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/thread2.h>
56 #include <sys/in_cksum.h>
60 #include <net/netisr.h>
62 #include <net/route.h>
64 #include <netinet/in.h>
65 #include <netinet/in_systm.h>
66 #include <netinet/ip.h>
67 #include <netinet/in_pcb.h>
68 #include <netinet/in_var.h>
69 #include <netinet/ip_var.h>
71 #include <netproto/mpls/mpls_var.h>
73 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
76 #include <netinet6/ipsec.h>
77 #include <netproto/key/key.h>
79 #include <netproto/key/key_debug.h>
81 #define KEYDEBUG(lev,arg)
86 #include <netproto/ipsec/ipsec.h>
87 #include <netproto/ipsec/xform.h>
88 #include <netproto/ipsec/key.h>
91 #include <net/ipfw/ip_fw.h>
92 #include <net/dummynet/ip_dummynet.h>
94 #define print_ip(x, a, y) kprintf("%s %d.%d.%d.%d%s",\
95 x, (ntohl(a.s_addr)>>24)&0xFF,\
96 (ntohl(a.s_addr)>>16)&0xFF,\
97 (ntohl(a.s_addr)>>8)&0xFF,\
98 (ntohl(a.s_addr))&0xFF, y);
102 #ifdef MBUF_STRESS_TEST
103 int mbuf_frag_size = 0;
104 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
105 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
108 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
109 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
110 static void ip_mloopback
111 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
112 static int ip_getmoptions
113 (struct sockopt *, struct ip_moptions *);
114 static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
115 static int ip_setmoptions
116 (struct sockopt *, struct ip_moptions **);
118 int ip_optcopy(struct ip *, struct ip *);
120 extern int route_assert_owner_access;
121 extern void db_print_backtrace(void);
123 extern struct protosw inetsw[];
126 ip_localforward(struct mbuf *m, const struct sockaddr_in *dst, int hlen)
128 struct in_ifaddr_container *iac;
131 * We need to figure out if we have been forwarded to a local
132 * socket. If so, then we should somehow "loop back" to
133 * ip_input(), and get directed to the PCB as if we had received
134 * this packet. This is because it may be difficult to identify
135 * the packets you want to forward until they are being output
136 * and have selected an interface (e.g. locally initiated
137 * packets). If we used the loopback inteface, we would not be
138 * able to control what happens as the packet runs through
139 * ip_input() as it is done through a ISR.
141 LIST_FOREACH(iac, INADDR_HASH(dst->sin_addr.s_addr), ia_hash) {
143 * If the addr to forward to is one of ours, we pretend
144 * to be the destination for this packet.
146 if (IA_SIN(iac->ia)->sin_addr.s_addr == dst->sin_addr.s_addr)
152 if (m->m_pkthdr.rcvif == NULL)
153 m->m_pkthdr.rcvif = ifunit("lo0");
154 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
155 m->m_pkthdr.csum_flags |= CSUM_DATA_VALID |
157 m->m_pkthdr.csum_data = 0xffff;
159 m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID;
162 * Make sure that the IP header is in one mbuf,
163 * required by ip_input
165 if (m->m_len < hlen) {
166 m = m_pullup(m, hlen);
168 /* The packet was freed; we are done */
172 ip = mtod(m, struct ip *);
174 ip->ip_len = htons(ip->ip_len);
175 ip->ip_off = htons(ip->ip_off);
178 return 1; /* The packet gets forwarded locally */
184 * IP output. The packet in mbuf chain m contains a skeletal IP
185 * header (with len, off, ttl, proto, tos, src, dst).
186 * The mbuf chain containing the packet will be freed.
187 * The mbuf opt, if present, will not be freed.
190 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
191 int flags, struct ip_moptions *imo, struct inpcb *inp)
194 struct ifnet *ifp = NULL; /* keep compiler happy */
196 int hlen = sizeof(struct ip);
198 struct sockaddr_in *dst = NULL; /* keep compiler happy */
199 struct in_ifaddr *ia = NULL;
200 int isbroadcast, sw_csum;
201 struct in_addr pkt_dst;
202 struct route iproute;
205 struct secpolicy *sp = NULL;
206 struct socket *so = inp ? inp->inp_socket : NULL;
209 struct secpolicy *sp = NULL;
210 struct tdb_ident *tdbi;
211 #endif /* FAST_IPSEC */
212 struct sockaddr_in *next_hop = NULL;
213 int src_was_INADDR_ANY = 0; /* as the name says... */
220 bzero(ro, sizeof *ro);
221 } else if (ro->ro_rt != NULL && ro->ro_rt->rt_cpuid != mycpuid) {
222 if (flags & IP_DEBUGROUTE) {
223 if (route_assert_owner_access) {
225 "rt rt_cpuid %d accessed on cpu %d\n",
226 ro->ro_rt->rt_cpuid, mycpuid);
228 kprintf("ip_output: "
229 "rt rt_cpuid %d accessed on cpu %d\n",
230 ro->ro_rt->rt_cpuid, mycpuid);
231 db_print_backtrace();
237 * If the cached rtentry's owner CPU is not the current CPU,
238 * then don't touch the cached rtentry (remote free is too
239 * expensive in this context); just relocate the route.
