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 1.55 2008/09/06 14:12:50 sephe Exp $
38 #include "opt_ipdivert.h"
39 #include "opt_ipfilter.h"
40 #include "opt_ipsec.h"
41 #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/thread2.h>
55 #include <sys/in_cksum.h>
58 #include <net/netisr.h>
60 #include <net/route.h>
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/ip.h>
65 #include <netinet/in_pcb.h>
66 #include <netinet/in_var.h>
67 #include <netinet/ip_var.h>
69 #include <netinet/ip_divert.h>
72 #include <netproto/mpls/mpls_var.h>
74 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
77 #include <netinet6/ipsec.h>
78 #include <netproto/key/key.h>
80 #include <netproto/key/key_debug.h>
82 #define KEYDEBUG(lev,arg)
87 #include <netproto/ipsec/ipsec.h>
88 #include <netproto/ipsec/xform.h>
89 #include <netproto/ipsec/key.h>
92 #include <net/ipfw/ip_fw.h>
93 #include <net/dummynet/ip_dummynet.h>
95 #define print_ip(x, a, y) kprintf("%s %d.%d.%d.%d%s",\
96 x, (ntohl(a.s_addr)>>24)&0xFF,\
97 (ntohl(a.s_addr)>>16)&0xFF,\
98 (ntohl(a.s_addr)>>8)&0xFF,\
99 (ntohl(a.s_addr))&0xFF, y);
103 #ifdef MBUF_STRESS_TEST
104 int mbuf_frag_size = 0;
105 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
106 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
109 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
110 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
111 static void ip_mloopback
112 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
113 static int ip_getmoptions
114 (struct sockopt *, struct ip_moptions *);
115 static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
116 static int ip_setmoptions
117 (struct sockopt *, struct ip_moptions **);
119 int ip_optcopy(struct ip *, struct ip *);
122 extern struct protosw inetsw[];
125 * IP output. The packet in mbuf chain m contains a skeletal IP
126 * header (with len, off, ttl, proto, tos, src, dst).
127 * The mbuf chain containing the packet will be freed.
128 * The mbuf opt, if present, will not be freed.
131 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
132 int flags, struct ip_moptions *imo, struct inpcb *inp)
135 struct ifnet *ifp = NULL; /* keep compiler happy */
137 int hlen = sizeof(struct ip);
138 int len, off, error = 0;
139 struct sockaddr_in *dst = NULL; /* keep compiler happy */
140 struct in_ifaddr *ia = NULL;
141 int isbroadcast, sw_csum;
142 struct in_addr pkt_dst;
143 struct route iproute;
146 struct secpolicy *sp = NULL;
147 struct socket *so = inp ? inp->inp_socket : NULL;
150 struct secpolicy *sp = NULL;
151 struct tdb_ident *tdbi;
152 #endif /* FAST_IPSEC */
153 struct sockaddr_in *next_hop = NULL;
154 int src_was_INADDR_ANY = 0; /* as the name says... */
161 bzero(ro, sizeof *ro);
164 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
166 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
167 KKASSERT(mtag != NULL);
168 next_hop = m_tag_data(mtag);
171 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
172 struct dn_pkt *dn_pkt;
174 /* Extract info from dummynet tag */
175 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
176 KKASSERT(mtag != NULL);
177 dn_pkt = m_tag_data(mtag);
180 * The packet was already tagged, so part of the
181 * processing was already done, and we need to go down.
182 * Get the calculated parameters from the tag.
186 KKASSERT(ro == &iproute);
187 *ro = dn_pkt->ro; /* structure copy */
189 dst = dn_pkt->dn_dst;
190 if (dst == (struct sockaddr_in *)&(dn_pkt->ro.ro_dst)) {
191 /* If 'dst' points into dummynet tag, adjust it */
192 dst = (struct sockaddr_in *)&(ro->ro_dst);
195 ip = mtod(m, struct ip *);
196 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
198 ia = ifatoia(ro->ro_rt->rt_ifa);
204 m = ip_insertoptions(m, opt, &len);
208 ip = mtod(m, struct ip *);
209 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
214 if (!(flags & (IP_FORWARDING|IP_RAWOUTPUT))) {
215 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
217 ip->ip_id = ip_newid();
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(pkt_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) {
391 m->m_flags &= ~M_MCAST;
395 * If the source address is not specified yet, use the address
396 * of the outoing interface. In case, keep note we did that, so
397 * if the the firewall changes the next-hop causing the output
398 * interface to change, we can fix that.
400 if (ip->ip_src.s_addr == INADDR_ANY) {
401 /* Interface may have no addresses. */
403 ip->ip_src = IA_SIN(ia)->sin_addr;
404 src_was_INADDR_ANY = 1;
410 * Disable packet drop hack.
411 * Packetdrop should be done by queueing.
415 * Verify that we have any chance at all of being able to queue
416 * the packet or packet fragments
418 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
419 ifp->if_snd.ifq_maxlen) {
421 ipstat.ips_odropped++;
427 * Look for broadcast address and
428 * verify user is allowed to send
432 if (!(ifp->if_flags & IFF_BROADCAST)) {
433 error = EADDRNOTAVAIL;
436 if (!(flags & IP_ALLOWBROADCAST)) {
440 /* don't allow broadcast messages to be fragmented */
441 if (ip->ip_len > ifp->if_mtu) {
445 m->m_flags |= M_BCAST;
447 m->m_flags &= ~M_BCAST;
452 /* get SP for this packet */
454 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
456 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
459 ipsecstat.out_inval++;
466 switch (sp->policy) {
467 case IPSEC_POLICY_DISCARD:
469 * This packet is just discarded.
