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 1.22 2004/10/15 22:59:10 hsu Exp $
38 #include "opt_ipdivert.h"
39 #include "opt_ipfilter.h"
40 #include "opt_ipsec.h"
41 #include "opt_random_ip_id.h"
42 #include "opt_mbuf_stress_test.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/malloc.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
53 #include <sys/sysctl.h>
54 #include <sys/in_cksum.h>
57 #include <net/netisr.h>
59 #include <net/route.h>
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/ip.h>
64 #include <netinet/in_pcb.h>
65 #include <netinet/in_var.h>
66 #include <netinet/ip_var.h>
68 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
71 #include <netinet6/ipsec.h>
72 #include <netproto/key/key.h>
74 #include <netproto/key/key_debug.h>
76 #define KEYDEBUG(lev,arg)
81 #include <netproto/ipsec/ipsec.h>
82 #include <netproto/ipsec/xform.h>
83 #include <netproto/ipsec/key.h>
86 #include <net/ipfw/ip_fw.h>
87 #include <net/dummynet/ip_dummynet.h>
89 #define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\
90 x, (ntohl(a.s_addr)>>24)&0xFF,\
91 (ntohl(a.s_addr)>>16)&0xFF,\
92 (ntohl(a.s_addr)>>8)&0xFF,\
93 (ntohl(a.s_addr))&0xFF, y);
97 #ifdef MBUF_STRESS_TEST
98 int mbuf_frag_size = 0;
99 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
100 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
103 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
104 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
105 static void ip_mloopback
106 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
107 static int ip_getmoptions
108 (struct sockopt *, struct ip_moptions *);
109 static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
110 static int ip_setmoptions
111 (struct sockopt *, struct ip_moptions **);
113 int ip_optcopy(struct ip *, struct ip *);
116 extern struct protosw inetsw[];
119 * IP output. The packet in mbuf chain m contains a skeletal IP
120 * header (with len, off, ttl, proto, tos, src, dst).
121 * The mbuf chain containing the packet will be freed.
122 * The mbuf opt, if present, will not be freed.
125 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
126 int flags, struct ip_moptions *imo, struct inpcb *inp)
129 struct ifnet *ifp = NULL; /* keep compiler happy */
131 int hlen = sizeof (struct ip);
132 int len, off, error = 0;
133 struct sockaddr_in *dst = NULL; /* keep compiler happy */
134 struct in_ifaddr *ia = NULL;
135 int isbroadcast, sw_csum;
136 struct in_addr pkt_dst;
137 struct route iproute;
139 struct secpolicy *sp = NULL;
140 struct socket *so = inp ? inp->inp_socket : NULL;
144 struct secpolicy *sp = NULL;
145 struct tdb_ident *tdbi;
147 #endif /* FAST_IPSEC */
148 struct ip_fw_args args;
149 int src_was_INADDR_ANY = 0; /* as the name says... */
153 args.next_hop = NULL;
154 args.divert_rule = 0; /* divert cookie */
156 /* Grab info from MT_TAG mbufs prepended to the chain. */
157 for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) {
158 switch(m0->_m_tag_id) {
160 printf("ip_output: unrecognised MT_TAG tag %d\n",
164 case PACKET_TAG_DUMMYNET:
166 * the packet was already tagged, so part of the
167 * processing was already done, and we need to go down.
168 * Get parameters from the header.
170 args.rule = ((struct dn_pkt *)m0)->rule;
172 ro = & ( ((struct dn_pkt *)m0)->ro ) ;
174 dst = ((struct dn_pkt *)m0)->dn_dst ;
175 ifp = ((struct dn_pkt *)m0)->ifp ;
176 flags = ((struct dn_pkt *)m0)->flags ;
179 case PACKET_TAG_DIVERT:
180 args.divert_rule = (int)m0->m_data & 0xffff;
183 case PACKET_TAG_IPFORWARD:
184 args.next_hop = (struct sockaddr_in *)m0->m_data;
190 KASSERT(!m || (m->m_flags & M_PKTHDR) != 0, ("ip_output: no HDR"));
194 bzero(ro, sizeof(*ro));
197 if (args.rule != NULL) { /* dummynet already saw us */
198 ip = mtod(m, struct ip *);
199 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
201 ia = ifatoia(ro->ro_rt->rt_ifa);
207 m = ip_insertoptions(m, opt, &len);
211 ip = mtod(m, struct ip *);
212 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
217 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
218 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
221 ip->ip_id = ip_randomid();
223 ip->ip_id = htons(ip_id++);
225 ipstat.ips_localout++;
227 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
230 dst = (struct sockaddr_in *)&ro->ro_dst;
232 * If there is a cached route,
233 * check that it is to the same destination
234 * and is still up. If not, free it and try again.
235 * The address family should also be checked in case of sharing the
238 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
239 dst->sin_family != AF_INET ||
240 dst->sin_addr.s_addr != pkt_dst.s_addr)) {
242 ro->ro_rt = (struct rtentry *)0;
244 if (ro->ro_rt == 0) {
245 bzero(dst, sizeof(*dst));
246 dst->sin_family = AF_INET;
247 dst->sin_len = sizeof(*dst);
248 dst->sin_addr = pkt_dst;
251 * If routing to interface only,
252 * short circuit routing lookup.
254 if (flags & IP_ROUTETOIF) {
255 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
256 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
257 ipstat.ips_noroute++;
263 isbroadcast = in_broadcast(dst->sin_addr, ifp);
264 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
265 imo != NULL && imo->imo_multicast_ifp != NULL) {
267 * Bypass the normal routing lookup for multicast
268 * packets if the interface is specified.
270 ifp = imo->imo_multicast_ifp;
272 isbroadcast = 0; /* fool gcc */
275 * If this is the case, we probably don't want to allocate
276 * a protocol-cloned route since we didn't get one from the
277 * ULP. This lets TCP do its thing, while not burdening
278 * forwarding or ICMP with the overhead of cloning a route.