242 bzero(ro, sizeof *ro);
245 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
247 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
248 KKASSERT(mtag != NULL);
249 next_hop = m_tag_data(mtag);
252 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
253 struct dn_pkt *dn_pkt;
255 /* Extract info from dummynet tag */
256 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
257 KKASSERT(mtag != NULL);
258 dn_pkt = m_tag_data(mtag);
261 * The packet was already tagged, so part of the
262 * processing was already done, and we need to go down.
263 * Get the calculated parameters from the tag.
267 KKASSERT(ro == &iproute);
268 *ro = dn_pkt->ro; /* structure copy */
269 KKASSERT(ro->ro_rt == NULL || ro->ro_rt->rt_cpuid == mycpuid);
271 dst = dn_pkt->dn_dst;
272 if (dst == (struct sockaddr_in *)&(dn_pkt->ro.ro_dst)) {
273 /* If 'dst' points into dummynet tag, adjust it */
274 dst = (struct sockaddr_in *)&(ro->ro_dst);
277 ip = mtod(m, struct ip *);
278 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
280 ia = ifatoia(ro->ro_rt->rt_ifa);
286 m = ip_insertoptions(m, opt, &len);
290 ip = mtod(m, struct ip *);
295 if (!(flags & (IP_FORWARDING|IP_RAWOUTPUT))) {
296 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
298 ip->ip_id = ip_newid();
299 ipstat.ips_localout++;
301 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
305 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
308 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
311 * Multicast is not MPSAFE yet. Caller must hold
312 * BGL when output a multicast IP packet.
314 ASSERT_MP_LOCK_HELD(curthread);
318 dst = (struct sockaddr_in *)&ro->ro_dst;
320 * If there is a cached route,
321 * check that it is to the same destination
322 * and is still up. If not, free it and try again.
323 * The address family should also be checked in case of sharing the
327 (!(ro->ro_rt->rt_flags & RTF_UP) ||
328 dst->sin_family != AF_INET ||
329 dst->sin_addr.s_addr != pkt_dst.s_addr)) {
331 ro->ro_rt = (struct rtentry *)NULL;
333 if (ro->ro_rt == NULL) {
334 bzero(dst, sizeof *dst);
335 dst->sin_family = AF_INET;
336 dst->sin_len = sizeof *dst;
337 dst->sin_addr = pkt_dst;
340 * If routing to interface only,
341 * short circuit routing lookup.
343 if (flags & IP_ROUTETOIF) {
344 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
345 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) {
346 ipstat.ips_noroute++;
352 isbroadcast = in_broadcast(dst->sin_addr, ifp);
353 } else if (IN_MULTICAST(ntohl(pkt_dst.s_addr)) &&
354 imo != NULL && imo->imo_multicast_ifp != NULL) {
356 * Bypass the normal routing lookup for multicast
357 * packets if the interface is specified.
359 ifp = imo->imo_multicast_ifp;
361 isbroadcast = 0; /* fool gcc */
364 * If this is the case, we probably don't want to allocate
365 * a protocol-cloned route since we didn't get one from the
366 * ULP. This lets TCP do its thing, while not burdening
367 * forwarding or ICMP with the overhead of cloning a route.
368 * Of course, we still want to do any cloning requested by
369 * the link layer, as this is probably required in all cases
370 * for correct operation (as it is for ARP).
372 if (ro->ro_rt == NULL)
373 rtalloc_ign(ro, RTF_PRCLONING);
374 if (ro->ro_rt == NULL) {
375 ipstat.ips_noroute++;
376 error = EHOSTUNREACH;
379 ia = ifatoia(ro->ro_rt->rt_ifa);
380 ifp = ro->ro_rt->rt_ifp;
382 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
383 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
384 if (ro->ro_rt->rt_flags & RTF_HOST)
385 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
387 isbroadcast = in_broadcast(dst->sin_addr, ifp);
389 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
390 struct in_multi *inm;
392 m->m_flags |= M_MCAST;
394 * IP destination address is multicast. Make sure "dst"
395 * still points to the address in "ro". (It may have been
396 * changed to point to a gateway address, above.)
398 dst = (struct sockaddr_in *)&ro->ro_dst;
400 * See if the caller provided any multicast options
403 ip->ip_ttl = imo->imo_multicast_ttl;
404 if (imo->imo_multicast_vif != -1) {
407 ip_mcast_src(imo->imo_multicast_vif) :
411 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
414 * Confirm that the outgoing interface supports multicast.