471 ipsecstat.out_polvio++;
474 case IPSEC_POLICY_BYPASS:
475 case IPSEC_POLICY_NONE:
476 /* no need to do IPsec. */
479 case IPSEC_POLICY_IPSEC:
480 if (sp->req == NULL) {
481 /* acquire a policy */
482 error = key_spdacquire(sp);
487 case IPSEC_POLICY_ENTRUST:
489 kprintf("ip_output: Invalid policy found. %d\n", sp->policy);
492 struct ipsec_output_state state;
493 bzero(&state, sizeof state);
495 if (flags & IP_ROUTETOIF) {
497 bzero(&iproute, sizeof iproute);
500 state.dst = (struct sockaddr *)dst;
506 * delayed checksums are not currently compatible with IPsec
508 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
510 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
513 ip->ip_len = htons(ip->ip_len);
514 ip->ip_off = htons(ip->ip_off);
516 error = ipsec4_output(&state, sp, flags);
519 if (flags & IP_ROUTETOIF) {
521 * if we have tunnel mode SA, we may need to ignore
524 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
525 flags &= ~IP_ROUTETOIF;
530 dst = (struct sockaddr_in *)state.dst;
532 /* mbuf is already reclaimed in ipsec4_output. */
542 kprintf("ip4_output (ipsec): error code %d\n", error);
545 /* don't show these error codes to the user */
553 /* be sure to update variables that are affected by ipsec4_output() */
554 ip = mtod(m, struct ip *);
556 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
558 hlen = ip->ip_hl << 2;
560 if (ro->ro_rt == NULL) {
561 if (!(flags & IP_ROUTETOIF)) {
562 kprintf("ip_output: "
563 "can't update route after IPsec processing\n");
564 error = EHOSTUNREACH; /*XXX*/
568 ia = ifatoia(ro->ro_rt->rt_ifa);
569 ifp = ro->ro_rt->rt_ifp;
572 /* make it flipped, again. */
573 ip->ip_len = ntohs(ip->ip_len);
574 ip->ip_off = ntohs(ip->ip_off);
579 * Check the security policy (SP) for the packet and, if
580 * required, do IPsec-related processing. There are two
581 * cases here; the first time a packet is sent through
582 * it will be untagged and handled by ipsec4_checkpolicy.
583 * If the packet is resubmitted to ip_output (e.g. after
584 * AH, ESP, etc. processing), there will be a tag to bypass
585 * the lookup and related policy checking.
587 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
590 tdbi = (struct tdb_ident *)m_tag_data(mtag);
591 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
593 error = -EINVAL; /* force silent drop */
594 m_tag_delete(m, mtag);
596 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
600 * There are four return cases:
601 * sp != NULL apply IPsec policy
602 * sp == NULL, error == 0 no IPsec handling needed
603 * sp == NULL, error == -EINVAL discard packet w/o error
604 * sp == NULL, error != 0 discard packet, report error
607 /* Loop detection, check if ipsec processing already done */
608 KASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
609 for (mtag = m_tag_first(m); mtag != NULL;
610 mtag = m_tag_next(m, mtag)) {
611 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
613 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
614 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
617 * Check if policy has an SA associated with it.
618 * This can happen when an SP has yet to acquire
619 * an SA; e.g. on first reference. If it occurs,
620 * then we let ipsec4_process_packet do its thing.
622 if (sp->req->sav == NULL)
624 tdbi = (struct tdb_ident *)m_tag_data(mtag);
625 if (tdbi->spi == sp->req->sav->spi &&
626 tdbi->proto == sp->req->sav->sah->saidx.proto &&
627 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
628 sizeof(union sockaddr_union)) == 0) {
630 * No IPsec processing is needed, free
633 * NB: null pointer to avoid free at
636 KEY_FREESP(&sp), sp = NULL;
643 * Do delayed checksums now because we send before
644 * this is done in the normal processing path.
646 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
648 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
651 ip->ip_len = htons(ip->ip_len);
652 ip->ip_off = htons(ip->ip_off);
654 /* NB: callee frees mbuf */
655 error = ipsec4_process_packet(m, sp->req, flags, 0);
657 * Preserve KAME behaviour: ENOENT can be returned
658 * when an SA acquire is in progress. Don't propagate
659 * this to user-level; it confuses applications.
661 * XXX this will go away when the SADB is redone.
672 * Hack: -EINVAL is used to signal that a packet
673 * should be silently discarded. This is typically
674 * because we asked key management for an SA and
675 * it was delayed (e.g. kicked up to IKE).
677 if (error == -EINVAL)
681 /* No IPsec processing for this packet. */
685 * If deferred crypto processing is needed, check that
686 * the interface supports it.
688 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
689 if (mtag != NULL && !(ifp->if_capenable & IFCAP_IPSEC)) {
690 /* notify IPsec to do its own crypto */
691 ipsp_skipcrypto_unmark((struct tdb_ident *)m_tag_data(mtag));
692 error = EHOSTUNREACH;
698 #endif /* FAST_IPSEC */
701 * - Xlate: translate packet's addr/port (NAT).
