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 1.67 2008/10/28 03:07:28 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>
54 #include <sys/sysctl.h>
55 #include <sys/in_cksum.h>
58 #include <sys/thread2.h>
59 #include <sys/mplock2.h>
60 #include <sys/msgport2.h>
63 #include <net/netisr.h>
65 #include <net/route.h>
67 #include <netinet/in.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/ip.h>
70 #include <netinet/in_pcb.h>
71 #include <netinet/in_var.h>
72 #include <netinet/ip_var.h>
74 #include <netproto/mpls/mpls_var.h>
76 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
79 #include <netinet6/ipsec.h>
80 #include <netproto/key/key.h>
82 #include <netproto/key/key_debug.h>
84 #define KEYDEBUG(lev,arg)
89 #include <netproto/ipsec/ipsec.h>
90 #include <netproto/ipsec/xform.h>
91 #include <netproto/ipsec/key.h>
94 #include <net/ipfw/ip_fw.h>
95 #include <net/dummynet/ip_dummynet.h>
97 #define print_ip(x, a, y) kprintf("%s %d.%d.%d.%d%s",\
98 x, (ntohl(a.s_addr)>>24)&0xFF,\
99 (ntohl(a.s_addr)>>16)&0xFF,\
100 (ntohl(a.s_addr)>>8)&0xFF,\
101 (ntohl(a.s_addr))&0xFF, y);
105 #ifdef MBUF_STRESS_TEST
106 int mbuf_frag_size = 0;
107 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
108 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
111 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
112 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
113 static void ip_mloopback
114 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
115 static int ip_getmoptions
116 (struct sockopt *, struct ip_moptions *);
117 static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
118 static int ip_setmoptions
119 (struct sockopt *, struct ip_moptions **);
121 int ip_optcopy(struct ip *, struct ip *);
123 extern int route_assert_owner_access;
125 extern struct protosw inetsw[];
128 ip_localforward(struct mbuf *m, const struct sockaddr_in *dst, int hlen)
130 struct in_ifaddr_container *iac;
133 * We need to figure out if we have been forwarded to a local
134 * socket. If so, then we should somehow "loop back" to
135 * ip_input(), and get directed to the PCB as if we had received
136 * this packet. This is because it may be difficult to identify
137 * the packets you want to forward until they are being output
138 * and have selected an interface (e.g. locally initiated
139 * packets). If we used the loopback inteface, we would not be
140 * able to control what happens as the packet runs through
141 * ip_input() as it is done through a ISR.
143 LIST_FOREACH(iac, INADDR_HASH(dst->sin_addr.s_addr), ia_hash) {
145 * If the addr to forward to is one of ours, we pretend
146 * to be the destination for this packet.
148 if (IA_SIN(iac->ia)->sin_addr.s_addr == dst->sin_addr.s_addr)
154 if (m->m_pkthdr.rcvif == NULL)
155 m->m_pkthdr.rcvif = ifunit("lo0");
156 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
157 m->m_pkthdr.csum_flags |= CSUM_DATA_VALID |
159 m->m_pkthdr.csum_data = 0xffff;
161 m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID;
164 * Make sure that the IP header is in one mbuf,
165 * required by ip_input
167 if (m->m_len < hlen) {
168 m = m_pullup(m, hlen);
170 /* The packet was freed; we are done */
174 ip = mtod(m, struct ip *);
176 ip->ip_len = htons(ip->ip_len);
177 ip->ip_off = htons(ip->ip_off);
180 return 1; /* The packet gets forwarded locally */
186 * IP output. The packet in mbuf chain m contains a skeletal IP
187 * header (with len, off, ttl, proto, tos, src, dst).
188 * The mbuf chain containing the packet will be freed.
189 * The mbuf opt, if present, will not be freed.
192 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
193 int flags, struct ip_moptions *imo, struct inpcb *inp)
196 struct ifnet *ifp = NULL; /* keep compiler happy */
198 int hlen = sizeof(struct ip);
200 struct sockaddr_in *dst = NULL; /* keep compiler happy */
201 struct in_ifaddr *ia = NULL;
202 int isbroadcast, sw_csum;
203 struct in_addr pkt_dst;
204 struct route iproute;
207 struct secpolicy *sp = NULL;
208 struct socket *so = inp ? inp->inp_socket : NULL;
211 struct secpolicy *sp = NULL;
212 struct tdb_ident *tdbi;
213 #endif /* FAST_IPSEC */
214 struct sockaddr_in *next_hop = NULL;
215 int src_was_INADDR_ANY = 0; /* as the name says... */
222 bzero(ro, sizeof *ro);
223 } else if (ro->ro_rt != NULL && ro->ro_rt->rt_cpuid != mycpuid) {
224 if (flags & IP_DEBUGROUTE) {
225 if (route_assert_owner_access) {
227 "rt rt_cpuid %d accessed on cpu %d\n",
228 ro->ro_rt->rt_cpuid, mycpuid);
230 kprintf("ip_output: "
231 "rt rt_cpuid %d accessed on cpu %d\n",
232 ro->ro_rt->rt_cpuid, mycpuid);
239 * If the cached rtentry's owner CPU is not the current CPU,
240 * then don't touch the cached rtentry (remote free is too
241 * expensive in this context); just relocate the route.
244 bzero(ro, sizeof *ro);
247 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
249 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
250 KKASSERT(mtag != NULL);
251 next_hop = m_tag_data(mtag);
254 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
255 struct dn_pkt *dn_pkt;
257 /* Extract info from dummynet tag */
258 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
259 KKASSERT(mtag != NULL);
260 dn_pkt = m_tag_data(mtag);
263 * The packet was already tagged, so part of the
264 * processing was already done, and we need to go down.
265 * Get the calculated parameters from the tag.