416 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
417 if (!(ifp->if_flags & IFF_MULTICAST)) {
418 ipstat.ips_noroute++;
424 * If source address not specified yet, use address
425 * of outgoing interface.
427 if (ip->ip_src.s_addr == INADDR_ANY) {
428 /* Interface may have no addresses. */
430 ip->ip_src = IA_SIN(ia)->sin_addr;
433 IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
435 (imo == NULL || imo->imo_multicast_loop)) {
437 * If we belong to the destination multicast group
438 * on the outgoing interface, and the caller did not
439 * forbid loopback, loop back a copy.
441 ip_mloopback(ifp, m, dst, hlen);
444 * If we are acting as a multicast router, perform
445 * multicast forwarding as if the packet had just
446 * arrived on the interface to which we are about
447 * to send. The multicast forwarding function
448 * recursively calls this function, using the
449 * IP_FORWARDING flag to prevent infinite recursion.
451 * Multicasts that are looped back by ip_mloopback(),
452 * above, will be forwarded by the ip_input() routine,
455 if (ip_mrouter && !(flags & IP_FORWARDING)) {
457 * If rsvp daemon is not running, do not
458 * set ip_moptions. This ensures that the packet
459 * is multicast and not just sent down one link
460 * as prescribed by rsvpd.
465 ip_mforward(ip, ifp, m, imo) != 0) {
473 * Multicasts with a time-to-live of zero may be looped-
474 * back, above, but must not be transmitted on a network.
475 * Also, multicasts addressed to the loopback interface
476 * are not sent -- the above call to ip_mloopback() will
477 * loop back a copy if this host actually belongs to the
478 * destination group on the loopback interface.
480 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
487 m->m_flags &= ~M_MCAST;
491 * If the source address is not specified yet, use the address
492 * of the outoing interface. In case, keep note we did that, so
493 * if the the firewall changes the next-hop causing the output
494 * interface to change, we can fix that.
496 if (ip->ip_src.s_addr == INADDR_ANY || src_was_INADDR_ANY) {
497 /* Interface may have no addresses. */
499 ip->ip_src = IA_SIN(ia)->sin_addr;
500 src_was_INADDR_ANY = 1;
506 * Disable packet drop hack.
507 * Packetdrop should be done by queueing.
511 * Verify that we have any chance at all of being able to queue
512 * the packet or packet fragments
514 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
515 ifp->if_snd.ifq_maxlen) {
517 ipstat.ips_odropped++;
523 * Look for broadcast address and
524 * verify user is allowed to send
528 if (!(ifp->if_flags & IFF_BROADCAST)) {
529 error = EADDRNOTAVAIL;
532 if (!(flags & IP_ALLOWBROADCAST)) {
536 /* don't allow broadcast messages to be fragmented */
537 if (ip->ip_len > ifp->if_mtu) {
541 m->m_flags |= M_BCAST;
543 m->m_flags &= ~M_BCAST;
548 /* get SP for this packet */
550 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
552 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
555 ipsecstat.out_inval++;
562 switch (sp->policy) {
563 case IPSEC_POLICY_DISCARD:
565 * This packet is just discarded.
567 ipsecstat.out_polvio++;
570 case IPSEC_POLICY_BYPASS:
571 case IPSEC_POLICY_NONE:
572 /* no need to do IPsec. */
575 case IPSEC_POLICY_IPSEC:
576 if (sp->req == NULL) {
577 /* acquire a policy */
578 error = key_spdacquire(sp);
583 case IPSEC_POLICY_ENTRUST:
585 kprintf("ip_output: Invalid policy found. %d\n", sp->policy);
588 struct ipsec_output_state state;
589 bzero(&state, sizeof state);
591 if (flags & IP_ROUTETOIF) {
593 bzero(&iproute, sizeof iproute);
596 state.dst = (struct sockaddr *)dst;
602 * delayed checksums are not currently compatible with IPsec
604 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
606 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
609 ip->ip_len = htons(ip->ip_len);
610 ip->ip_off = htons(ip->ip_off);
612 error = ipsec4_output(&state, sp, flags);
615 if (flags & IP_ROUTETOIF) {
617 * if we have tunnel mode SA, we may need to ignore
620 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
621 flags &= ~IP_ROUTETOIF;
626 dst = (struct sockaddr_in *)state.dst;
628 /* mbuf is already reclaimed in ipsec4_output. */
638 kprintf("ip4_output (ipsec): error code %d\n", error);
641 /* don't show these error codes to the user */
649 /* be sure to update variables that are affected by ipsec4_output() */
650 ip = mtod(m, struct ip *);
652 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
654 hlen = ip->ip_hl << 2;
656 if (ro->ro_rt == NULL) {
657 if (!(flags & IP_ROUTETOIF)) {
658 kprintf("ip_output: "
659 "can't update route after IPsec processing\n");
660 error = EHOSTUNREACH; /*XXX*/
664 ia = ifatoia(ro->ro_rt->rt_ifa);
665 ifp = ro->ro_rt->rt_ifp;
668 /* make it flipped, again. */
669 ip->ip_len = ntohs(ip->ip_len);
670 ip->ip_off = ntohs(ip->ip_off);
675 * Check the security policy (SP) for the packet and, if
676 * required, do IPsec-related processing. There are two
677 * cases here; the first time a packet is sent through
678 * it will be untagged and handled by ipsec4_checkpolicy.