702 * - Firewall: deny/allow/etc.
703 * - Wrap: fake packet's addr/port <unimpl.>
704 * - Encapsulate: put it in another IP and send out. <unimp.>
708 * Run through list of hooks for output packets.
710 if (pfil_has_hooks(&inet_pfil_hook)) {
711 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT);
712 if (error != 0 || m == NULL)
714 ip = mtod(m, struct ip *);
718 * Check with the firewall...
719 * but not if we are already being fwd'd from a firewall.
721 if (fw_enable && IPFW_LOADED && !next_hop) {
722 struct sockaddr_in *old = dst;
723 struct ip_fw_args args;
725 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
726 /* Extract info from dummynet tag */
727 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
728 KKASSERT(mtag != NULL);
730 ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
731 KKASSERT(args.rule != NULL);
733 m_tag_delete(m, mtag);
734 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
742 off = ip_fw_chk_ptr(&args);
745 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
746 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
747 KKASSERT(mtag != NULL);
748 next_hop = m_tag_data(mtag);
753 * On return we must do the following:
754 * m == NULL -> drop the pkt (old interface, deprecated)
755 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface)
756 * 1<=off<= 0xffff -> DIVERT
757 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe
758 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet
759 * dst != old -> IPFIREWALL_FORWARD
760 * off==0, dst==old -> accept
761 * If some of the above modules are not compiled in, then
762 * we should't have to check the corresponding condition
763 * (because the ipfw control socket should not accept
764 * unsupported rules), but better play safe and drop
765 * packets in case of doubt.
767 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
773 ip = mtod(m, struct ip *);
774 if (off == 0 && dst == old) /* common case */
776 if (off & IP_FW_PORT_DYNT_FLAG) {
778 * pass the pkt to dummynet. Need to include
779 * pipe number, m, ifp, ro, dst because these are
780 * not recomputed in the next pass.
781 * All other parameters have been already used and
782 * so they are not needed anymore.
783 * XXX note: if the ifp or ro entry are deleted
784 * while a pkt is in dummynet, we are in trouble!
791 ip_fw_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT, &args);
795 if (off != 0 && !(off & IP_FW_PORT_DYNT_FLAG)) {
796 struct mbuf *clone = NULL;
798 /* Clone packet if we're doing a 'tee' */
799 if ((off & IP_FW_PORT_TEE_FLAG))
800 clone = m_dup(m, MB_DONTWAIT);
804 * delayed checksums are not currently compatible
805 * with divert sockets.
807 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
809 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
812 /* Restore packet header fields to original values */
813 ip->ip_len = htons(ip->ip_len);
814 ip->ip_off = htons(ip->ip_off);
816 /* Deliver packet to divert input routine */
819 /* If 'tee', continue with original packet */
822 ip = mtod(m, struct ip *);
829 /* IPFIREWALL_FORWARD */
831 * Check dst to make sure it is directly reachable on the
832 * interface we previously thought it was.
833 * If it isn't (which may be likely in some situations) we have
834 * to re-route it (ie, find a route for the next-hop and the
835 * associated interface) and set them here. This is nested
836 * forwarding which in most cases is undesirable, except where
837 * such control is nigh impossible. So we do it here.
840 if (off == 0 && old != dst) { /* FORWARD, dst has changed */
843 * XXX To improve readability, this block should be
844 * changed into a function call as below:
846 error = ip_ipforward(&m, &dst, &ifp);
849 if (m == NULL) /* ip_input consumed the mbuf */
852 struct in_ifaddr *ia;
853 struct in_ifaddr_container *iac;
856 * XXX sro_fwd below is static, and a pointer
857 * to it gets passed to routines downstream.
858 * This could have surprisingly bad results in
859 * practice, because its content is overwritten
860 * by subsequent packets.
862 /* There must be a better way to do this next line... */
863 static struct route sro_fwd;
864 struct route *ro_fwd = &sro_fwd;
867 print_ip("IPFIREWALL_FORWARD: New dst ip: ",
868 dst->sin_addr, "\n");
872 * We need to figure out if we have been forwarded
873 * to a local socket. If so, then we should somehow
874 * "loop back" to ip_input, and get directed to the
875 * PCB as if we had received this packet. This is
876 * because it may be dificult to identify the packets
877 * you want to forward until they are being output
878 * and have selected an interface. (e.g. locally
879 * initiated packets) If we used the loopback inteface,
880 * we would not be able to control what happens
881 * as the packet runs through ip_input() as
882 * it is done through a ISR.
885 LIST_FOREACH(iac, INADDR_HASH(dst->sin_addr.s_addr),
888 * If the addr to forward to is one
889 * of ours, we pretend to
890 * be the destination for this packet.
892 if (IA_SIN(iac->ia)->sin_addr.s_addr ==
893 dst->sin_addr.s_addr) {
898 if (ia != NULL) { /* tell ip_input "dont filter" */
899 if (m->m_pkthdr.rcvif == NULL)
900 m->m_pkthdr.rcvif = ifunit("lo0");
901 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
902 m->m_pkthdr.csum_flags |=
903 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
904 m->m_pkthdr.csum_data = 0xffff;
906 m->m_pkthdr.csum_flags |=
907 CSUM_IP_CHECKED | CSUM_IP_VALID;
908 ip->ip_len = htons(ip->ip_len);
909 ip->ip_off = htons(ip->ip_off);
913 /* Some of the logic for this was nicked from above.