279 * Of course, we still want to do any cloning requested by
280 * the link layer, as this is probably required in all cases
281 * for correct operation (as it is for ARP).
284 rtalloc_ign(ro, RTF_PRCLONING);
285 if (ro->ro_rt == 0) {
286 ipstat.ips_noroute++;
287 error = EHOSTUNREACH;
290 ia = ifatoia(ro->ro_rt->rt_ifa);
291 ifp = ro->ro_rt->rt_ifp;
293 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
294 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
295 if (ro->ro_rt->rt_flags & RTF_HOST)
296 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
298 isbroadcast = in_broadcast(dst->sin_addr, ifp);
300 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
301 struct in_multi *inm;
303 m->m_flags |= M_MCAST;
305 * IP destination address is multicast. Make sure "dst"
306 * still points to the address in "ro". (It may have been
307 * changed to point to a gateway address, above.)
309 dst = (struct sockaddr_in *)&ro->ro_dst;
311 * See if the caller provided any multicast options
314 ip->ip_ttl = imo->imo_multicast_ttl;
315 if (imo->imo_multicast_vif != -1)
318 ip_mcast_src(imo->imo_multicast_vif) :
321 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
323 * Confirm that the outgoing interface supports multicast.
325 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
326 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
327 ipstat.ips_noroute++;
333 * If source address not specified yet, use address
334 * of outgoing interface.
336 if (ip->ip_src.s_addr == INADDR_ANY) {
337 /* Interface may have no addresses. */
339 ip->ip_src = IA_SIN(ia)->sin_addr;
342 IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
344 (imo == NULL || imo->imo_multicast_loop)) {
346 * If we belong to the destination multicast group
347 * on the outgoing interface, and the caller did not
348 * forbid loopback, loop back a copy.
350 ip_mloopback(ifp, m, dst, hlen);
354 * If we are acting as a multicast router, perform
355 * multicast forwarding as if the packet had just
356 * arrived on the interface to which we are about
357 * to send. The multicast forwarding function
358 * recursively calls this function, using the
359 * IP_FORWARDING flag to prevent infinite recursion.
361 * Multicasts that are looped back by ip_mloopback(),
362 * above, will be forwarded by the ip_input() routine,
365 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
367 * If rsvp daemon is not running, do not
368 * set ip_moptions. This ensures that the packet
369 * is multicast and not just sent down one link
370 * as prescribed by rsvpd.
375 ip_mforward(ip, ifp, m, imo) != 0) {
383 * Multicasts with a time-to-live of zero may be looped-
384 * back, above, but must not be transmitted on a network.
385 * Also, multicasts addressed to the loopback interface
386 * are not sent -- the above call to ip_mloopback() will
387 * loop back a copy if this host actually belongs to the
388 * destination group on the loopback interface.
390 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
399 * If the source address is not specified yet, use the address
400 * of the outoing interface. In case, keep note we did that, so
401 * if the the firewall changes the next-hop causing the output
402 * interface to change, we can fix that.
404 if (ip->ip_src.s_addr == INADDR_ANY) {
405 /* Interface may have no addresses. */
407 ip->ip_src = IA_SIN(ia)->sin_addr;
408 src_was_INADDR_ANY = 1;
413 * Verify that we have any chance at all of being able to queue
414 * the packet or packet fragments
416 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
417 ifp->if_snd.ifq_maxlen) {
419 ipstat.ips_odropped++;
424 * Look for broadcast address and
425 * verify user is allowed to send
429 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
430 error = EADDRNOTAVAIL;
433 if ((flags & IP_ALLOWBROADCAST) == 0) {
437 /* don't allow broadcast messages to be fragmented */
438 if (ip->ip_len > ifp->if_mtu) {
442 m->m_flags |= M_BCAST;
444 m->m_flags &= ~M_BCAST;
449 /* get SP for this packet */
451 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
453 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
456 ipsecstat.out_inval++;
463 switch (sp->policy) {
464 case IPSEC_POLICY_DISCARD:
466 * This packet is just discarded.
468 ipsecstat.out_polvio++;
471 case IPSEC_POLICY_BYPASS:
472 case IPSEC_POLICY_NONE:
473 /* no need to do IPsec. */
476 case IPSEC_POLICY_IPSEC:
477 if (sp->req == NULL) {
478 /* acquire a policy */
479 error = key_spdacquire(sp);
484 case IPSEC_POLICY_ENTRUST:
486 printf("ip_output: Invalid policy found. %d\n", sp->policy);
489 struct ipsec_output_state state;
490 bzero(&state, sizeof(state));
492 if (flags & IP_ROUTETOIF) {
494 bzero(&iproute, sizeof(iproute));
497 state.dst = (struct sockaddr *)dst;
503 * delayed checksums are not currently compatible with IPsec
505 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
507 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
510 ip->ip_len = htons(ip->ip_len);
511 ip->ip_off = htons(ip->ip_off);
513 error = ipsec4_output(&state, sp, flags);
516 if (flags & IP_ROUTETOIF) {
518 * if we have tunnel mode SA, we may need to ignore
521 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
522 flags &= ~IP_ROUTETOIF;
527 dst = (struct sockaddr_in *)state.dst;
529 /* mbuf is already reclaimed in ipsec4_output. */
539 printf("ip4_output (ipsec): error code %d\n", error);
542 /* don't show these error codes to the user */
550 /* be sure to update variables that are affected by ipsec4_output() */
551 ip = mtod(m, struct ip *);
553 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
555 hlen = ip->ip_hl << 2;
557 if (ro->ro_rt == NULL) {
558 if ((flags & IP_ROUTETOIF) == 0) {
560 "can't update route after IPsec processing\n");
561 error = EHOSTUNREACH; /*XXX*/
565 ia = ifatoia(ro->ro_rt->rt_ifa);
566 ifp = ro->ro_rt->rt_ifp;
569 /* make it flipped, again. */
570 ip->ip_len = ntohs(ip->ip_len);
571 ip->ip_off = ntohs(ip->ip_off);
576 * Check the security policy (SP) for the packet and, if
577 * required, do IPsec-related processing. There are two
578 * cases here; the first time a packet is sent through
579 * it will be untagged and handled by ipsec4_checkpolicy.