679 * If the packet is resubmitted to ip_output (e.g. after
680 * AH, ESP, etc. processing), there will be a tag to bypass
681 * the lookup and related policy checking.
683 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
686 tdbi = (struct tdb_ident *)m_tag_data(mtag);
687 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
689 error = -EINVAL; /* force silent drop */
690 m_tag_delete(m, mtag);
692 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
696 * There are four return cases:
697 * sp != NULL apply IPsec policy
698 * sp == NULL, error == 0 no IPsec handling needed
699 * sp == NULL, error == -EINVAL discard packet w/o error
700 * sp == NULL, error != 0 discard packet, report error
703 /* Loop detection, check if ipsec processing already done */
704 KASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
705 for (mtag = m_tag_first(m); mtag != NULL;
706 mtag = m_tag_next(m, mtag)) {
707 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
709 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
710 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
713 * Check if policy has an SA associated with it.
714 * This can happen when an SP has yet to acquire
715 * an SA; e.g. on first reference. If it occurs,
716 * then we let ipsec4_process_packet do its thing.
718 if (sp->req->sav == NULL)
720 tdbi = (struct tdb_ident *)m_tag_data(mtag);
721 if (tdbi->spi == sp->req->sav->spi &&
722 tdbi->proto == sp->req->sav->sah->saidx.proto &&
723 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
724 sizeof(union sockaddr_union)) == 0) {
726 * No IPsec processing is needed, free
729 * NB: null pointer to avoid free at
732 KEY_FREESP(&sp), sp = NULL;
739 * Do delayed checksums now because we send before
740 * this is done in the normal processing path.
742 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
744 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
747 ip->ip_len = htons(ip->ip_len);
748 ip->ip_off = htons(ip->ip_off);
750 /* NB: callee frees mbuf */
751 error = ipsec4_process_packet(m, sp->req, flags, 0);
753 * Preserve KAME behaviour: ENOENT can be returned
754 * when an SA acquire is in progress. Don't propagate
755 * this to user-level; it confuses applications.
757 * XXX this will go away when the SADB is redone.
768 * Hack: -EINVAL is used to signal that a packet
769 * should be silently discarded. This is typically
770 * because we asked key management for an SA and
771 * it was delayed (e.g. kicked up to IKE).
773 if (error == -EINVAL)
777 /* No IPsec processing for this packet. */
781 * If deferred crypto processing is needed, check that
782 * the interface supports it.
784 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
785 if (mtag != NULL && !(ifp->if_capenable & IFCAP_IPSEC)) {
786 /* notify IPsec to do its own crypto */
787 ipsp_skipcrypto_unmark((struct tdb_ident *)m_tag_data(mtag));
788 error = EHOSTUNREACH;
794 #endif /* FAST_IPSEC */
796 /* We are already being fwd'd from a firewall. */
797 if (next_hop != NULL)
801 if (!pfil_has_hooks(&inet_pfil_hook)) {
802 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
804 * Strip dummynet tags from stranded packets
806 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
807 KKASSERT(mtag != NULL);
808 m_tag_delete(m, mtag);
809 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
816 * - Xlate: translate packet's addr/port (NAT).
817 * - Firewall: deny/allow/etc.
818 * - Wrap: fake packet's addr/port <unimpl.>
819 * - Encapsulate: put it in another IP and send out. <unimp.>
823 * Run through list of hooks for output packets.
825 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT);
826 if (error != 0 || m == NULL)
828 ip = mtod(m, struct ip *);
830 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
832 * Check dst to make sure it is directly reachable on the
833 * interface we previously thought it was.
834 * If it isn't (which may be likely in some situations) we have
835 * to re-route it (ie, find a route for the next-hop and the
836 * associated interface) and set them here. This is nested
837 * forwarding which in most cases is undesirable, except where
838 * such control is nigh impossible. So we do it here.
841 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
842 KKASSERT(mtag != NULL);
843 next_hop = m_tag_data(mtag);
846 * Try local forwarding first
848 if (ip_localforward(m, next_hop, hlen))
852 * Relocate the route based on next_hop.
853 * If the current route is inp's cache, keep it untouched.