915 * This rewrites the cached route in a local PCB.
916 * Is this what we want to do?
918 bcopy(dst, &ro_fwd->ro_dst, sizeof *dst);
919 ro_fwd->ro_rt = NULL;
921 rtalloc_ign(ro_fwd, RTF_PRCLONING);
922 if (ro_fwd->ro_rt == NULL) {
923 ipstat.ips_noroute++;
924 error = EHOSTUNREACH;
928 ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
929 ifp = ro_fwd->ro_rt->rt_ifp;
930 ro_fwd->ro_rt->rt_use++;
931 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
932 dst = (struct sockaddr_in *)
933 ro_fwd->ro_rt->rt_gateway;
934 if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
936 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
938 isbroadcast = in_broadcast(dst->sin_addr, ifp);
939 if (ro->ro_rt != NULL)
941 ro->ro_rt = ro_fwd->ro_rt;
942 dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
944 #endif /* ... block to be put into a function */
946 * If we added a default src ip earlier,
947 * which would have been gotten from the-then
948 * interface, do it again, from the new one.
950 if (src_was_INADDR_ANY)
951 ip->ip_src = IA_SIN(ia)->sin_addr;
956 * if we get here, none of the above matches, and
957 * we have to drop the pkt
960 error = EACCES; /* not sure this is the right error msg */
965 /* 127/8 must not appear on wire - RFC1122. */
966 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
967 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
968 if (!(ifp->if_flags & IFF_LOOPBACK)) {
969 ipstat.ips_badaddr++;
970 error = EADDRNOTAVAIL;
975 m->m_pkthdr.csum_flags |= CSUM_IP;
976 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
977 if (sw_csum & CSUM_DELAY_DATA) {
979 sw_csum &= ~CSUM_DELAY_DATA;
981 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
984 * If small enough for interface, or the interface will take
985 * care of the fragmentation for us, can just send directly.
987 if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
988 !(ip->ip_off & IP_DF))) {
989 ip->ip_len = htons(ip->ip_len);
990 ip->ip_off = htons(ip->ip_off);
992 if (sw_csum & CSUM_DELAY_IP) {
993 if (ip->ip_vhl == IP_VHL_BORING) {
994 ip->ip_sum = in_cksum_hdr(ip);
996 ip->ip_sum = in_cksum(m, hlen);
1000 /* Record statistics for this interface address. */
1001 if (!(flags & IP_FORWARDING) && ia) {
1002 ia->ia_ifa.if_opackets++;
1003 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1007 /* clean ipsec history once it goes out of the node */
1011 #ifdef MBUF_STRESS_TEST
1012 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) {
1013 struct mbuf *m1, *m2;
1016 tmp = length = m->m_pkthdr.len;
1018 while ((length -= mbuf_frag_size) >= 1) {
1019 m1 = m_split(m, length, MB_DONTWAIT);
1023 while (m2->m_next != NULL)
1027 m->m_pkthdr.len = tmp;
1032 if (!mpls_output_process(m, ro->ro_rt))
1035 error = ifp->if_output(ifp, m, (struct sockaddr *)dst,
1040 if (ip->ip_off & IP_DF) {
1043 * This case can happen if the user changed the MTU
1044 * of an interface after enabling IP on it. Because
1045 * most netifs don't keep track of routes pointing to
1046 * them, there is no way for one to update all its
1047 * routes when the MTU is changed.
1049 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
1050 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
1051 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
1052 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
1054 ipstat.ips_cantfrag++;
1059 * Too large for interface; fragment if possible. If successful,
1060 * on return, m will point to a list of packets to be sent.
1062 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
1067 m->m_nextpkt = NULL;
1069 /* clean ipsec history once it goes out of the node */
1073 /* Record statistics for this interface address. */
1075 ia->ia_ifa.if_opackets++;
1076 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1079 if (!mpls_output_process(m, ro->ro_rt))
1082 error = ifp->if_output(ifp, m, (struct sockaddr *)dst,
1090 ipstat.ips_fragmented++;
1093 if (ro == &iproute && ro->ro_rt != NULL) {
1099 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1100 kprintf("DP ip_output call free SP:%p\n", sp));
1115 * Create a chain of fragments which fit the given mtu. m_frag points to the
1116 * mbuf to be fragmented; on return it points to the chain with the fragments.
1117 * Return 0 if no error. If error, m_frag may contain a partially built
1118 * chain of fragments that should be freed by the caller.
1120 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
1121 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
1124 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
1125 u_long if_hwassist_flags, int sw_csum)
1128 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1129 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
1131 struct mbuf *m0 = *m_frag; /* the original packet */
1133 struct mbuf **mnext;
1136 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
1137 ipstat.ips_cantfrag++;
1142 * Must be able to put at least 8 bytes per fragment.
1148 * If the interface will not calculate checksums on
1149 * fragmented packets, then do it here.
1151 if ((m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) &&
1152 !(if_hwassist_flags & CSUM_IP_FRAGS)) {
1153 in_delayed_cksum(m0);
1154 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1157 if (len > PAGE_SIZE) {
1159 * Fragment large datagrams such that each segment
1160 * contains a multiple of PAGE_SIZE amount of data,
1161 * plus headers. This enables a receiver to perform
1162 * page-flipping zero-copy optimizations.