269 KKASSERT(ro == &iproute);
270 *ro = dn_pkt->ro; /* structure copy */
271 KKASSERT(ro->ro_rt == NULL || ro->ro_rt->rt_cpuid == mycpuid);
273 dst = dn_pkt->dn_dst;
274 if (dst == (struct sockaddr_in *)&(dn_pkt->ro.ro_dst)) {
275 /* If 'dst' points into dummynet tag, adjust it */
276 dst = (struct sockaddr_in *)&(ro->ro_dst);
279 ip = mtod(m, struct ip *);
280 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
282 ia = ifatoia(ro->ro_rt->rt_ifa);
288 m = ip_insertoptions(m, opt, &len);
292 ip = mtod(m, struct ip *);
297 if (!(flags & (IP_FORWARDING|IP_RAWOUTPUT))) {
298 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
300 ip->ip_id = ip_newid();
301 ipstat.ips_localout++;
303 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
307 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
310 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
313 * Multicast is not MPSAFE yet. Caller must hold
314 * BGL when output a multicast IP packet.
316 ASSERT_MP_LOCK_HELD(curthread);
320 dst = (struct sockaddr_in *)&ro->ro_dst;
322 * If there is a cached route,
323 * check that it is to the same destination
324 * and is still up. If not, free it and try again.
325 * The address family should also be checked in case of sharing the
329 (!(ro->ro_rt->rt_flags & RTF_UP) ||
330 dst->sin_family != AF_INET ||
331 dst->sin_addr.s_addr != pkt_dst.s_addr)) {
335 if (ro->ro_rt == NULL) {
336 bzero(dst, sizeof *dst);
337 dst->sin_family = AF_INET;
338 dst->sin_len = sizeof *dst;
339 dst->sin_addr = pkt_dst;
342 * If routing to interface only,
343 * short circuit routing lookup.
345 if (flags & IP_ROUTETOIF) {
346 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
347 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) {
348 ipstat.ips_noroute++;
354 isbroadcast = in_broadcast(dst->sin_addr, ifp);
355 } else if (IN_MULTICAST(ntohl(pkt_dst.s_addr)) &&
356 imo != NULL && imo->imo_multicast_ifp != NULL) {
358 * Bypass the normal routing lookup for multicast
359 * packets if the interface is specified.
361 ifp = imo->imo_multicast_ifp;
363 isbroadcast = 0; /* fool gcc */
366 * If this is the case, we probably don't want to allocate
367 * a protocol-cloned route since we didn't get one from the
368 * ULP. This lets TCP do its thing, while not burdening
369 * forwarding or ICMP with the overhead of cloning a route.
370 * Of course, we still want to do any cloning requested by
371 * the link layer, as this is probably required in all cases
372 * for correct operation (as it is for ARP).
374 if (ro->ro_rt == NULL)
375 rtalloc_ign(ro, RTF_PRCLONING);
376 if (ro->ro_rt == NULL) {
377 ipstat.ips_noroute++;
378 error = EHOSTUNREACH;
381 ia = ifatoia(ro->ro_rt->rt_ifa);
382 ifp = ro->ro_rt->rt_ifp;
384 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
385 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
386 if (ro->ro_rt->rt_flags & RTF_HOST)
387 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
389 isbroadcast = in_broadcast(dst->sin_addr, ifp);
391 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
392 struct in_multi *inm;
394 m->m_flags |= M_MCAST;
396 * IP destination address is multicast. Make sure "dst"
397 * still points to the address in "ro". (It may have been
398 * changed to point to a gateway address, above.)
400 dst = (struct sockaddr_in *)&ro->ro_dst;
402 * See if the caller provided any multicast options
405 ip->ip_ttl = imo->imo_multicast_ttl;
406 if (imo->imo_multicast_vif != -1) {
409 ip_mcast_src(imo->imo_multicast_vif) :
413 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
416 * Confirm that the outgoing interface supports multicast.
418 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
419 if (!(ifp->if_flags & IFF_MULTICAST)) {
420 ipstat.ips_noroute++;
426 * If source address not specified yet, use address
427 * of outgoing interface.
429 if (ip->ip_src.s_addr == INADDR_ANY) {
430 /* Interface may have no addresses. */
432 ip->ip_src = IA_SIN(ia)->sin_addr;
435 IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
437 (imo == NULL || imo->imo_multicast_loop)) {
439 * If we belong to the destination multicast group
440 * on the outgoing interface, and the caller did not
441 * forbid loopback, loop back a copy.
443 ip_mloopback(ifp, m, dst, hlen);
446 * If we are acting as a multicast router, perform
447 * multicast forwarding as if the packet had just
448 * arrived on the interface to which we are about
449 * to send. The multicast forwarding function
450 * recursively calls this function, using the
451 * IP_FORWARDING flag to prevent infinite recursion.
453 * Multicasts that are looped back by ip_mloopback(),
454 * above, will be forwarded by the ip_input() routine,
457 if (ip_mrouter && !(flags & IP_FORWARDING)) {
459 * If rsvp daemon is not running, do not
460 * set ip_moptions. This ensures that the packet
461 * is multicast and not just sent down one link
462 * as prescribed by rsvpd.
467 ip_mforward(ip, ifp, m, imo) != 0) {
475 * Multicasts with a time-to-live of zero may be looped-
476 * back, above, but must not be transmitted on a network.
477 * Also, multicasts addressed to the loopback interface
478 * are not sent -- the above call to ip_mloopback() will
479 * loop back a copy if this host actually belongs to the
480 * destination group on the loopback interface.
482 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
489 m->m_flags &= ~M_MCAST;
493 * If the source address is not specified yet, use the address
494 * of the outoing interface. In case, keep note we did that, so
495 * if the the firewall changes the next-hop causing the output
496 * interface to change, we can fix that.
498 if (ip->ip_src.s_addr == INADDR_ANY || src_was_INADDR_ANY) {
499 /* Interface may have no addresses. */
501 ip->ip_src = IA_SIN(ia)->sin_addr;
502 src_was_INADDR_ANY = 1;
508 * Disable packet drop hack.
509 * Packetdrop should be done by queueing.