855 if (ro == &iproute && ro->ro_rt != NULL) {
860 bzero(ro, sizeof *ro);
863 * Forwarding to broadcast address is not allowed.
864 * XXX Should we follow IP_ROUTETOIF?
866 flags &= ~(IP_ALLOWBROADCAST | IP_ROUTETOIF);
868 /* We are doing forwarding now */
869 flags |= IP_FORWARDING;
874 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
875 struct dn_pkt *dn_pkt;
877 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
878 KKASSERT(mtag != NULL);
879 dn_pkt = m_tag_data(mtag);
882 * Under certain cases it is not possible to recalculate
883 * 'ro' and 'dst', let alone 'flags', so just save them in
884 * dummynet tag and avoid the possible wrong reculcalation
885 * when we come back to ip_output() again.
887 * All other parameters have been already used and so they
888 * are not needed anymore.
889 * XXX if the ifp is deleted while a pkt is in dummynet,
890 * we are in trouble! (TODO use ifnet_detach_event)
892 * We need to copy *ro because for ICMP pkts (and maybe
893 * others) the caller passed a pointer into the stack;
894 * dst might also be a pointer into *ro so it needs to
899 ro->ro_rt->rt_refcnt++;
900 if (dst == (struct sockaddr_in *)&ro->ro_dst) {
901 /* 'dst' points into 'ro' */
902 dst = (struct sockaddr_in *)&(dn_pkt->ro.ro_dst);
904 dn_pkt->dn_dst = dst;
905 dn_pkt->flags = flags;
911 /* 127/8 must not appear on wire - RFC1122. */
912 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
913 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
914 if (!(ifp->if_flags & IFF_LOOPBACK)) {
915 ipstat.ips_badaddr++;
916 error = EADDRNOTAVAIL;
921 m->m_pkthdr.csum_flags |= CSUM_IP;
922 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
923 if (sw_csum & CSUM_DELAY_DATA) {
925 sw_csum &= ~CSUM_DELAY_DATA;
927 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
930 * If small enough for interface, or the interface will take
931 * care of the fragmentation for us, can just send directly.
933 if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
934 !(ip->ip_off & IP_DF))) {
935 ip->ip_len = htons(ip->ip_len);
936 ip->ip_off = htons(ip->ip_off);
938 if (sw_csum & CSUM_DELAY_IP) {
939 if (ip->ip_vhl == IP_VHL_BORING)
940 ip->ip_sum = in_cksum_hdr(ip);
942 ip->ip_sum = in_cksum(m, hlen);
945 /* Record statistics for this interface address. */
946 if (!(flags & IP_FORWARDING) && ia) {
947 ia->ia_ifa.if_opackets++;
948 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
952 /* clean ipsec history once it goes out of the node */
956 #ifdef MBUF_STRESS_TEST
957 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) {
958 struct mbuf *m1, *m2;
961 tmp = length = m->m_pkthdr.len;
963 while ((length -= mbuf_frag_size) >= 1) {
964 m1 = m_split(m, length, MB_DONTWAIT);
968 while (m2->m_next != NULL)
972 m->m_pkthdr.len = tmp;
977 if (!mpls_output_process(m, ro->ro_rt))
980 error = ifp->if_output(ifp, m, (struct sockaddr *)dst,
985 if (ip->ip_off & IP_DF) {
988 * This case can happen if the user changed the MTU
989 * of an interface after enabling IP on it. Because
990 * most netifs don't keep track of routes pointing to
991 * them, there is no way for one to update all its
992 * routes when the MTU is changed.
994 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
995 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
996 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
997 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
999 ipstat.ips_cantfrag++;
1004 * Too large for interface; fragment if possible. If successful,
1005 * on return, m will point to a list of packets to be sent.
1007 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
1012 m->m_nextpkt = NULL;
1014 /* clean ipsec history once it goes out of the node */
1018 /* Record statistics for this interface address. */
1020 ia->ia_ifa.if_opackets++;
1021 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1024 if (!mpls_output_process(m, ro->ro_rt))
1027 error = ifp->if_output(ifp, m, (struct sockaddr *)dst,
1035 ipstat.ips_fragmented++;
1038 if (ro == &iproute && ro->ro_rt != NULL) {
1044 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1045 kprintf("DP ip_output call free SP:%p\n", sp));
1060 * Create a chain of fragments which fit the given mtu. m_frag points to the
1061 * mbuf to be fragmented; on return it points to the chain with the fragments.
1062 * Return 0 if no error. If error, m_frag may contain a partially built
1063 * chain of fragments that should be freed by the caller.
1065 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
1066 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
1069 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
1070 u_long if_hwassist_flags, int sw_csum)
1073 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1074 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
1076 struct mbuf *m0 = *m_frag; /* the original packet */
1078 struct mbuf **mnext;
1081 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
1082 ipstat.ips_cantfrag++;
1087 * Must be able to put at least 8 bytes per fragment.