580 * If the packet is resubmitted to ip_output (e.g. after
581 * AH, ESP, etc. processing), there will be a tag to bypass
582 * the lookup and related policy checking.
584 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
587 tdbi = (struct tdb_ident *)(mtag + 1);
588 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
590 error = -EINVAL; /* force silent drop */
591 m_tag_delete(m, mtag);
593 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
597 * There are four return cases:
598 * sp != NULL apply IPsec policy
599 * sp == NULL, error == 0 no IPsec handling needed
600 * sp == NULL, error == -EINVAL discard packet w/o error
601 * sp == NULL, error != 0 discard packet, report error
604 /* Loop detection, check if ipsec processing already done */
605 KASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
606 for (mtag = m_tag_first(m); mtag != NULL;
607 mtag = m_tag_next(m, mtag)) {
608 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
610 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
611 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
614 * Check if policy has an SA associated with it.
615 * This can happen when an SP has yet to acquire
616 * an SA; e.g. on first reference. If it occurs,
617 * then we let ipsec4_process_packet do its thing.
619 if (sp->req->sav == NULL)
621 tdbi = (struct tdb_ident *)(mtag + 1);
622 if (tdbi->spi == sp->req->sav->spi &&
623 tdbi->proto == sp->req->sav->sah->saidx.proto &&
624 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
625 sizeof (union sockaddr_union)) == 0) {
627 * No IPsec processing is needed, free
630 * NB: null pointer to avoid free at
633 KEY_FREESP(&sp), sp = NULL;
640 * Do delayed checksums now because we send before
641 * this is done in the normal processing path.
643 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
645 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
648 ip->ip_len = htons(ip->ip_len);
649 ip->ip_off = htons(ip->ip_off);
651 /* NB: callee frees mbuf */
652 error = ipsec4_process_packet(m, sp->req, flags, 0);
654 * Preserve KAME behaviour: ENOENT can be returned
655 * when an SA acquire is in progress. Don't propagate
656 * this to user-level; it confuses applications.
658 * XXX this will go away when the SADB is redone.
669 * Hack: -EINVAL is used to signal that a packet
670 * should be silently discarded. This is typically
671 * because we asked key management for an SA and
672 * it was delayed (e.g. kicked up to IKE).
674 if (error == -EINVAL)
678 /* No IPsec processing for this packet. */
682 * If deferred crypto processing is needed, check that
683 * the interface supports it.
685 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
686 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) {
687 /* notify IPsec to do its own crypto */
688 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
689 error = EHOSTUNREACH;
695 #endif /* FAST_IPSEC */
698 * - Xlate: translate packet's addr/port (NAT).
699 * - Firewall: deny/allow/etc.
700 * - Wrap: fake packet's addr/port <unimpl.>
701 * - Encapsulate: put it in another IP and send out. <unimp.>
705 * Run through list of hooks for output packets.
707 if (pfil_has_hooks(&inet_pfil_hook)) {
708 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT);
709 if (error != 0 || m == NULL)
711 ip = mtod(m, struct ip *);
715 * Check with the firewall...
716 * but not if we are already being fwd'd from a firewall.
718 if (fw_enable && IPFW_LOADED && !args.next_hop) {
719 struct sockaddr_in *old = dst;
724 off = ip_fw_chk_ptr(&args);
729 * On return we must do the following:
730 * m == NULL -> drop the pkt (old interface, deprecated)
731 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface)
732 * 1<=off<= 0xffff -> DIVERT
733 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe
734 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet
735 * dst != old -> IPFIREWALL_FORWARD
736 * off==0, dst==old -> accept
737 * If some of the above modules are not compiled in, then
738 * we should't have to check the corresponding condition
739 * (because the ipfw control socket should not accept
740 * unsupported rules), but better play safe and drop
741 * packets in case of doubt.
743 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
749 ip = mtod(m, struct ip *);
750 if (off == 0 && dst == old) /* common case */
752 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) {
754 * pass the pkt to dummynet. Need to include
755 * pipe number, m, ifp, ro, dst because these are
756 * not recomputed in the next pass.
757 * All other parameters have been already used and
758 * so they are not needed anymore.
759 * XXX note: if the ifp or ro entry are deleted
760 * while a pkt is in dummynet, we are in trouble!
766 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT,
771 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) {
772 struct mbuf *clone = NULL;
774 /* Clone packet if we're doing a 'tee' */
775 if ((off & IP_FW_PORT_TEE_FLAG) != 0)
776 clone = m_dup(m, MB_DONTWAIT);
780 * delayed checksums are not currently compatible
781 * with divert sockets.
783 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
785 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
788 /* Restore packet header fields to original values */
789 ip->ip_len = htons(ip->ip_len);
790 ip->ip_off = htons(ip->ip_off);
792 /* Deliver packet to divert input routine */
793 divert_packet(m, 0, off & 0xffff, args.divert_rule);
795 /* If 'tee', continue with original packet */
798 ip = mtod(m, struct ip *);
805 /* IPFIREWALL_FORWARD */
807 * Check dst to make sure it is directly reachable on the
808 * interface we previously thought it was.
809 * If it isn't (which may be likely in some situations) we have
810 * to re-route it (ie, find a route for the next-hop and the
811 * associated interface) and set them here. This is nested
812 * forwarding which in most cases is undesirable, except where
813 * such control is nigh impossible. So we do it here.
816 if (off == 0 && old != dst) { /* FORWARD, dst has changed */
819 * XXX To improve readability, this block should be
820 * changed into a function call as below:
822 error = ip_ipforward(&m, &dst, &ifp);
825 if (m == NULL) /* ip_input consumed the mbuf */
828 struct in_ifaddr *ia;
831 * XXX sro_fwd below is static, and a pointer
832 * to it gets passed to routines downstream.