513 * Verify that we have any chance at all of being able to queue
514 * the packet or packet fragments
516 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
517 ifp->if_snd.ifq_maxlen) {
519 ipstat.ips_odropped++;
525 * Look for broadcast address and
526 * verify user is allowed to send
530 if (!(ifp->if_flags & IFF_BROADCAST)) {
531 error = EADDRNOTAVAIL;
534 if (!(flags & IP_ALLOWBROADCAST)) {
538 /* don't allow broadcast messages to be fragmented */
539 if (ip->ip_len > ifp->if_mtu) {
543 m->m_flags |= M_BCAST;
545 m->m_flags &= ~M_BCAST;
550 /* get SP for this packet */
552 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
554 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
557 ipsecstat.out_inval++;
564 switch (sp->policy) {
565 case IPSEC_POLICY_DISCARD:
567 * This packet is just discarded.
569 ipsecstat.out_polvio++;
572 case IPSEC_POLICY_BYPASS:
573 case IPSEC_POLICY_NONE:
574 case IPSEC_POLICY_TCP:
575 /* no need to do IPsec. */
578 case IPSEC_POLICY_IPSEC:
579 if (sp->req == NULL) {
580 /* acquire a policy */
581 error = key_spdacquire(sp);
586 case IPSEC_POLICY_ENTRUST:
588 kprintf("ip_output: Invalid policy found. %d\n", sp->policy);
591 struct ipsec_output_state state;
592 bzero(&state, sizeof state);
594 if (flags & IP_ROUTETOIF) {
596 bzero(&iproute, sizeof iproute);
599 state.dst = (struct sockaddr *)dst;
605 * delayed checksums are not currently compatible with IPsec
607 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
609 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
612 ip->ip_len = htons(ip->ip_len);
613 ip->ip_off = htons(ip->ip_off);
615 error = ipsec4_output(&state, sp, flags);
618 if (flags & IP_ROUTETOIF) {
620 * if we have tunnel mode SA, we may need to ignore
623 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
624 flags &= ~IP_ROUTETOIF;
629 dst = (struct sockaddr_in *)state.dst;
631 /* mbuf is already reclaimed in ipsec4_output. */
641 kprintf("ip4_output (ipsec): error code %d\n", error);
644 /* don't show these error codes to the user */
652 /* be sure to update variables that are affected by ipsec4_output() */
653 ip = mtod(m, struct ip *);
655 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
657 hlen = ip->ip_hl << 2;
659 if (ro->ro_rt == NULL) {
660 if (!(flags & IP_ROUTETOIF)) {
661 kprintf("ip_output: "
662 "can't update route after IPsec processing\n");
663 error = EHOSTUNREACH; /*XXX*/
667 ia = ifatoia(ro->ro_rt->rt_ifa);
668 ifp = ro->ro_rt->rt_ifp;
671 /* make it flipped, again. */
672 ip->ip_len = ntohs(ip->ip_len);
673 ip->ip_off = ntohs(ip->ip_off);
678 * Check the security policy (SP) for the packet and, if
679 * required, do IPsec-related processing. There are two
680 * cases here; the first time a packet is sent through
681 * it will be untagged and handled by ipsec4_checkpolicy.
682 * If the packet is resubmitted to ip_output (e.g. after
683 * AH, ESP, etc. processing), there will be a tag to bypass
684 * the lookup and related policy checking.
686 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
689 tdbi = (struct tdb_ident *)m_tag_data(mtag);
690 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
692 error = -EINVAL; /* force silent drop */
693 m_tag_delete(m, mtag);
695 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
699 * There are four return cases:
700 * sp != NULL apply IPsec policy
701 * sp == NULL, error == 0 no IPsec handling needed
702 * sp == NULL, error == -EINVAL discard packet w/o error
703 * sp == NULL, error != 0 discard packet, report error
706 /* Loop detection, check if ipsec processing already done */
707 KASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
708 for (mtag = m_tag_first(m); mtag != NULL;
709 mtag = m_tag_next(m, mtag)) {
710 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
712 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
713 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
716 * Check if policy has an SA associated with it.
717 * This can happen when an SP has yet to acquire
718 * an SA; e.g. on first reference. If it occurs,
719 * then we let ipsec4_process_packet do its thing.
721 if (sp->req->sav == NULL)
723 tdbi = (struct tdb_ident *)m_tag_data(mtag);
724 if (tdbi->spi == sp->req->sav->spi &&
725 tdbi->proto == sp->req->sav->sah->saidx.proto &&
726 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
727 sizeof(union sockaddr_union)) == 0) {
729 * No IPsec processing is needed, free
732 * NB: null pointer to avoid free at
735 KEY_FREESP(&sp), sp = NULL;
742 * Do delayed checksums now because we send before
743 * this is done in the normal processing path.
745 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
747 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
750 ip->ip_len = htons(ip->ip_len);
751 ip->ip_off = htons(ip->ip_off);
753 /* NB: callee frees mbuf */
754 error = ipsec4_process_packet(m, sp->req, flags, 0);
756 * Preserve KAME behaviour: ENOENT can be returned
757 * when an SA acquire is in progress. Don't propagate
758 * this to user-level; it confuses applications.
760 * XXX this will go away when the SADB is redone.
771 * Hack: -EINVAL is used to signal that a packet
772 * should be silently discarded. This is typically
773 * because we asked key management for an SA and
774 * it was delayed (e.g. kicked up to IKE).
776 if (error == -EINVAL)
780 /* No IPsec processing for this packet. */
784 * If deferred crypto processing is needed, check that
785 * the interface supports it.
787 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
788 if (mtag != NULL && !(ifp->if_capenable & IFCAP_IPSEC)) {
789 /* notify IPsec to do its own crypto */
790 ipsp_skipcrypto_unmark((struct tdb_ident *)m_tag_data(mtag));
791 error = EHOSTUNREACH;
797 #endif /* FAST_IPSEC */
799 /* We are already being fwd'd from a firewall. */
800 if (next_hop != NULL)
804 if (!pfil_has_hooks(&inet_pfil_hook)) {
805 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
807 * Strip dummynet tags from stranded packets
809 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
810 KKASSERT(mtag != NULL);
811 m_tag_delete(m, mtag);
812 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
819 * - Xlate: translate packet's addr/port (NAT).