1093 * If the interface will not calculate checksums on
1094 * fragmented packets, then do it here.
1096 if ((m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) &&
1097 !(if_hwassist_flags & CSUM_IP_FRAGS)) {
1098 in_delayed_cksum(m0);
1099 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1102 if (len > PAGE_SIZE) {
1104 * Fragment large datagrams such that each segment
1105 * contains a multiple of PAGE_SIZE amount of data,
1106 * plus headers. This enables a receiver to perform
1107 * page-flipping zero-copy optimizations.
1109 * XXX When does this help given that sender and receiver
1110 * could have different page sizes, and also mtu could
1111 * be less than the receiver's page size ?
1116 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
1120 * firstlen (off - hlen) must be aligned on an
1124 goto smart_frag_failure;
1125 off = ((off - hlen) & ~7) + hlen;
1126 newlen = (~PAGE_MASK) & mtu;
1127 if ((newlen + sizeof(struct ip)) > mtu) {
1128 /* we failed, go back the default */
1139 firstlen = off - hlen;
1140 mnext = &m0->m_nextpkt; /* pointer to next packet */
1143 * Loop through length of segment after first fragment,
1144 * make new header and copy data of each part and link onto chain.
1145 * Here, m0 is the original packet, m is the fragment being created.
1146 * The fragments are linked off the m_nextpkt of the original
1147 * packet, which after processing serves as the first fragment.
1149 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
1150 struct ip *mhip; /* ip header on the fragment */
1152 int mhlen = sizeof(struct ip);
1154 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1157 ipstat.ips_odropped++;
1160 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
1162 * In the first mbuf, leave room for the link header, then
1163 * copy the original IP header including options. The payload
1164 * goes into an additional mbuf chain returned by m_copy().
1166 m->m_data += max_linkhdr;
1167 mhip = mtod(m, struct ip *);
1169 if (hlen > sizeof(struct ip)) {
1170 mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip);
1171 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
1174 /* XXX do we need to add ip->ip_off below ? */
1175 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
1176 if (off + len >= ip->ip_len) { /* last fragment */
1177 len = ip->ip_len - off;
1178 m->m_flags |= M_LASTFRAG;
1180 mhip->ip_off |= IP_MF;
1181 mhip->ip_len = htons((u_short)(len + mhlen));
1182 m->m_next = m_copy(m0, off, len);
1183 if (m->m_next == NULL) { /* copy failed */
1185 error = ENOBUFS; /* ??? */
1186 ipstat.ips_odropped++;
1189 m->m_pkthdr.len = mhlen + len;
1190 m->m_pkthdr.rcvif = (struct ifnet *)NULL;
1191 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1192 mhip->ip_off = htons(mhip->ip_off);
1194 if (sw_csum & CSUM_DELAY_IP)
1195 mhip->ip_sum = in_cksum(m, mhlen);
1197 mnext = &m->m_nextpkt;
1199 ipstat.ips_ofragments += nfrags;
1201 /* set first marker for fragment chain */
1202 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1203 m0->m_pkthdr.csum_data = nfrags;
1206 * Update first fragment by trimming what's been copied out
1207 * and updating header.
1209 m_adj(m0, hlen + firstlen - ip->ip_len);
1210 m0->m_pkthdr.len = hlen + firstlen;
1211 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1212 ip->ip_off |= IP_MF;
1213 ip->ip_off = htons(ip->ip_off);
1215 if (sw_csum & CSUM_DELAY_IP)
1216 ip->ip_sum = in_cksum(m0, hlen);
1224 in_delayed_cksum(struct mbuf *m)
1227 u_short csum, offset;
1229 ip = mtod(m, struct ip *);
1230 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1231 csum = in_cksum_skip(m, ip->ip_len, offset);
1232 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1234 offset += m->m_pkthdr.csum_data; /* checksum offset */
1236 if (offset + sizeof(u_short) > m->m_len) {
1237 kprintf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1238 m->m_len, offset, ip->ip_p);
1241 * this shouldn't happen, but if it does, the
1242 * correct behavior may be to insert the checksum
1243 * in the existing chain instead of rearranging it.
1245 m = m_pullup(m, offset + sizeof(u_short));
1247 *(u_short *)(m->m_data + offset) = csum;
1251 * Insert IP options into preformed packet.
1252 * Adjust IP destination as required for IP source routing,
1253 * as indicated by a non-zero in_addr at the start of the options.
1255 * XXX This routine assumes that the packet has no options in place.