833 * This could have surprisingly bad results in
834 * practice, because its content is overwritten
835 * by subsequent packets.
837 /* There must be a better way to do this next line... */
838 static struct route sro_fwd;
839 struct route *ro_fwd = &sro_fwd;
842 print_ip("IPFIREWALL_FORWARD: New dst ip: ",
843 dst->sin_addr, "\n");
847 * We need to figure out if we have been forwarded
848 * to a local socket. If so, then we should somehow
849 * "loop back" to ip_input, and get directed to the
850 * PCB as if we had received this packet. This is
851 * because it may be dificult to identify the packets
852 * you want to forward until they are being output
853 * and have selected an interface. (e.g. locally
854 * initiated packets) If we used the loopback inteface,
855 * we would not be able to control what happens
856 * as the packet runs through ip_input() as
857 * it is done through a ISR.
860 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) {
862 * If the addr to forward to is one
863 * of ours, we pretend to
864 * be the destination for this packet.
866 if (IA_SIN(ia)->sin_addr.s_addr ==
867 dst->sin_addr.s_addr)
870 if (ia) { /* tell ip_input "dont filter" */
873 tag.mh_type = MT_TAG;
874 tag.mh_flags = PACKET_TAG_IPFORWARD;
875 tag.mh_data = (caddr_t)args.next_hop;
878 if (m->m_pkthdr.rcvif == NULL)
879 m->m_pkthdr.rcvif = ifunit("lo0");
880 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
881 m->m_pkthdr.csum_flags |=
882 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
883 m0->m_pkthdr.csum_data = 0xffff;
885 m->m_pkthdr.csum_flags |=
886 CSUM_IP_CHECKED | CSUM_IP_VALID;
887 ip->ip_len = htons(ip->ip_len);
888 ip->ip_off = htons(ip->ip_off);
889 ip_input((struct mbuf *)&tag);
892 /* Some of the logic for this was
895 * This rewrites the cached route in a local PCB.
896 * Is this what we want to do?
898 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst));
901 rtalloc_ign(ro_fwd, RTF_PRCLONING);
903 if (ro_fwd->ro_rt == 0) {
904 ipstat.ips_noroute++;
905 error = EHOSTUNREACH;
909 ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
910 ifp = ro_fwd->ro_rt->rt_ifp;
911 ro_fwd->ro_rt->rt_use++;
912 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
913 dst = (struct sockaddr_in *)
914 ro_fwd->ro_rt->rt_gateway;
915 if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
917 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
919 isbroadcast = in_broadcast(dst->sin_addr, ifp);
922 ro->ro_rt = ro_fwd->ro_rt;
923 dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
925 #endif /* ... block to be put into a function */
927 * If we added a default src ip earlier,
928 * which would have been gotten from the-then
929 * interface, do it again, from the new one.
931 if (src_was_INADDR_ANY)
932 ip->ip_src = IA_SIN(ia)->sin_addr;
937 * if we get here, none of the above matches, and
938 * we have to drop the pkt
941 error = EACCES; /* not sure this is the right error msg */
946 /* 127/8 must not appear on wire - RFC1122. */
947 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
948 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
949 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
950 ipstat.ips_badaddr++;
951 error = EADDRNOTAVAIL;
956 m->m_pkthdr.csum_flags |= CSUM_IP;
957 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
958 if (sw_csum & CSUM_DELAY_DATA) {
960 sw_csum &= ~CSUM_DELAY_DATA;
962 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
965 * If small enough for interface, or the interface will take
966 * care of the fragmentation for us, can just send directly.
968 if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
969 (ip->ip_off & IP_DF) == 0)) {
970 ip->ip_len = htons(ip->ip_len);
971 ip->ip_off = htons(ip->ip_off);
973 if (sw_csum & CSUM_DELAY_IP) {
974 if (ip->ip_vhl == IP_VHL_BORING) {
975 ip->ip_sum = in_cksum_hdr(ip);
977 ip->ip_sum = in_cksum(m, hlen);
981 /* Record statistics for this interface address. */
982 if (!(flags & IP_FORWARDING) && ia) {
983 ia->ia_ifa.if_opackets++;
984 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
988 /* clean ipsec history once it goes out of the node */
992 #ifdef MBUF_STRESS_TEST
993 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) {
994 struct mbuf *m1, *m2;
997 tmp = length = m->m_pkthdr.len;
999 while ((length -= mbuf_frag_size) >= 1) {
1000 m1 = m_split(m, length, MB_DONTWAIT);
1003 m1->m_flags &= ~M_PKTHDR;
1005 while (m2->m_next != NULL)
1009 m->m_pkthdr.len = tmp;
1012 error = (*ifp->if_output)(ifp, m,
1013 (struct sockaddr *)dst, ro->ro_rt);
1017 if (ip->ip_off & IP_DF) {
1020 * This case can happen if the user changed the MTU
1021 * of an interface after enabling IP on it. Because
1022 * most netifs don't keep track of routes pointing to
1023 * them, there is no way for one to update all its
1024 * routes when the MTU is changed.
1026 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
1027 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
1028 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
1029 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
1031 ipstat.ips_cantfrag++;
1036 * Too large for interface; fragment if possible. If successful,
1037 * on return, m will point to a list of packets to be sent.
1039 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
1046 /* clean ipsec history once it goes out of the node */
1050 /* Record statistics for this interface address. */
1052 ia->ia_ifa.if_opackets++;
1053 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1056 error = (*ifp->if_output)(ifp, m,
1057 (struct sockaddr *)dst, ro->ro_rt);
1063 ipstat.ips_fragmented++;
1066 if (ro == &iproute && ro->ro_rt) {
1072 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1073 printf("DP ip_output call free SP:%p\n", sp));
1088 * Create a chain of fragments which fit the given mtu. m_frag points to the
1089 * mbuf to be fragmented; on return it points to the chain with the fragments.