820 * - Firewall: deny/allow/etc.
821 * - Wrap: fake packet's addr/port <unimpl.>
822 * - Encapsulate: put it in another IP and send out. <unimp.>
826 * Run through list of hooks for output packets.
828 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT);
829 if (error != 0 || m == NULL)
831 ip = mtod(m, struct ip *);
833 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
835 * Check dst to make sure it is directly reachable on the
836 * interface we previously thought it was.
837 * If it isn't (which may be likely in some situations) we have
838 * to re-route it (ie, find a route for the next-hop and the
839 * associated interface) and set them here. This is nested
840 * forwarding which in most cases is undesirable, except where
841 * such control is nigh impossible. So we do it here.
844 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
845 KKASSERT(mtag != NULL);
846 next_hop = m_tag_data(mtag);
849 * Try local forwarding first
851 if (ip_localforward(m, next_hop, hlen))
855 * Relocate the route based on next_hop.
856 * If the current route is inp's cache, keep it untouched.
858 if (ro == &iproute && ro->ro_rt != NULL) {
863 bzero(ro, sizeof *ro);
866 * Forwarding to broadcast address is not allowed.
867 * XXX Should we follow IP_ROUTETOIF?
869 flags &= ~(IP_ALLOWBROADCAST | IP_ROUTETOIF);
871 /* We are doing forwarding now */
872 flags |= IP_FORWARDING;
877 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
878 struct dn_pkt *dn_pkt;
880 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
881 KKASSERT(mtag != NULL);
882 dn_pkt = m_tag_data(mtag);
885 * Under certain cases it is not possible to recalculate
886 * 'ro' and 'dst', let alone 'flags', so just save them in
887 * dummynet tag and avoid the possible wrong reculcalation
888 * when we come back to ip_output() again.
890 * All other parameters have been already used and so they
891 * are not needed anymore.
892 * XXX if the ifp is deleted while a pkt is in dummynet,
893 * we are in trouble! (TODO use ifnet_detach_event)
895 * We need to copy *ro because for ICMP pkts (and maybe
896 * others) the caller passed a pointer into the stack;
897 * dst might also be a pointer into *ro so it needs to
902 ro->ro_rt->rt_refcnt++;
903 if (dst == (struct sockaddr_in *)&ro->ro_dst) {
904 /* 'dst' points into 'ro' */
905 dst = (struct sockaddr_in *)&(dn_pkt->ro.ro_dst);
907 dn_pkt->dn_dst = dst;
908 dn_pkt->flags = flags;
914 /* 127/8 must not appear on wire - RFC1122. */
915 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
916 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
917 if (!(ifp->if_flags & IFF_LOOPBACK)) {
918 ipstat.ips_badaddr++;
919 error = EADDRNOTAVAIL;
924 m->m_pkthdr.csum_flags |= CSUM_IP;
925 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
926 if (sw_csum & CSUM_DELAY_DATA) {
928 sw_csum &= ~CSUM_DELAY_DATA;
930 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
933 * If small enough for interface, or the interface will take
934 * care of the fragmentation for us, can just send directly.
936 if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
937 !(ip->ip_off & IP_DF))) {
938 ip->ip_len = htons(ip->ip_len);
939 ip->ip_off = htons(ip->ip_off);
941 if (sw_csum & CSUM_DELAY_IP) {
942 if (ip->ip_vhl == IP_VHL_BORING)
943 ip->ip_sum = in_cksum_hdr(ip);
945 ip->ip_sum = in_cksum(m, hlen);
948 /* Record statistics for this interface address. */
949 if (!(flags & IP_FORWARDING) && ia) {
950 ia->ia_ifa.if_opackets++;
951 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
955 /* clean ipsec history once it goes out of the node */
959 #ifdef MBUF_STRESS_TEST
960 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) {
961 struct mbuf *m1, *m2;
964 tmp = length = m->m_pkthdr.len;
966 while ((length -= mbuf_frag_size) >= 1) {
967 m1 = m_split(m, length, MB_DONTWAIT);
971 while (m2->m_next != NULL)
975 m->m_pkthdr.len = tmp;
980 if (!mpls_output_process(m, ro->ro_rt))
983 error = ifp->if_output(ifp, m, (struct sockaddr *)dst,
988 if (ip->ip_off & IP_DF) {
991 * This case can happen if the user changed the MTU
992 * of an interface after enabling IP on it. Because
993 * most netifs don't keep track of routes pointing to
994 * them, there is no way for one to update all its
995 * routes when the MTU is changed.
997 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
998 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
999 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
1000 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
1002 ipstat.ips_cantfrag++;
1007 * Too large for interface; fragment if possible. If successful,
1008 * on return, m will point to a list of packets to be sent.
1010 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
1015 m->m_nextpkt = NULL;
1017 /* clean ipsec history once it goes out of the node */
1021 /* Record statistics for this interface address. */
1023 ia->ia_ifa.if_opackets++;
1024 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1027 if (!mpls_output_process(m, ro->ro_rt))
1030 error = ifp->if_output(ifp, m, (struct sockaddr *)dst,
1038 ipstat.ips_fragmented++;
1041 if (ro == &iproute && ro->ro_rt != NULL) {
1047 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1048 kprintf("DP ip_output call free SP:%p\n", sp));
1063 * Create a chain of fragments which fit the given mtu. m_frag points to the
1064 * mbuf to be fragmented; on return it points to the chain with the fragments.
1065 * Return 0 if no error. If error, m_frag may contain a partially built
1066 * chain of fragments that should be freed by the caller.
1068 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
1069 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
1072 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
1073 u_long if_hwassist_flags, int sw_csum)
1076 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1077 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
1079 struct mbuf *m0 = *m_frag; /* the original packet */
1081 struct mbuf **mnext;
1084 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
1085 ipstat.ips_cantfrag++;
1090 * Must be able to put at least 8 bytes per fragment.
1096 * If the interface will not calculate checksums on
1097 * fragmented packets, then do it here.