1164 * XXX When does this help given that sender and receiver
1165 * could have different page sizes, and also mtu could
1166 * be less than the receiver's page size ?
1171 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
1175 * firstlen (off - hlen) must be aligned on an
1179 goto smart_frag_failure;
1180 off = ((off - hlen) & ~7) + hlen;
1181 newlen = (~PAGE_MASK) & mtu;
1182 if ((newlen + sizeof(struct ip)) > mtu) {
1183 /* we failed, go back the default */
1194 firstlen = off - hlen;
1195 mnext = &m0->m_nextpkt; /* pointer to next packet */
1198 * Loop through length of segment after first fragment,
1199 * make new header and copy data of each part and link onto chain.
1200 * Here, m0 is the original packet, m is the fragment being created.
1201 * The fragments are linked off the m_nextpkt of the original
1202 * packet, which after processing serves as the first fragment.
1204 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
1205 struct ip *mhip; /* ip header on the fragment */
1207 int mhlen = sizeof(struct ip);
1209 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1212 ipstat.ips_odropped++;
1215 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
1217 * In the first mbuf, leave room for the link header, then
1218 * copy the original IP header including options. The payload
1219 * goes into an additional mbuf chain returned by m_copy().
1221 m->m_data += max_linkhdr;
1222 mhip = mtod(m, struct ip *);
1224 if (hlen > sizeof(struct ip)) {
1225 mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip);
1226 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
1229 /* XXX do we need to add ip->ip_off below ? */
1230 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
1231 if (off + len >= ip->ip_len) { /* last fragment */
1232 len = ip->ip_len - off;
1233 m->m_flags |= M_LASTFRAG;
1235 mhip->ip_off |= IP_MF;
1236 mhip->ip_len = htons((u_short)(len + mhlen));
1237 m->m_next = m_copy(m0, off, len);
1238 if (m->m_next == NULL) { /* copy failed */
1240 error = ENOBUFS; /* ??? */
1241 ipstat.ips_odropped++;
1244 m->m_pkthdr.len = mhlen + len;
1245 m->m_pkthdr.rcvif = (struct ifnet *)NULL;
1246 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1247 mhip->ip_off = htons(mhip->ip_off);
1249 if (sw_csum & CSUM_DELAY_IP)
1250 mhip->ip_sum = in_cksum(m, mhlen);
1252 mnext = &m->m_nextpkt;
1254 ipstat.ips_ofragments += nfrags;
1256 /* set first marker for fragment chain */
1257 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1258 m0->m_pkthdr.csum_data = nfrags;
1261 * Update first fragment by trimming what's been copied out
1262 * and updating header.
1264 m_adj(m0, hlen + firstlen - ip->ip_len);
1265 m0->m_pkthdr.len = hlen + firstlen;
1266 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1267 ip->ip_off |= IP_MF;
1268 ip->ip_off = htons(ip->ip_off);
1270 if (sw_csum & CSUM_DELAY_IP)
1271 ip->ip_sum = in_cksum(m0, hlen);
1279 in_delayed_cksum(struct mbuf *m)
1282 u_short csum, offset;
1284 ip = mtod(m, struct ip *);
1285 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1286 csum = in_cksum_skip(m, ip->ip_len, offset);
1287 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1289 offset += m->m_pkthdr.csum_data; /* checksum offset */
1291 if (offset + sizeof(u_short) > m->m_len) {
1292 kprintf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1293 m->m_len, offset, ip->ip_p);
1296 * this shouldn't happen, but if it does, the
1297 * correct behavior may be to insert the checksum
1298 * in the existing chain instead of rearranging it.
1300 m = m_pullup(m, offset + sizeof(u_short));
1302 *(u_short *)(m->m_data + offset) = csum;
1306 * Insert IP options into preformed packet.
1307 * Adjust IP destination as required for IP source routing,
1308 * as indicated by a non-zero in_addr at the start of the options.
1310 * XXX This routine assumes that the packet has no options in place.
1312 static struct mbuf *
1313 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
1315 struct ipoption *p = mtod(opt, struct ipoption *);
1317 struct ip *ip = mtod(m, struct ip *);
1320 optlen = opt->m_len - sizeof p->ipopt_dst;
1321 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
1323 return (m); /* XXX should fail */
1325 if (p->ipopt_dst.s_addr)
1326 ip->ip_dst = p->ipopt_dst;
1327 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1328 MGETHDR(n, MB_DONTWAIT, MT_HEADER);
1333 n->m_pkthdr.rcvif = (struct ifnet *)NULL;
1334 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1335 m->m_len -= sizeof(struct ip);
1336 m->m_data += sizeof(struct ip);
1339 m->m_len = optlen + sizeof(struct ip);
1340 m->m_data += max_linkhdr;
1341 memcpy(mtod(m, void *), ip, sizeof(struct ip));
1343 m->m_data -= optlen;
1345 m->m_pkthdr.len += optlen;
1346 ovbcopy(ip, mtod(m, caddr_t), sizeof(struct ip));
1348 ip = mtod(m, struct ip *);
1349 bcopy(p->ipopt_list, ip + 1, optlen);
1350 *phlen = sizeof(struct ip) + optlen;
1351 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1352 ip->ip_len += optlen;
1357 * Copy options from ip to jp,
1358 * omitting those not copied during fragmentation.