1090 * Return 0 if no error. If error, m_frag may contain a partially built
1091 * chain of fragments that should be freed by the caller.
1093 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
1094 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
1097 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
1098 u_long if_hwassist_flags, int sw_csum)
1101 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1102 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
1104 struct mbuf *m0 = *m_frag; /* the original packet */
1106 struct mbuf **mnext;
1109 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
1110 ipstat.ips_cantfrag++;
1115 * Must be able to put at least 8 bytes per fragment.
1121 * If the interface will not calculate checksums on
1122 * fragmented packets, then do it here.
1124 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
1125 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
1126 in_delayed_cksum(m0);
1127 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1130 if (len > PAGE_SIZE) {
1132 * Fragment large datagrams such that each segment
1133 * contains a multiple of PAGE_SIZE amount of data,
1134 * plus headers. This enables a receiver to perform
1135 * page-flipping zero-copy optimizations.
1137 * XXX When does this help given that sender and receiver
1138 * could have different page sizes, and also mtu could
1139 * be less than the receiver's page size ?
1144 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
1148 * firstlen (off - hlen) must be aligned on an
1152 goto smart_frag_failure;
1153 off = ((off - hlen) & ~7) + hlen;
1154 newlen = (~PAGE_MASK) & mtu;
1155 if ((newlen + sizeof (struct ip)) > mtu) {
1156 /* we failed, go back the default */
1167 firstlen = off - hlen;
1168 mnext = &m0->m_nextpkt; /* pointer to next packet */
1171 * Loop through length of segment after first fragment,
1172 * make new header and copy data of each part and link onto chain.
1173 * Here, m0 is the original packet, m is the fragment being created.
1174 * The fragments are linked off the m_nextpkt of the original
1175 * packet, which after processing serves as the first fragment.
1177 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
1178 struct ip *mhip; /* ip header on the fragment */
1180 int mhlen = sizeof (struct ip);
1182 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1185 ipstat.ips_odropped++;
1188 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
1190 * In the first mbuf, leave room for the link header, then
1191 * copy the original IP header including options. The payload
1192 * goes into an additional mbuf chain returned by m_copy().
1194 m->m_data += max_linkhdr;
1195 mhip = mtod(m, struct ip *);
1197 if (hlen > sizeof (struct ip)) {
1198 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1199 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
1202 /* XXX do we need to add ip->ip_off below ? */
1203 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
1204 if (off + len >= ip->ip_len) { /* last fragment */
1205 len = ip->ip_len - off;
1206 m->m_flags |= M_LASTFRAG;
1208 mhip->ip_off |= IP_MF;
1209 mhip->ip_len = htons((u_short)(len + mhlen));
1210 m->m_next = m_copy(m0, off, len);
1211 if (m->m_next == 0) { /* copy failed */
1213 error = ENOBUFS; /* ??? */
1214 ipstat.ips_odropped++;
1217 m->m_pkthdr.len = mhlen + len;
1218 m->m_pkthdr.rcvif = (struct ifnet *)0;
1219 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1220 mhip->ip_off = htons(mhip->ip_off);
1222 if (sw_csum & CSUM_DELAY_IP)
1223 mhip->ip_sum = in_cksum(m, mhlen);
1225 mnext = &m->m_nextpkt;
1227 ipstat.ips_ofragments += nfrags;
1229 /* set first marker for fragment chain */
1230 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1231 m0->m_pkthdr.csum_data = nfrags;
1234 * Update first fragment by trimming what's been copied out
1235 * and updating header.
1237 m_adj(m0, hlen + firstlen - ip->ip_len);
1238 m0->m_pkthdr.len = hlen + firstlen;
1239 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1240 ip->ip_off |= IP_MF;
1241 ip->ip_off = htons(ip->ip_off);
1243 if (sw_csum & CSUM_DELAY_IP)
1244 ip->ip_sum = in_cksum(m0, hlen);
1252 in_delayed_cksum(struct mbuf *m)
1255 u_short csum, offset;
1257 ip = mtod(m, struct ip *);
1258 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1259 csum = in_cksum_skip(m, ip->ip_len, offset);
1260 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1262 offset += m->m_pkthdr.csum_data; /* checksum offset */
1264 if (offset + sizeof(u_short) > m->m_len) {
1265 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1266 m->m_len, offset, ip->ip_p);
1269 * this shouldn't happen, but if it does, the
1270 * correct behavior may be to insert the checksum
1271 * in the existing chain instead of rearranging it.
1273 m = m_pullup(m, offset + sizeof(u_short));
1275 *(u_short *)(m->m_data + offset) = csum;
1279 * Insert IP options into preformed packet.
1280 * Adjust IP destination as required for IP source routing,
1281 * as indicated by a non-zero in_addr at the start of the options.
1283 * XXX This routine assumes that the packet has no options in place.
1285 static struct mbuf *
1286 ip_insertoptions(m, opt, phlen)
1291 struct ipoption *p = mtod(opt, struct ipoption *);
1293 struct ip *ip = mtod(m, struct ip *);
1296 optlen = opt->m_len - sizeof(p->ipopt_dst);
1297 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
1299 return (m); /* XXX should fail */
1301 if (p->ipopt_dst.s_addr)
1302 ip->ip_dst = p->ipopt_dst;
1303 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1304 MGETHDR(n, MB_DONTWAIT, MT_HEADER);
1309 n->m_pkthdr.rcvif = (struct ifnet *)0;
1310 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1311 m->m_len -= sizeof(struct ip);
1312 m->m_data += sizeof(struct ip);
1315 m->m_len = optlen + sizeof(struct ip);
1316 m->m_data += max_linkhdr;
1317 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip));
1319 m->m_data -= optlen;
1321 m->m_pkthdr.len += optlen;
1322 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1324 ip = mtod(m, struct ip *);
1325 bcopy(p->ipopt_list, ip + 1, optlen);
1326 *phlen = sizeof(struct ip) + optlen;
1327 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1328 ip->ip_len += optlen;
1333 * Copy options from ip to jp,
1334 * omitting those not copied during fragmentation.