1099 if ((m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) &&
1100 !(if_hwassist_flags & CSUM_IP_FRAGS)) {
1101 in_delayed_cksum(m0);
1102 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1105 if (len > PAGE_SIZE) {
1107 * Fragment large datagrams such that each segment
1108 * contains a multiple of PAGE_SIZE amount of data,
1109 * plus headers. This enables a receiver to perform
1110 * page-flipping zero-copy optimizations.
1112 * XXX When does this help given that sender and receiver
1113 * could have different page sizes, and also mtu could
1114 * be less than the receiver's page size ?
1119 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
1123 * firstlen (off - hlen) must be aligned on an
1127 goto smart_frag_failure;
1128 off = ((off - hlen) & ~7) + hlen;
1129 newlen = (~PAGE_MASK) & mtu;
1130 if ((newlen + sizeof(struct ip)) > mtu) {
1131 /* we failed, go back the default */
1142 firstlen = off - hlen;
1143 mnext = &m0->m_nextpkt; /* pointer to next packet */
1146 * Loop through length of segment after first fragment,
1147 * make new header and copy data of each part and link onto chain.
1148 * Here, m0 is the original packet, m is the fragment being created.
1149 * The fragments are linked off the m_nextpkt of the original
1150 * packet, which after processing serves as the first fragment.
1152 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
1153 struct ip *mhip; /* ip header on the fragment */
1155 int mhlen = sizeof(struct ip);
1157 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1160 ipstat.ips_odropped++;
1163 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
1165 * In the first mbuf, leave room for the link header, then
1166 * copy the original IP header including options. The payload
1167 * goes into an additional mbuf chain returned by m_copy().
1169 m->m_data += max_linkhdr;
1170 mhip = mtod(m, struct ip *);
1172 if (hlen > sizeof(struct ip)) {
1173 mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip);
1174 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
1177 /* XXX do we need to add ip->ip_off below ? */
1178 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
1179 if (off + len >= ip->ip_len) { /* last fragment */
1180 len = ip->ip_len - off;
1181 m->m_flags |= M_LASTFRAG;
1183 mhip->ip_off |= IP_MF;
1184 mhip->ip_len = htons((u_short)(len + mhlen));
1185 m->m_next = m_copy(m0, off, len);
1186 if (m->m_next == NULL) { /* copy failed */
1188 error = ENOBUFS; /* ??? */
1189 ipstat.ips_odropped++;
1192 m->m_pkthdr.len = mhlen + len;
1193 m->m_pkthdr.rcvif = NULL;
1194 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1195 mhip->ip_off = htons(mhip->ip_off);
1197 if (sw_csum & CSUM_DELAY_IP)
1198 mhip->ip_sum = in_cksum(m, mhlen);
1200 mnext = &m->m_nextpkt;
1202 ipstat.ips_ofragments += nfrags;
1204 /* set first marker for fragment chain */
1205 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1206 m0->m_pkthdr.csum_data = nfrags;
1209 * Update first fragment by trimming what's been copied out
1210 * and updating header.
1212 m_adj(m0, hlen + firstlen - ip->ip_len);
1213 m0->m_pkthdr.len = hlen + firstlen;
1214 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1215 ip->ip_off |= IP_MF;
1216 ip->ip_off = htons(ip->ip_off);
1218 if (sw_csum & CSUM_DELAY_IP)
1219 ip->ip_sum = in_cksum(m0, hlen);
1227 in_delayed_cksum(struct mbuf *m)
1230 u_short csum, offset;
1232 ip = mtod(m, struct ip *);
1233 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1234 csum = in_cksum_skip(m, ip->ip_len, offset);
1235 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1237 offset += m->m_pkthdr.csum_data; /* checksum offset */
1239 if (offset + sizeof(u_short) > m->m_len) {
1240 kprintf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1241 m->m_len, offset, ip->ip_p);
1244 * this shouldn't happen, but if it does, the
1245 * correct behavior may be to insert the checksum
1246 * in the existing chain instead of rearranging it.
1248 m = m_pullup(m, offset + sizeof(u_short));
1250 *(u_short *)(m->m_data + offset) = csum;
1254 * Insert IP options into preformed packet.
1255 * Adjust IP destination as required for IP source routing,
1256 * as indicated by a non-zero in_addr at the start of the options.
1258 * XXX This routine assumes that the packet has no options in place.
1260 static struct mbuf *
1261 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
1263 struct ipoption *p = mtod(opt, struct ipoption *);
1265 struct ip *ip = mtod(m, struct ip *);
1268 optlen = opt->m_len - sizeof p->ipopt_dst;
1269 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
1271 return (m); /* XXX should fail */
1273 if (p->ipopt_dst.s_addr)
1274 ip->ip_dst = p->ipopt_dst;
1275 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1276 MGETHDR(n, MB_DONTWAIT, MT_HEADER);
1281 n->m_pkthdr.rcvif = NULL;
1282 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1283 m->m_len -= sizeof(struct ip);
1284 m->m_data += sizeof(struct ip);
1287 m->m_len = optlen + sizeof(struct ip);
1288 m->m_data += max_linkhdr;
1289 memcpy(mtod(m, void *), ip, sizeof(struct ip));
1291 m->m_data -= optlen;
1293 m->m_pkthdr.len += optlen;
1294 ovbcopy(ip, mtod(m, caddr_t), sizeof(struct ip));
1296 ip = mtod(m, struct ip *);
1297 bcopy(p->ipopt_list, ip + 1, optlen);
1298 *phlen = sizeof(struct ip) + optlen;
1299 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1300 ip->ip_len += optlen;
1305 * Copy options from ip to jp,
1306 * omitting those not copied during fragmentation.