1361 ip_optcopy(struct ip *ip, struct ip *jp)
1364 int opt, optlen, cnt;
1366 cp = (u_char *)(ip + 1);
1367 dp = (u_char *)(jp + 1);
1368 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1369 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1371 if (opt == IPOPT_EOL)
1373 if (opt == IPOPT_NOP) {
1374 /* Preserve for IP mcast tunnel's LSRR alignment. */
1380 KASSERT(cnt >= IPOPT_OLEN + sizeof *cp,
1381 ("ip_optcopy: malformed ipv4 option"));
1382 optlen = cp[IPOPT_OLEN];
1383 KASSERT(optlen >= IPOPT_OLEN + sizeof *cp && optlen <= cnt,
1384 ("ip_optcopy: malformed ipv4 option"));
1386 /* bogus lengths should have been caught by ip_dooptions */
1389 if (IPOPT_COPIED(opt)) {
1390 bcopy(cp, dp, optlen);
1394 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1400 * IP socket option processing.
1403 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1405 struct inpcb *inp = so->so_pcb;
1409 if (sopt->sopt_level != IPPROTO_IP) {
1413 switch (sopt->sopt_dir) {
1415 switch (sopt->sopt_name) {
1422 if (sopt->sopt_valsize > MLEN) {
1426 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1431 m->m_len = sopt->sopt_valsize;
1432 error = soopt_to_kbuf(sopt, mtod(m, void *), m->m_len,
1434 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1442 case IP_RECVRETOPTS:
1443 case IP_RECVDSTADDR:
1447 error = soopt_to_kbuf(sopt, &optval, sizeof optval,
1451 switch (sopt->sopt_name) {
1453 inp->inp_ip_tos = optval;
1457 inp->inp_ip_ttl = optval;
1460 if (optval > 0 && optval <= MAXTTL)
1461 inp->inp_ip_minttl = optval;
1465 #define OPTSET(bit) \
1467 inp->inp_flags |= bit; \
1469 inp->inp_flags &= ~bit;
1472 OPTSET(INP_RECVOPTS);
1475 case IP_RECVRETOPTS:
1476 OPTSET(INP_RECVRETOPTS);
1479 case IP_RECVDSTADDR:
1480 OPTSET(INP_RECVDSTADDR);
1488 OPTSET(INP_RECVTTL);
1498 case IP_MULTICAST_IF:
1499 case IP_MULTICAST_VIF:
1500 case IP_MULTICAST_TTL:
1501 case IP_MULTICAST_LOOP:
1502 case IP_ADD_MEMBERSHIP:
1503 case IP_DROP_MEMBERSHIP:
1504 error = ip_setmoptions(sopt, &inp->inp_moptions);
1508 error = soopt_to_kbuf(sopt, &optval, sizeof optval,
1514 case IP_PORTRANGE_DEFAULT:
1515 inp->inp_flags &= ~(INP_LOWPORT);
1516 inp->inp_flags &= ~(INP_HIGHPORT);
1519 case IP_PORTRANGE_HIGH:
1520 inp->inp_flags &= ~(INP_LOWPORT);
1521 inp->inp_flags |= INP_HIGHPORT;
1524 case IP_PORTRANGE_LOW:
1525 inp->inp_flags &= ~(INP_HIGHPORT);
1526 inp->inp_flags |= INP_LOWPORT;
1535 #if defined(IPSEC) || defined(FAST_IPSEC)
1536 case IP_IPSEC_POLICY:
1544 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1546 soopt_to_mbuf(sopt, m);
1547 priv = (sopt->sopt_td != NULL &&
1548 suser(sopt->sopt_td) != 0) ? 0 : 1;
1549 req = mtod(m, caddr_t);
1551 optname = sopt->sopt_name;
1552 error = ipsec4_set_policy(inp, optname, req, len, priv);
1559 error = ENOPROTOOPT;
1565 switch (sopt->sopt_name) {
1568 if (inp->inp_options)
1569 soopt_from_kbuf(sopt, mtod(inp->inp_options,
1571 inp->inp_options->m_len);
1573 sopt->sopt_valsize = 0;
1580 case IP_RECVRETOPTS:
1581 case IP_RECVDSTADDR:
1586 switch (sopt->sopt_name) {
1589 optval = inp->inp_ip_tos;
1593 optval = inp->inp_ip_ttl;
1596 optval = inp->inp_ip_minttl;
1599 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1602 optval = OPTBIT(INP_RECVOPTS);
1605 case IP_RECVRETOPTS:
1606 optval = OPTBIT(INP_RECVRETOPTS);
1609 case IP_RECVDSTADDR:
1610 optval = OPTBIT(INP_RECVDSTADDR);
1614 optval = OPTBIT(INP_RECVTTL);
1618 optval = OPTBIT(INP_RECVIF);
1622 if (inp->inp_flags & INP_HIGHPORT)
1623 optval = IP_PORTRANGE_HIGH;
1624 else if (inp->inp_flags & INP_LOWPORT)
1625 optval = IP_PORTRANGE_LOW;
1631 optval = OPTBIT(INP_FAITH);
1634 soopt_from_kbuf(sopt, &optval, sizeof optval);
1637 case IP_MULTICAST_IF:
1638 case IP_MULTICAST_VIF:
1639 case IP_MULTICAST_TTL:
1640 case IP_MULTICAST_LOOP:
1641 case IP_ADD_MEMBERSHIP:
1642 case IP_DROP_MEMBERSHIP:
1643 error = ip_getmoptions(sopt, inp->inp_moptions);
1646 #if defined(IPSEC) || defined(FAST_IPSEC)
1647 case IP_IPSEC_POLICY:
1649 struct mbuf *m = NULL;
1654 req = mtod(m, caddr_t);
1657 error = ipsec4_get_policy(so->so_pcb, req, len, &m);
1659 error = soopt_from_mbuf(sopt, m); /* XXX */
1667 error = ENOPROTOOPT;
1676 * Set up IP options in pcb for insertion in output packets.