1341 int opt, optlen, cnt;
1343 cp = (u_char *)(ip + 1);
1344 dp = (u_char *)(jp + 1);
1345 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1346 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1348 if (opt == IPOPT_EOL)
1350 if (opt == IPOPT_NOP) {
1351 /* Preserve for IP mcast tunnel's LSRR alignment. */
1357 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp),
1358 ("ip_optcopy: malformed ipv4 option"));
1359 optlen = cp[IPOPT_OLEN];
1360 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt,
1361 ("ip_optcopy: malformed ipv4 option"));
1363 /* bogus lengths should have been caught by ip_dooptions */
1366 if (IPOPT_COPIED(opt)) {
1367 bcopy(cp, dp, optlen);
1371 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1377 * IP socket option processing.
1380 ip_ctloutput(so, sopt)
1382 struct sockopt *sopt;
1384 struct inpcb *inp = sotoinpcb(so);
1388 if (sopt->sopt_level != IPPROTO_IP) {
1392 switch (sopt->sopt_dir) {
1394 switch (sopt->sopt_name) {
1401 if (sopt->sopt_valsize > MLEN) {
1405 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1410 m->m_len = sopt->sopt_valsize;
1411 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1414 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1421 case IP_RECVRETOPTS:
1422 case IP_RECVDSTADDR:
1425 error = sooptcopyin(sopt, &optval, sizeof optval,
1430 switch (sopt->sopt_name) {
1432 inp->inp_ip_tos = optval;
1436 inp->inp_ip_ttl = optval;
1438 #define OPTSET(bit) \
1440 inp->inp_flags |= bit; \
1442 inp->inp_flags &= ~bit;
1445 OPTSET(INP_RECVOPTS);
1448 case IP_RECVRETOPTS:
1449 OPTSET(INP_RECVRETOPTS);
1452 case IP_RECVDSTADDR:
1453 OPTSET(INP_RECVDSTADDR);
1467 case IP_MULTICAST_IF:
1468 case IP_MULTICAST_VIF:
1469 case IP_MULTICAST_TTL:
1470 case IP_MULTICAST_LOOP:
1471 case IP_ADD_MEMBERSHIP:
1472 case IP_DROP_MEMBERSHIP:
1473 error = ip_setmoptions(sopt, &inp->inp_moptions);
1477 error = sooptcopyin(sopt, &optval, sizeof optval,
1483 case IP_PORTRANGE_DEFAULT:
1484 inp->inp_flags &= ~(INP_LOWPORT);
1485 inp->inp_flags &= ~(INP_HIGHPORT);
1488 case IP_PORTRANGE_HIGH:
1489 inp->inp_flags &= ~(INP_LOWPORT);
1490 inp->inp_flags |= INP_HIGHPORT;
1493 case IP_PORTRANGE_LOW:
1494 inp->inp_flags &= ~(INP_HIGHPORT);
1495 inp->inp_flags |= INP_LOWPORT;
1504 #if defined(IPSEC) || defined(FAST_IPSEC)
1505 case IP_IPSEC_POLICY:
1513 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1515 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1517 priv = (sopt->sopt_td != NULL &&
1518 suser(sopt->sopt_td) != 0) ? 0 : 1;
1519 req = mtod(m, caddr_t);
1521 optname = sopt->sopt_name;
1522 error = ipsec4_set_policy(inp, optname, req, len, priv);
1529 error = ENOPROTOOPT;
1535 switch (sopt->sopt_name) {
1538 if (inp->inp_options)
1539 error = sooptcopyout(sopt,
1540 mtod(inp->inp_options,
1542 inp->inp_options->m_len);
1544 sopt->sopt_valsize = 0;
1550 case IP_RECVRETOPTS:
1551 case IP_RECVDSTADDR:
1555 switch (sopt->sopt_name) {
1558 optval = inp->inp_ip_tos;
1562 optval = inp->inp_ip_ttl;
1565 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1568 optval = OPTBIT(INP_RECVOPTS);
1571 case IP_RECVRETOPTS:
1572 optval = OPTBIT(INP_RECVRETOPTS);
1575 case IP_RECVDSTADDR:
1576 optval = OPTBIT(INP_RECVDSTADDR);
1580 optval = OPTBIT(INP_RECVIF);
1584 if (inp->inp_flags & INP_HIGHPORT)
1585 optval = IP_PORTRANGE_HIGH;
1586 else if (inp->inp_flags & INP_LOWPORT)
1587 optval = IP_PORTRANGE_LOW;
1593 optval = OPTBIT(INP_FAITH);
1596 error = sooptcopyout(sopt, &optval, sizeof optval);
1599 case IP_MULTICAST_IF:
1600 case IP_MULTICAST_VIF:
1601 case IP_MULTICAST_TTL:
1602 case IP_MULTICAST_LOOP:
1603 case IP_ADD_MEMBERSHIP:
1604 case IP_DROP_MEMBERSHIP:
1605 error = ip_getmoptions(sopt, inp->inp_moptions);
1608 #if defined(IPSEC) || defined(FAST_IPSEC)
1609 case IP_IPSEC_POLICY:
1611 struct mbuf *m = NULL;
1616 req = mtod(m, caddr_t);
1619 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m);
1621 error = soopt_mcopyout(sopt, m); /* XXX */
1629 error = ENOPROTOOPT;
1638 * Set up IP options in pcb for insertion in output packets.
1639 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1640 * with destination address if source routed.
1643 ip_pcbopts(optname, pcbopt, m)
1645 struct mbuf **pcbopt;
1652 /* turn off any old options */
1654 (void)m_free(*pcbopt);
1656 if (m == (struct mbuf *)0 || m->m_len == 0) {
1658 * Only turning off any previous options.