1309 ip_optcopy(struct ip *ip, struct ip *jp)
1312 int opt, optlen, cnt;
1314 cp = (u_char *)(ip + 1);
1315 dp = (u_char *)(jp + 1);
1316 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1317 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1319 if (opt == IPOPT_EOL)
1321 if (opt == IPOPT_NOP) {
1322 /* Preserve for IP mcast tunnel's LSRR alignment. */
1328 KASSERT(cnt >= IPOPT_OLEN + sizeof *cp,
1329 ("ip_optcopy: malformed ipv4 option"));
1330 optlen = cp[IPOPT_OLEN];
1331 KASSERT(optlen >= IPOPT_OLEN + sizeof *cp && optlen <= cnt,
1332 ("ip_optcopy: malformed ipv4 option"));
1334 /* bogus lengths should have been caught by ip_dooptions */
1337 if (IPOPT_COPIED(opt)) {
1338 bcopy(cp, dp, optlen);
1342 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1348 * IP socket option processing.
1351 ip_ctloutput(netmsg_t msg)
1353 struct socket *so = msg->base.nm_so;
1354 struct sockopt *sopt = msg->ctloutput.nm_sopt;
1355 struct inpcb *inp = so->so_pcb;
1359 if (sopt->sopt_level != IPPROTO_IP) {
1364 switch (sopt->sopt_dir) {
1366 switch (sopt->sopt_name) {
1373 if (sopt->sopt_valsize > MLEN) {
1377 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1382 m->m_len = sopt->sopt_valsize;
1383 error = soopt_to_kbuf(sopt, mtod(m, void *), m->m_len,
1385 error = ip_pcbopts(sopt->sopt_name,
1386 &inp->inp_options, m);
1394 case IP_RECVRETOPTS:
1395 case IP_RECVDSTADDR:
1399 error = soopt_to_kbuf(sopt, &optval, sizeof optval,
1403 switch (sopt->sopt_name) {
1405 inp->inp_ip_tos = optval;
1409 inp->inp_ip_ttl = optval;
1412 if (optval >= 0 && optval <= MAXTTL)
1413 inp->inp_ip_minttl = optval;
1417 #define OPTSET(bit) \
1419 inp->inp_flags |= bit; \
1421 inp->inp_flags &= ~bit;
1424 OPTSET(INP_RECVOPTS);
1427 case IP_RECVRETOPTS:
1428 OPTSET(INP_RECVRETOPTS);
1431 case IP_RECVDSTADDR:
1432 OPTSET(INP_RECVDSTADDR);
1440 OPTSET(INP_RECVTTL);
1450 case IP_MULTICAST_IF:
1451 case IP_MULTICAST_VIF:
1452 case IP_MULTICAST_TTL:
1453 case IP_MULTICAST_LOOP:
1454 case IP_ADD_MEMBERSHIP:
1455 case IP_DROP_MEMBERSHIP:
1456 error = ip_setmoptions(sopt, &inp->inp_moptions);
1460 error = soopt_to_kbuf(sopt, &optval, sizeof optval,
1466 case IP_PORTRANGE_DEFAULT:
1467 inp->inp_flags &= ~(INP_LOWPORT);
1468 inp->inp_flags &= ~(INP_HIGHPORT);
1471 case IP_PORTRANGE_HIGH:
1472 inp->inp_flags &= ~(INP_LOWPORT);
1473 inp->inp_flags |= INP_HIGHPORT;
1476 case IP_PORTRANGE_LOW:
1477 inp->inp_flags &= ~(INP_HIGHPORT);
1478 inp->inp_flags |= INP_LOWPORT;
1487 #if defined(IPSEC) || defined(FAST_IPSEC)
1488 case IP_IPSEC_POLICY:
1496 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1498 soopt_to_mbuf(sopt, m);
1499 priv = (sopt->sopt_td != NULL &&
1500 priv_check(sopt->sopt_td, PRIV_ROOT) != 0) ? 0 : 1;
1501 req = mtod(m, caddr_t);
1503 optname = sopt->sopt_name;
1504 error = ipsec4_set_policy(inp, optname, req, len, priv);
1511 error = ENOPROTOOPT;
1517 switch (sopt->sopt_name) {
1520 if (inp->inp_options)
1521 soopt_from_kbuf(sopt, mtod(inp->inp_options,
1523 inp->inp_options->m_len);
1525 sopt->sopt_valsize = 0;
1532 case IP_RECVRETOPTS:
1533 case IP_RECVDSTADDR:
1538 switch (sopt->sopt_name) {
1541 optval = inp->inp_ip_tos;
1545 optval = inp->inp_ip_ttl;
1548 optval = inp->inp_ip_minttl;
1551 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1554 optval = OPTBIT(INP_RECVOPTS);
1557 case IP_RECVRETOPTS:
1558 optval = OPTBIT(INP_RECVRETOPTS);
1561 case IP_RECVDSTADDR:
1562 optval = OPTBIT(INP_RECVDSTADDR);
1566 optval = OPTBIT(INP_RECVTTL);
1570 optval = OPTBIT(INP_RECVIF);
1574 if (inp->inp_flags & INP_HIGHPORT)
1575 optval = IP_PORTRANGE_HIGH;
1576 else if (inp->inp_flags & INP_LOWPORT)
1577 optval = IP_PORTRANGE_LOW;
1583 optval = OPTBIT(INP_FAITH);
1586 soopt_from_kbuf(sopt, &optval, sizeof optval);
1589 case IP_MULTICAST_IF:
1590 case IP_MULTICAST_VIF:
1591 case IP_MULTICAST_TTL:
1592 case IP_MULTICAST_LOOP:
1593 case IP_ADD_MEMBERSHIP:
1594 case IP_DROP_MEMBERSHIP:
1595 error = ip_getmoptions(sopt, inp->inp_moptions);
1598 #if defined(IPSEC) || defined(FAST_IPSEC)
1599 case IP_IPSEC_POLICY:
1601 struct mbuf *m = NULL;
1606 req = mtod(m, caddr_t);
1609 error = ipsec4_get_policy(so->so_pcb, req, len, &m);
1611 error = soopt_from_mbuf(sopt, m); /* XXX */
1619 error = ENOPROTOOPT;
1625 lwkt_replymsg(&msg->lmsg, error);
1629 * Set up IP options in pcb for insertion in output packets.
1630 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1631 * with destination address if source routed.
1634 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
1640 /* turn off any old options */
1644 if (m == NULL || m->m_len == 0) {
1646 * Only turning off any previous options.