1677 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1678 * with destination address if source routed.
1681 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
1687 /* turn off any old options */
1691 if (m == NULL || m->m_len == 0) {
1693 * Only turning off any previous options.
1700 if (m->m_len % sizeof(int32_t))
1703 * IP first-hop destination address will be stored before
1704 * actual options; move other options back
1705 * and clear it when none present.
1707 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1710 m->m_len += sizeof(struct in_addr);
1711 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1712 ovbcopy(mtod(m, caddr_t), cp, cnt);
1713 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1715 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1716 opt = cp[IPOPT_OPTVAL];
1717 if (opt == IPOPT_EOL)
1719 if (opt == IPOPT_NOP)
1722 if (cnt < IPOPT_OLEN + sizeof *cp)
1724 optlen = cp[IPOPT_OLEN];
1725 if (optlen < IPOPT_OLEN + sizeof *cp || optlen > cnt)
1736 * user process specifies route as:
1738 * D must be our final destination (but we can't
1739 * check that since we may not have connected yet).
1740 * A is first hop destination, which doesn't appear in
1741 * actual IP option, but is stored before the options.
1743 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1745 m->m_len -= sizeof(struct in_addr);
1746 cnt -= sizeof(struct in_addr);
1747 optlen -= sizeof(struct in_addr);
1748 cp[IPOPT_OLEN] = optlen;
1750 * Move first hop before start of options.
1752 bcopy(&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1753 sizeof(struct in_addr));
1755 * Then copy rest of options back
1756 * to close up the deleted entry.
1758 ovbcopy(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr),
1759 &cp[IPOPT_OFFSET+1],
1760 cnt - (IPOPT_MINOFF - 1));
1764 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1776 * The whole multicast option thing needs to be re-thought.
1777 * Several of these options are equally applicable to non-multicast
1778 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1779 * standard option (IP_TTL).
1783 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1785 static struct ifnet *
1786 ip_multicast_if(struct in_addr *a, int *ifindexp)
1793 if (ntohl(a->s_addr) >> 24 == 0) {
1794 ifindex = ntohl(a->s_addr) & 0xffffff;
1795 if (ifindex < 0 || if_index < ifindex)
1797 ifp = ifindex2ifnet[ifindex];
1799 *ifindexp = ifindex;
1801 ifp = INADDR_TO_IFP(a);
1807 * Set the IP multicast options in response to user setsockopt().
1810 ip_setmoptions(struct sockopt *sopt, struct ip_moptions **imop)
1814 struct in_addr addr;
1815 struct ip_mreq mreq;
1817 struct ip_moptions *imo = *imop;
1822 * No multicast option buffer attached to the pcb;
1823 * allocate one and initialize to default values.
1825 imo = kmalloc(sizeof *imo, M_IPMOPTS, M_WAITOK);
1828 imo->imo_multicast_ifp = NULL;
1829 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1830 imo->imo_multicast_vif = -1;
1831 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1832 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1833 imo->imo_num_memberships = 0;
1835 switch (sopt->sopt_name) {
1836 /* store an index number for the vif you wanna use in the send */
1837 case IP_MULTICAST_VIF:
1838 if (legal_vif_num == 0) {
1842 error = soopt_to_kbuf(sopt, &i, sizeof i, sizeof i);
1845 if (!legal_vif_num(i) && (i != -1)) {
1849 imo->imo_multicast_vif = i;
1852 case IP_MULTICAST_IF:
1854 * Select the interface for outgoing multicast packets.
1856 error = soopt_to_kbuf(sopt, &addr, sizeof addr, sizeof addr);
1861 * INADDR_ANY is used to remove a previous selection.
1862 * When no interface is selected, a default one is
1863 * chosen every time a multicast packet is sent.
1865 if (addr.s_addr == INADDR_ANY) {
1866 imo->imo_multicast_ifp = NULL;
1870 * The selected interface is identified by its local
1871 * IP address. Find the interface and confirm that
1872 * it supports multicasting.
1875 ifp = ip_multicast_if(&addr, &ifindex);
1876 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1878 error = EADDRNOTAVAIL;
1881 imo->imo_multicast_ifp = ifp;
1883 imo->imo_multicast_addr = addr;
1885 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1889 case IP_MULTICAST_TTL:
1891 * Set the IP time-to-live for outgoing multicast packets.
1892 * The original multicast API required a char argument,
1893 * which is inconsistent with the rest of the socket API.
1894 * We allow either a char or an int.
1896 if (sopt->sopt_valsize == 1) {
1898 error = soopt_to_kbuf(sopt, &ttl, 1, 1);
1901 imo->imo_multicast_ttl = ttl;
1904 error = soopt_to_kbuf(sopt, &ttl, sizeof ttl, sizeof ttl);
1910 imo->imo_multicast_ttl = ttl;
1914 case IP_MULTICAST_LOOP:
1916 * Set the loopback flag for outgoing multicast packets.