1665 if (m->m_len % sizeof(int32_t))
1668 * IP first-hop destination address will be stored before
1669 * actual options; move other options back
1670 * and clear it when none present.
1672 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1675 m->m_len += sizeof(struct in_addr);
1676 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1677 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
1678 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1680 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1681 opt = cp[IPOPT_OPTVAL];
1682 if (opt == IPOPT_EOL)
1684 if (opt == IPOPT_NOP)
1687 if (cnt < IPOPT_OLEN + sizeof(*cp))
1689 optlen = cp[IPOPT_OLEN];
1690 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1701 * user process specifies route as:
1703 * D must be our final destination (but we can't
1704 * check that since we may not have connected yet).
1705 * A is first hop destination, which doesn't appear in
1706 * actual IP option, but is stored before the options.
1708 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1710 m->m_len -= sizeof(struct in_addr);
1711 cnt -= sizeof(struct in_addr);
1712 optlen -= sizeof(struct in_addr);
1713 cp[IPOPT_OLEN] = optlen;
1715 * Move first hop before start of options.
1717 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1718 sizeof(struct in_addr));
1720 * Then copy rest of options back
1721 * to close up the deleted entry.
1723 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1724 sizeof(struct in_addr)),
1725 &cp[IPOPT_OFFSET+1],
1726 (unsigned)cnt - (IPOPT_MINOFF - 1));
1730 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1742 * The whole multicast option thing needs to be re-thought.
1743 * Several of these options are equally applicable to non-multicast
1744 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1745 * standard option (IP_TTL).
1749 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1751 static struct ifnet *
1752 ip_multicast_if(a, ifindexp)
1761 if (ntohl(a->s_addr) >> 24 == 0) {
1762 ifindex = ntohl(a->s_addr) & 0xffffff;
1763 if (ifindex < 0 || if_index < ifindex)
1765 ifp = ifindex2ifnet[ifindex];
1767 *ifindexp = ifindex;
1769 INADDR_TO_IFP(*a, ifp);
1775 * Set the IP multicast options in response to user setsockopt().
1778 ip_setmoptions(sopt, imop)
1779 struct sockopt *sopt;
1780 struct ip_moptions **imop;
1784 struct in_addr addr;
1785 struct ip_mreq mreq;
1787 struct ip_moptions *imo = *imop;
1789 struct sockaddr_in *dst;
1795 * No multicast option buffer attached to the pcb;
1796 * allocate one and initialize to default values.
1798 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
1804 imo->imo_multicast_ifp = NULL;
1805 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1806 imo->imo_multicast_vif = -1;
1807 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1808 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1809 imo->imo_num_memberships = 0;
1812 switch (sopt->sopt_name) {
1813 /* store an index number for the vif you wanna use in the send */
1814 case IP_MULTICAST_VIF:
1815 if (legal_vif_num == 0) {
1819 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
1822 if (!legal_vif_num(i) && (i != -1)) {
1826 imo->imo_multicast_vif = i;
1829 case IP_MULTICAST_IF:
1831 * Select the interface for outgoing multicast packets.
1833 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr);
1837 * INADDR_ANY is used to remove a previous selection.
1838 * When no interface is selected, a default one is
1839 * chosen every time a multicast packet is sent.
1841 if (addr.s_addr == INADDR_ANY) {
1842 imo->imo_multicast_ifp = NULL;
1846 * The selected interface is identified by its local
1847 * IP address. Find the interface and confirm that
1848 * it supports multicasting.
1851 ifp = ip_multicast_if(&addr, &ifindex);
1852 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1854 error = EADDRNOTAVAIL;
1857 imo->imo_multicast_ifp = ifp;
1859 imo->imo_multicast_addr = addr;
1861 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1865 case IP_MULTICAST_TTL:
1867 * Set the IP time-to-live for outgoing multicast packets.
1868 * The original multicast API required a char argument,
1869 * which is inconsistent with the rest of the socket API.
1870 * We allow either a char or an int.
1872 if (sopt->sopt_valsize == 1) {
1874 error = sooptcopyin(sopt, &ttl, 1, 1);
1877 imo->imo_multicast_ttl = ttl;
1880 error = sooptcopyin(sopt, &ttl, sizeof ttl,
1887 imo->imo_multicast_ttl = ttl;
1891 case IP_MULTICAST_LOOP:
1893 * Set the loopback flag for outgoing multicast packets.
1894 * Must be zero or one. The original multicast API required a
1895 * char argument, which is inconsistent with the rest
1896 * of the socket API. We allow either a char or an int.
1898 if (sopt->sopt_valsize == 1) {
1900 error = sooptcopyin(sopt, &loop, 1, 1);
1903 imo->imo_multicast_loop = !!loop;
1906 error = sooptcopyin(sopt, &loop, sizeof loop,
1910 imo->imo_multicast_loop = !!loop;
1914 case IP_ADD_MEMBERSHIP:
1916 * Add a multicast group membership.
1917 * Group must be a valid IP multicast address.
1919 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1923 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1929 * If no interface address was provided, use the interface of
1930 * the route to the given multicast address.
1932 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1933 bzero((caddr_t)&ro, sizeof(ro));
1934 dst = (struct sockaddr_in *)&ro.ro_dst;
1935 dst->sin_len = sizeof(*dst);
1936 dst->sin_family = AF_INET;
1937 dst->sin_addr = mreq.imr_multiaddr;
1939 if (ro.ro_rt == NULL) {
1940 error = EADDRNOTAVAIL;
1944 ifp = ro.ro_rt->rt_ifp;
1948 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1952 * See if we found an interface, and confirm that it
1953 * supports multicast.
1955 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1956 error = EADDRNOTAVAIL;
1961 * See if the membership already exists or if all the
1962 * membership slots are full.