1257 static struct mbuf *
1258 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
1260 struct ipoption *p = mtod(opt, struct ipoption *);
1262 struct ip *ip = mtod(m, struct ip *);
1265 optlen = opt->m_len - sizeof p->ipopt_dst;
1266 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
1268 return (m); /* XXX should fail */
1270 if (p->ipopt_dst.s_addr)
1271 ip->ip_dst = p->ipopt_dst;
1272 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1273 MGETHDR(n, MB_DONTWAIT, MT_HEADER);
1278 n->m_pkthdr.rcvif = (struct ifnet *)NULL;
1279 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1280 m->m_len -= sizeof(struct ip);
1281 m->m_data += sizeof(struct ip);
1284 m->m_len = optlen + sizeof(struct ip);
1285 m->m_data += max_linkhdr;
1286 memcpy(mtod(m, void *), ip, sizeof(struct ip));
1288 m->m_data -= optlen;
1290 m->m_pkthdr.len += optlen;
1291 ovbcopy(ip, mtod(m, caddr_t), sizeof(struct ip));
1293 ip = mtod(m, struct ip *);
1294 bcopy(p->ipopt_list, ip + 1, optlen);
1295 *phlen = sizeof(struct ip) + optlen;
1296 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1297 ip->ip_len += optlen;
1302 * Copy options from ip to jp,
1303 * omitting those not copied during fragmentation.
1306 ip_optcopy(struct ip *ip, struct ip *jp)
1309 int opt, optlen, cnt;
1311 cp = (u_char *)(ip + 1);
1312 dp = (u_char *)(jp + 1);
1313 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1314 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1316 if (opt == IPOPT_EOL)
1318 if (opt == IPOPT_NOP) {
1319 /* Preserve for IP mcast tunnel's LSRR alignment. */
1325 KASSERT(cnt >= IPOPT_OLEN + sizeof *cp,
1326 ("ip_optcopy: malformed ipv4 option"));
1327 optlen = cp[IPOPT_OLEN];
1328 KASSERT(optlen >= IPOPT_OLEN + sizeof *cp && optlen <= cnt,
1329 ("ip_optcopy: malformed ipv4 option"));
1331 /* bogus lengths should have been caught by ip_dooptions */
1334 if (IPOPT_COPIED(opt)) {
1335 bcopy(cp, dp, optlen);
1339 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1345 * IP socket option processing.
1348 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1350 struct inpcb *inp = so->so_pcb;
1354 if (sopt->sopt_level != IPPROTO_IP) {
1358 switch (sopt->sopt_dir) {
1360 switch (sopt->sopt_name) {
1367 if (sopt->sopt_valsize > MLEN) {
1371 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1376 m->m_len = sopt->sopt_valsize;
1377 error = soopt_to_kbuf(sopt, mtod(m, void *), m->m_len,
1379 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1387 case IP_RECVRETOPTS:
1388 case IP_RECVDSTADDR:
1392 error = soopt_to_kbuf(sopt, &optval, sizeof optval,
1396 switch (sopt->sopt_name) {
1398 inp->inp_ip_tos = optval;
1402 inp->inp_ip_ttl = optval;
1405 if (optval > 0 && optval <= MAXTTL)
1406 inp->inp_ip_minttl = optval;
1410 #define OPTSET(bit) \
1412 inp->inp_flags |= bit; \
1414 inp->inp_flags &= ~bit;
1417 OPTSET(INP_RECVOPTS);
1420 case IP_RECVRETOPTS:
1421 OPTSET(INP_RECVRETOPTS);
1424 case IP_RECVDSTADDR:
1425 OPTSET(INP_RECVDSTADDR);
1433 OPTSET(INP_RECVTTL);
1443 case IP_MULTICAST_IF:
1444 case IP_MULTICAST_VIF:
1445 case IP_MULTICAST_TTL:
1446 case IP_MULTICAST_LOOP:
1447 case IP_ADD_MEMBERSHIP:
1448 case IP_DROP_MEMBERSHIP:
1449 error = ip_setmoptions(sopt, &inp->inp_moptions);
1453 error = soopt_to_kbuf(sopt, &optval, sizeof optval,
1459 case IP_PORTRANGE_DEFAULT:
1460 inp->inp_flags &= ~(INP_LOWPORT);
1461 inp->inp_flags &= ~(INP_HIGHPORT);
1464 case IP_PORTRANGE_HIGH:
1465 inp->inp_flags &= ~(INP_LOWPORT);
1466 inp->inp_flags |= INP_HIGHPORT;
1469 case IP_PORTRANGE_LOW:
1470 inp->inp_flags &= ~(INP_HIGHPORT);
1471 inp->inp_flags |= INP_LOWPORT;
1480 #if defined(IPSEC) || defined(FAST_IPSEC)
1481 case IP_IPSEC_POLICY:
1489 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1491 soopt_to_mbuf(sopt, m);
1492 priv = (sopt->sopt_td != NULL &&
1493 priv_check(sopt->sopt_td, PRIV_ROOT) != 0) ? 0 : 1;
1494 req = mtod(m, caddr_t);
1496 optname = sopt->sopt_name;
1497 error = ipsec4_set_policy(inp, optname, req, len, priv);
1504 error = ENOPROTOOPT;
