 * If an interface address was specified, get a pointer
2037 * to its ifnet structure.
2039 if (mreq.imr_interface.s_addr == INADDR_ANY)
2042 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
2044 error = EADDRNOTAVAIL;
2050 * Find the membership in the membership array.
2052 for (i = 0; i < imo->imo_num_memberships; ++i) {
2054 imo->imo_membership[i]->inm_ifp == ifp) &&
2055 imo->imo_membership[i]->inm_addr.s_addr ==
2056 mreq.imr_multiaddr.s_addr)
2059 if (i == imo->imo_num_memberships) {
2060 error = EADDRNOTAVAIL;
2065 * Give up the multicast address record to which the
2066 * membership points.
2068 in_delmulti(imo->imo_membership[i]);
2070 * Remove the gap in the membership array.
2072 for (++i; i < imo->imo_num_memberships; ++i)
2073 imo->imo_membership[i-1] = imo->imo_membership[i];
2074 --imo->imo_num_memberships;
2084 * If all options have default values, no need to keep the mbuf.
2086 if (imo->imo_multicast_ifp == NULL &&
2087 imo->imo_multicast_vif == -1 &&
2088 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2089 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2090 imo->imo_num_memberships == 0) {
2091 free(*imop, M_IPMOPTS);
2099 * Return the IP multicast options in response to user getsockopt().
2102 ip_getmoptions(sopt, imo)
2103 struct sockopt *sopt;
2104 struct ip_moptions *imo;
2106 struct in_addr addr;
2107 struct in_ifaddr *ia;
2112 switch (sopt->sopt_name) {
2113 case IP_MULTICAST_VIF:
2115 optval = imo->imo_multicast_vif;
2118 error = sooptcopyout(sopt, &optval, sizeof optval);
2121 case IP_MULTICAST_IF:
2122 if (imo == NULL || imo->imo_multicast_ifp == NULL)
2123 addr.s_addr = INADDR_ANY;
2124 else if (imo->imo_multicast_addr.s_addr) {
2125 /* return the value user has set */
2126 addr = imo->imo_multicast_addr;
2128 IFP_TO_IA(imo->imo_multicast_ifp, ia);
2129 addr.s_addr = (ia == NULL) ? INADDR_ANY
2130 : IA_SIN(ia)->sin_addr.s_addr;
2132 error = sooptcopyout(sopt, &addr, sizeof addr);
2135 case IP_MULTICAST_TTL:
2137 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
2139 optval = coptval = imo->imo_multicast_ttl;
2140 if (sopt->sopt_valsize == 1)
2141 error = sooptcopyout(sopt, &coptval, 1);
2143 error = sooptcopyout(sopt, &optval, sizeof optval);
2146 case IP_MULTICAST_LOOP:
2148 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
2150 optval = coptval = imo->imo_multicast_loop;
2151 if (sopt->sopt_valsize == 1)
2152 error = sooptcopyout(sopt, &coptval, 1);
2154 error = sooptcopyout(sopt, &optval, sizeof optval);
2158 error = ENOPROTOOPT;
2165 * Discard the IP multicast options.
2168 ip_freemoptions(imo)
2169 struct ip_moptions *imo;
2174 for (i = 0; i < imo->imo_num_memberships; ++i)
2175 in_delmulti(imo->imo_membership[i]);
2176 free(imo, M_IPMOPTS);
2181 * Routine called from ip_output() to loop back a copy of an IP multicast
2182 * packet to the input queue of a specified interface. Note that this
2183 * calls the output routine of the loopback "driver", but with an interface
2184 * pointer that might NOT be a loopback interface -- evil, but easier than
2185 * replicating that code here.
2188 ip_mloopback(ifp, m, dst, hlen)
2191 struct sockaddr_in *dst;
2197 copym = m_copy(m, 0, M_COPYALL);
2198 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2199 copym = m_pullup(copym, hlen);
2200 if (copym != NULL) {
2202 * We don't bother to fragment if the IP length is greater
2203 * than the interface's MTU. Can this possibly matter?
2205 ip = mtod(copym, struct ip *);
2206 ip->ip_len = htons(ip->ip_len);
2207 ip->ip_off = htons(ip->ip_off);
2209 if (ip->ip_vhl == IP_VHL_BORING) {
2210 ip->ip_sum = in_cksum_hdr(ip);
2212 ip->ip_sum = in_cksum(copym, hlen);
2216 * It's not clear whether there are any lingering
2217 * reentrancy problems in other areas which might
2218 * be exposed by using ip_input directly (in
2219 * particular, everything which modifies the packet
2220 * in-place). Yet another option is using the
2221 * protosw directly to deliver the looped back
2222 * packet. For the moment, we'll err on the side
2223 * of safety by using if_simloop().
2226 if (dst->sin_family != AF_INET) {
2227 printf("ip_mloopback: bad address family %d\n",
2229 dst->sin_family = AF_INET;
2234 copym->m_pkthdr.rcvif = ifp;
2237 /* if the checksum hasn't been computed, mark it as valid */
2238 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2239 copym->m_pkthdr.csum_flags |=
2240 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2241 copym->m_pkthdr.csum_data = 0xffff;
2243 if_simloop(ifp, copym, dst->sin_family, 0);