1653 if (m->m_len % sizeof(int32_t))
1656 * IP first-hop destination address will be stored before
1657 * actual options; move other options back
1658 * and clear it when none present.
1660 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1663 m->m_len += sizeof(struct in_addr);
1664 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1665 ovbcopy(mtod(m, caddr_t), cp, cnt);
1666 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1668 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1669 opt = cp[IPOPT_OPTVAL];
1670 if (opt == IPOPT_EOL)
1672 if (opt == IPOPT_NOP)
1675 if (cnt < IPOPT_OLEN + sizeof *cp)
1677 optlen = cp[IPOPT_OLEN];
1678 if (optlen < IPOPT_OLEN + sizeof *cp || optlen > cnt)
1689 * user process specifies route as:
1691 * D must be our final destination (but we can't
1692 * check that since we may not have connected yet).
1693 * A is first hop destination, which doesn't appear in
1694 * actual IP option, but is stored before the options.
1696 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1698 m->m_len -= sizeof(struct in_addr);
1699 cnt -= sizeof(struct in_addr);
1700 optlen -= sizeof(struct in_addr);
1701 cp[IPOPT_OLEN] = optlen;
1703 * Move first hop before start of options.
1705 bcopy(&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1706 sizeof(struct in_addr));
1708 * Then copy rest of options back
1709 * to close up the deleted entry.
1711 ovbcopy(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr),
1712 &cp[IPOPT_OFFSET+1],
1713 cnt - (IPOPT_MINOFF - 1));
1717 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1729 * The whole multicast option thing needs to be re-thought.
1730 * Several of these options are equally applicable to non-multicast
1731 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1732 * standard option (IP_TTL).
1736 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1738 static struct ifnet *
1739 ip_multicast_if(struct in_addr *a, int *ifindexp)
1746 if (ntohl(a->s_addr) >> 24 == 0) {
1747 ifindex = ntohl(a->s_addr) & 0xffffff;
1748 if (ifindex < 0 || if_index < ifindex)
1750 ifp = ifindex2ifnet[ifindex];
1752 *ifindexp = ifindex;
1754 ifp = INADDR_TO_IFP(a);
1760 * Set the IP multicast options in response to user setsockopt().
1763 ip_setmoptions(struct sockopt *sopt, struct ip_moptions **imop)
1767 struct in_addr addr;
1768 struct ip_mreq mreq;
1770 struct ip_moptions *imo = *imop;
1775 * No multicast option buffer attached to the pcb;
1776 * allocate one and initialize to default values.
1778 imo = kmalloc(sizeof *imo, M_IPMOPTS, M_WAITOK);
1781 imo->imo_multicast_ifp = NULL;
1782 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1783 imo->imo_multicast_vif = -1;
1784 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1785 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1786 imo->imo_num_memberships = 0;
1788 switch (sopt->sopt_name) {
1789 /* store an index number for the vif you wanna use in the send */
1790 case IP_MULTICAST_VIF:
1791 if (legal_vif_num == 0) {
1795 error = soopt_to_kbuf(sopt, &i, sizeof i, sizeof i);
1798 if (!legal_vif_num(i) && (i != -1)) {
1802 imo->imo_multicast_vif = i;
1805 case IP_MULTICAST_IF:
1807 * Select the interface for outgoing multicast packets.
1809 error = soopt_to_kbuf(sopt, &addr, sizeof addr, sizeof addr);
1814 * INADDR_ANY is used to remove a previous selection.
1815 * When no interface is selected, a default one is
1816 * chosen every time a multicast packet is sent.
1818 if (addr.s_addr == INADDR_ANY) {
1819 imo->imo_multicast_ifp = NULL;
1823 * The selected interface is identified by its local
1824 * IP address. Find the interface and confirm that
1825 * it supports multicasting.
1828 ifp = ip_multicast_if(&addr, &ifindex);
1829 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1831 error = EADDRNOTAVAIL;
1834 imo->imo_multicast_ifp = ifp;
1836 imo->imo_multicast_addr = addr;
1838 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1842 case IP_MULTICAST_TTL:
1844 * Set the IP time-to-live for outgoing multicast packets.
1845 * The original multicast API required a char argument,
1846 * which is inconsistent with the rest of the socket API.
1847 * We allow either a char or an int.
1849 if (sopt->sopt_valsize == 1) {
1851 error = soopt_to_kbuf(sopt, &ttl, 1, 1);
1854 imo->imo_multicast_ttl = ttl;
1857 error = soopt_to_kbuf(sopt, &ttl, sizeof ttl, sizeof ttl);
1863 imo->imo_multicast_ttl = ttl;
1867 case IP_MULTICAST_LOOP:
1869 * Set the loopback flag for outgoing multicast packets.
1870 * Must be zero or one. The original multicast API required a
1871 * char argument, which is inconsistent with the rest
1872 * of the socket API. We allow either a char or an int.
1874 if (sopt->sopt_valsize == 1) {
1877 error = soopt_to_kbuf(sopt, &loop, 1, 1);
1880 imo->imo_multicast_loop = !!loop;
1884 error = soopt_to_kbuf(sopt, &loop, sizeof loop,
1888 imo->imo_multicast_loop = !!loop;
1892 case IP_ADD_MEMBERSHIP:
1894 * Add a multicast group membership.
1895 * Group must be a valid IP multicast address.
1897 error = soopt_to_kbuf(sopt, &mreq, sizeof mreq, sizeof mreq);
1901 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1907 * If no interface address was provided, use the interface of
1908 * the route to the given multicast address.
1910 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1911 struct sockaddr_in dst;
1914 bzero(&dst, sizeof(struct sockaddr_in));
1915 dst.sin_len = sizeof(struct sockaddr_in);
1916 dst.sin_family = AF_INET;
1917 dst.sin_addr = mreq.imr_multiaddr;
1918 rt = rtlookup((struct sockaddr *)&dst);
1920 error = EADDRNOTAVAIL;
1927 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1931 * See if we found an interface, and confirm that it
1932 * supports multicast.