1964 for (i = 0; i < imo->imo_num_memberships; ++i) {
1965 if (imo->imo_membership[i]->inm_ifp == ifp &&
1966 imo->imo_membership[i]->inm_addr.s_addr
1967 == mreq.imr_multiaddr.s_addr)
1970 if (i < imo->imo_num_memberships) {
1975 if (i == IP_MAX_MEMBERSHIPS) {
1976 error = ETOOMANYREFS;
1981 * Everything looks good; add a new record to the multicast
1982 * address list for the given interface.
1984 if ((imo->imo_membership[i] =
1985 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
1990 ++imo->imo_num_memberships;
1994 case IP_DROP_MEMBERSHIP:
1996 * Drop a multicast group membership.
1997 * Group must be a valid IP multicast address.
1999 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
2003 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
2010 * If an interface address was specified, get a pointer
2011 * to its ifnet structure.
2013 if (mreq.imr_interface.s_addr == INADDR_ANY)
2016 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
2018 error = EADDRNOTAVAIL;
2024 * Find the membership in the membership array.
2026 for (i = 0; i < imo->imo_num_memberships; ++i) {
2028 imo->imo_membership[i]->inm_ifp == ifp) &&
2029 imo->imo_membership[i]->inm_addr.s_addr ==
2030 mreq.imr_multiaddr.s_addr)
2033 if (i == imo->imo_num_memberships) {
2034 error = EADDRNOTAVAIL;
2039 * Give up the multicast address record to which the
2040 * membership points.
2042 in_delmulti(imo->imo_membership[i]);
2044 * Remove the gap in the membership array.
2046 for (++i; i < imo->imo_num_memberships; ++i)
2047 imo->imo_membership[i-1] = imo->imo_membership[i];
2048 --imo->imo_num_memberships;
2058 * If all options have default values, no need to keep the mbuf.
2060 if (imo->imo_multicast_ifp == NULL &&
2061 imo->imo_multicast_vif == -1 &&
2062 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2063 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2064 imo->imo_num_memberships == 0) {
2065 free(*imop, M_IPMOPTS);
2073 * Return the IP multicast options in response to user getsockopt().
2076 ip_getmoptions(sopt, imo)
2077 struct sockopt *sopt;
2078 struct ip_moptions *imo;
2080 struct in_addr addr;
2081 struct in_ifaddr *ia;
2086 switch (sopt->sopt_name) {
2087 case IP_MULTICAST_VIF:
2089 optval = imo->imo_multicast_vif;
2092 error = sooptcopyout(sopt, &optval, sizeof optval);
2095 case IP_MULTICAST_IF:
2096 if (imo == NULL || imo->imo_multicast_ifp == NULL)
2097 addr.s_addr = INADDR_ANY;
2098 else if (imo->imo_multicast_addr.s_addr) {
2099 /* return the value user has set */
2100 addr = imo->imo_multicast_addr;
2102 IFP_TO_IA(imo->imo_multicast_ifp, ia);
2103 addr.s_addr = (ia == NULL) ? INADDR_ANY
2104 : IA_SIN(ia)->sin_addr.s_addr;
2106 error = sooptcopyout(sopt, &addr, sizeof addr);
2109 case IP_MULTICAST_TTL:
2111 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
2113 optval = coptval = imo->imo_multicast_ttl;
2114 if (sopt->sopt_valsize == 1)
2115 error = sooptcopyout(sopt, &coptval, 1);
2117 error = sooptcopyout(sopt, &optval, sizeof optval);
2120 case IP_MULTICAST_LOOP:
2122 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
2124 optval = coptval = imo->imo_multicast_loop;
2125 if (sopt->sopt_valsize == 1)
2126 error = sooptcopyout(sopt, &coptval, 1);
2128 error = sooptcopyout(sopt, &optval, sizeof optval);
2132 error = ENOPROTOOPT;
2139 * Discard the IP multicast options.
2142 ip_freemoptions(imo)
2143 struct ip_moptions *imo;
2148 for (i = 0; i < imo->imo_num_memberships; ++i)
2149 in_delmulti(imo->imo_membership[i]);
2150 free(imo, M_IPMOPTS);
2155 * Routine called from ip_output() to loop back a copy of an IP multicast
2156 * packet to the input queue of a specified interface. Note that this
2157 * calls the output routine of the loopback "driver", but with an interface
2158 * pointer that might NOT be a loopback interface -- evil, but easier than
2159 * replicating that code here.
2162 ip_mloopback(ifp, m, dst, hlen)
2165 struct sockaddr_in *dst;
2171 copym = m_copy(m, 0, M_COPYALL);
2172 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2173 copym = m_pullup(copym, hlen);
2174 if (copym != NULL) {
2176 * if the checksum hasn't been computed, mark it as valid
2178 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2179 in_delayed_cksum(copym);
2180 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2181 copym->m_pkthdr.csum_flags |=
2182 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2183 copym->m_pkthdr.csum_data = 0xffff;
2186 * We don't bother to fragment if the IP length is greater
2187 * than the interface's MTU. Can this possibly matter?
2189 ip = mtod(copym, struct ip *);
2190 ip->ip_len = htons(ip->ip_len);
2191 ip->ip_off = htons(ip->ip_off);
2193 if (ip->ip_vhl == IP_VHL_BORING) {
2194 ip->ip_sum = in_cksum_hdr(ip);
2196 ip->ip_sum = in_cksum(copym, hlen);
2200 * It's not clear whether there are any lingering
2201 * reentrancy problems in other areas which might
2202 * be exposed by using ip_input directly (in
2203 * particular, everything which modifies the packet
2204 * in-place). Yet another option is using the
2205 * protosw directly to deliver the looped back
2206 * packet. For the moment, we'll err on the side
2207 * of safety by using if_simloop().
2210 if (dst->sin_family != AF_INET) {
2211 printf("ip_mloopback: bad address family %d\n",
2213 dst->sin_family = AF_INET;
2218 copym->m_pkthdr.rcvif = ifp;
2221 if_simloop(ifp, copym, dst->sin_family, 0);