1917 * Must be zero or one. The original multicast API required a
1918 * char argument, which is inconsistent with the rest
1919 * of the socket API. We allow either a char or an int.
1921 if (sopt->sopt_valsize == 1) {
1924 error = soopt_to_kbuf(sopt, &loop, 1, 1);
1927 imo->imo_multicast_loop = !!loop;
1931 error = soopt_to_kbuf(sopt, &loop, sizeof loop,
1935 imo->imo_multicast_loop = !!loop;
1939 case IP_ADD_MEMBERSHIP:
1941 * Add a multicast group membership.
1942 * Group must be a valid IP multicast address.
1944 error = soopt_to_kbuf(sopt, &mreq, sizeof mreq, sizeof mreq);
1948 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1954 * If no interface address was provided, use the interface of
1955 * the route to the given multicast address.
1957 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1958 struct sockaddr_in dst;
1961 bzero(&dst, sizeof(struct sockaddr_in));
1962 dst.sin_len = sizeof(struct sockaddr_in);
1963 dst.sin_family = AF_INET;
1964 dst.sin_addr = mreq.imr_multiaddr;
1965 rt = rtlookup((struct sockaddr *)&dst);
1967 error = EADDRNOTAVAIL;
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)) {
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 = soopt_to_kbuf(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 kfree(*imop, M_IPMOPTS);
2099 * Return the IP multicast options in response to user getsockopt().
2102 ip_getmoptions(struct sockopt *sopt, struct ip_moptions *imo)
2104 struct in_addr addr;
2105 struct in_ifaddr *ia;
2110 switch (sopt->sopt_name) {
2111 case IP_MULTICAST_VIF:
2113 optval = imo->imo_multicast_vif;
2116 soopt_from_kbuf(sopt, &optval, sizeof optval);
2119 case IP_MULTICAST_IF:
2120 if (imo == NULL || imo->imo_multicast_ifp == NULL)
2121 addr.s_addr = INADDR_ANY;
2122 else if (imo->imo_multicast_addr.s_addr) {
2123 /* return the value user has set */
2124 addr = imo->imo_multicast_addr;
2126 ia = IFP_TO_IA(imo->imo_multicast_ifp);
2127 addr.s_addr = (ia == NULL) ? INADDR_ANY
2128 : IA_SIN(ia)->sin_addr.s_addr;
2130 soopt_from_kbuf(sopt, &addr, sizeof addr);
2133 case IP_MULTICAST_TTL:
2135 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
2137 optval = coptval = imo->imo_multicast_ttl;
2138 if (sopt->sopt_valsize == 1)
2139 soopt_from_kbuf(sopt, &coptval, 1);
2141 soopt_from_kbuf(sopt, &optval, sizeof optval);
2144 case IP_MULTICAST_LOOP:
2146 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
2148 optval = coptval = imo->imo_multicast_loop;
2149 if (sopt->sopt_valsize == 1)
2150 soopt_from_kbuf(sopt, &coptval, 1);
2152 soopt_from_kbuf(sopt, &optval, sizeof optval);
2156 error = ENOPROTOOPT;
2163 * Discard the IP multicast options.
2166 ip_freemoptions(struct ip_moptions *imo)
2171 for (i = 0; i < imo->imo_num_memberships; ++i)
2172 in_delmulti(imo->imo_membership[i]);
2173 kfree(imo, M_IPMOPTS);
2178 * Routine called from ip_output() to loop back a copy of an IP multicast
2179 * packet to the input queue of a specified interface. Note that this
2180 * calls the output routine of the loopback "driver", but with an interface
2181 * pointer that might NOT be a loopback interface -- evil, but easier than
2182 * replicating that code here.
2185 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
2191 copym = m_copypacket(m, MB_DONTWAIT);
2192 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2193 copym = m_pullup(copym, hlen);
2194 if (copym != NULL) {
2196 * if the checksum hasn't been computed, mark it as valid
2198 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2199 in_delayed_cksum(copym);
2200 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2201 copym->m_pkthdr.csum_flags |=
2202 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2203 copym->m_pkthdr.csum_data = 0xffff;
2206 * We don't bother to fragment if the IP length is greater
2207 * than the interface's MTU. Can this possibly matter?
2209 ip = mtod(copym, struct ip *);
2210 ip->ip_len = htons(ip->ip_len);
2211 ip->ip_off = htons(ip->ip_off);
2213 if (ip->ip_vhl == IP_VHL_BORING) {
2214 ip->ip_sum = in_cksum_hdr(ip);
2216 ip->ip_sum = in_cksum(copym, hlen);
2220 * It's not clear whether there are any lingering
2221 * reentrancy problems in other areas which might
2222 * be exposed by using ip_input directly (in
2223 * particular, everything which modifies the packet
2224 * in-place). Yet another option is using the
2225 * protosw directly to deliver the looped back
2226 * packet. For the moment, we'll err on the side
2227 * of safety by using if_simloop().
2230 if (dst->sin_family != AF_INET) {
2231 kprintf("ip_mloopback: bad address family %d\n",
2233 dst->sin_family = AF_INET;
2238 copym->m_pkthdr.rcvif = ifp;
2241 if_simloop(ifp, copym, dst->sin_family, 0);