1934 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1935 error = EADDRNOTAVAIL;
1940 * See if the membership already exists or if all the
1941 * membership slots are full.
1943 for (i = 0; i < imo->imo_num_memberships; ++i) {
1944 if (imo->imo_membership[i]->inm_ifp == ifp &&
1945 imo->imo_membership[i]->inm_addr.s_addr
1946 == mreq.imr_multiaddr.s_addr)
1949 if (i < imo->imo_num_memberships) {
1954 if (i == IP_MAX_MEMBERSHIPS) {
1955 error = ETOOMANYREFS;
1960 * Everything looks good; add a new record to the multicast
1961 * address list for the given interface.
1963 if ((imo->imo_membership[i] =
1964 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
1969 ++imo->imo_num_memberships;
1973 case IP_DROP_MEMBERSHIP:
1975 * Drop a multicast group membership.
1976 * Group must be a valid IP multicast address.
1978 error = soopt_to_kbuf(sopt, &mreq, sizeof mreq, sizeof mreq);
1982 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1989 * If an interface address was specified, get a pointer
1990 * to its ifnet structure.
1992 if (mreq.imr_interface.s_addr == INADDR_ANY)
1995 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1997 error = EADDRNOTAVAIL;
2003 * Find the membership in the membership array.
2005 for (i = 0; i < imo->imo_num_memberships; ++i) {
2007 imo->imo_membership[i]->inm_ifp == ifp) &&
2008 imo->imo_membership[i]->inm_addr.s_addr ==
2009 mreq.imr_multiaddr.s_addr)
2012 if (i == imo->imo_num_memberships) {
2013 error = EADDRNOTAVAIL;
2018 * Give up the multicast address record to which the
2019 * membership points.
2021 in_delmulti(imo->imo_membership[i]);
2023 * Remove the gap in the membership array.
2025 for (++i; i < imo->imo_num_memberships; ++i)
2026 imo->imo_membership[i-1] = imo->imo_membership[i];
2027 --imo->imo_num_memberships;
2037 * If all options have default values, no need to keep the mbuf.
2039 if (imo->imo_multicast_ifp == NULL &&
2040 imo->imo_multicast_vif == -1 &&
2041 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2042 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2043 imo->imo_num_memberships == 0) {
2044 kfree(*imop, M_IPMOPTS);
2052 * Return the IP multicast options in response to user getsockopt().
2055 ip_getmoptions(struct sockopt *sopt, struct ip_moptions *imo)
2057 struct in_addr addr;
2058 struct in_ifaddr *ia;
2063 switch (sopt->sopt_name) {
2064 case IP_MULTICAST_VIF:
2066 optval = imo->imo_multicast_vif;
2069 soopt_from_kbuf(sopt, &optval, sizeof optval);
2072 case IP_MULTICAST_IF:
2073 if (imo == NULL || imo->imo_multicast_ifp == NULL)
2074 addr.s_addr = INADDR_ANY;
2075 else if (imo->imo_multicast_addr.s_addr) {
2076 /* return the value user has set */
2077 addr = imo->imo_multicast_addr;
2079 ia = IFP_TO_IA(imo->imo_multicast_ifp);
2080 addr.s_addr = (ia == NULL) ? INADDR_ANY
2081 : IA_SIN(ia)->sin_addr.s_addr;
2083 soopt_from_kbuf(sopt, &addr, sizeof addr);
2086 case IP_MULTICAST_TTL:
2088 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
2090 optval = coptval = imo->imo_multicast_ttl;
2091 if (sopt->sopt_valsize == 1)
2092 soopt_from_kbuf(sopt, &coptval, 1);
2094 soopt_from_kbuf(sopt, &optval, sizeof optval);
2097 case IP_MULTICAST_LOOP:
2099 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
2101 optval = coptval = imo->imo_multicast_loop;
2102 if (sopt->sopt_valsize == 1)
2103 soopt_from_kbuf(sopt, &coptval, 1);
2105 soopt_from_kbuf(sopt, &optval, sizeof optval);
2109 error = ENOPROTOOPT;
2116 * Discard the IP multicast options.
2119 ip_freemoptions(struct ip_moptions *imo)
2124 for (i = 0; i < imo->imo_num_memberships; ++i)
2125 in_delmulti(imo->imo_membership[i]);
2126 kfree(imo, M_IPMOPTS);
2131 * Routine called from ip_output() to loop back a copy of an IP multicast
2132 * packet to the input queue of a specified interface. Note that this
2133 * calls the output routine of the loopback "driver", but with an interface
2134 * pointer that might NOT be a loopback interface -- evil, but easier than
2135 * replicating that code here.
2138 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
2144 copym = m_copypacket(m, MB_DONTWAIT);
2145 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2146 copym = m_pullup(copym, hlen);
2147 if (copym != NULL) {
2149 * if the checksum hasn't been computed, mark it as valid
2151 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2152 in_delayed_cksum(copym);
2153 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2154 copym->m_pkthdr.csum_flags |=
2155 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2156 copym->m_pkthdr.csum_data = 0xffff;
2159 * We don't bother to fragment if the IP length is greater
2160 * than the interface's MTU. Can this possibly matter?
2162 ip = mtod(copym, struct ip *);
2163 ip->ip_len = htons(ip->ip_len);
2164 ip->ip_off = htons(ip->ip_off);
2166 if (ip->ip_vhl == IP_VHL_BORING) {
2167 ip->ip_sum = in_cksum_hdr(ip);
2169 ip->ip_sum = in_cksum(copym, hlen);
2173 * It's not clear whether there are any lingering
2174 * reentrancy problems in other areas which might
2175 * be exposed by using ip_input directly (in
2176 * particular, everything which modifies the packet
2177 * in-place). Yet another option is using the
2178 * protosw directly to deliver the looped back
2179 * packet. For the moment, we'll err on the side
2180 * of safety by using if_simloop().
2183 if (dst->sin_family != AF_INET) {
2184 kprintf("ip_mloopback: bad address family %d\n",
2186 dst->sin_family = AF_INET;
2189 if_simloop(ifp, copym, dst->sin_family, 0);