1510 switch (sopt->sopt_name) {
1513 if (inp->inp_options)
1514 soopt_from_kbuf(sopt, mtod(inp->inp_options,
1516 inp->inp_options->m_len);
1518 sopt->sopt_valsize = 0;
1525 case IP_RECVRETOPTS:
1526 case IP_RECVDSTADDR:
1531 switch (sopt->sopt_name) {
1534 optval = inp->inp_ip_tos;
1538 optval = inp->inp_ip_ttl;
1541 optval = inp->inp_ip_minttl;
1544 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1547 optval = OPTBIT(INP_RECVOPTS);
1550 case IP_RECVRETOPTS:
1551 optval = OPTBIT(INP_RECVRETOPTS);
1554 case IP_RECVDSTADDR:
1555 optval = OPTBIT(INP_RECVDSTADDR);
1559 optval = OPTBIT(INP_RECVTTL);
1563 optval = OPTBIT(INP_RECVIF);
1567 if (inp->inp_flags & INP_HIGHPORT)
1568 optval = IP_PORTRANGE_HIGH;
1569 else if (inp->inp_flags & INP_LOWPORT)
1570 optval = IP_PORTRANGE_LOW;
1576 optval = OPTBIT(INP_FAITH);
1579 soopt_from_kbuf(sopt, &optval, sizeof optval);
1582 case IP_MULTICAST_IF:
1583 case IP_MULTICAST_VIF:
1584 case IP_MULTICAST_TTL:
1585 case IP_MULTICAST_LOOP:
1586 case IP_ADD_MEMBERSHIP:
1587 case IP_DROP_MEMBERSHIP:
1588 error = ip_getmoptions(sopt, inp->inp_moptions);
1591 #if defined(IPSEC) || defined(FAST_IPSEC)
1592 case IP_IPSEC_POLICY:
1594 struct mbuf *m = NULL;
1599 req = mtod(m, caddr_t);
1602 error = ipsec4_get_policy(so->so_pcb, req, len, &m);
1604 error = soopt_from_mbuf(sopt, m); /* XXX */
1612 error = ENOPROTOOPT;
1621 * Set up IP options in pcb for insertion in output packets.
1622 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1623 * with destination address if source routed.
1626 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
1632 /* turn off any old options */
1636 if (m == NULL || m->m_len == 0) {
1638 * Only turning off any previous options.
1645 if (m->m_len % sizeof(int32_t))
1648 * IP first-hop destination address will be stored before
1649 * actual options; move other options back
1650 * and clear it when none present.
1652 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1655 m->m_len += sizeof(struct in_addr);
1656 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1657 ovbcopy(mtod(m, caddr_t), cp, cnt);
1658 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1660 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1661 opt = cp[IPOPT_OPTVAL];
1662 if (opt == IPOPT_EOL)
1664 if (opt == IPOPT_NOP)
1667 if (cnt < IPOPT_OLEN + sizeof *cp)
1669 optlen = cp[IPOPT_OLEN];
1670 if (optlen < IPOPT_OLEN + sizeof *cp || optlen > cnt)
1681 * user process specifies route as:
1683 * D must be our final destination (but we can't
1684 * check that since we may not have connected yet).
1685 * A is first hop destination, which doesn't appear in
1686 * actual IP option, but is stored before the options.
1688 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1690 m->m_len -= sizeof(struct in_addr);
1691 cnt -= sizeof(struct in_addr);
1692 optlen -= sizeof(struct in_addr);
1693 cp[IPOPT_OLEN] = optlen;
1695 * Move first hop before start of options.
1697 bcopy(&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1698 sizeof(struct in_addr));
1700 * Then copy rest of options back
1701 * to close up the deleted entry.
1703 ovbcopy(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr),
1704 &cp[IPOPT_OFFSET+1],
1705 cnt - (IPOPT_MINOFF - 1));
1709 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1721 * The whole multicast option thing needs to be re-thought.
1722 * Several of these options are equally applicable to non-multicast
1723 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1724 * standard option (IP_TTL).
1728 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1730 static struct ifnet *
1731 ip_multicast_if(struct in_addr *a, int *ifindexp)
1738 if (ntohl(a->s_addr) >> 24 == 0) {
1739 ifindex = ntohl(a->s_addr) & 0xffffff;
1740 if (ifindex < 0 || if_index < ifindex)
1742 ifp = ifindex2ifnet[ifindex];
1744 *ifindexp = ifindex;
1746 ifp = INADDR_TO_IFP(a);
1752 * Set the IP multicast options in response to user setsockopt().
1755 ip_setmoptions(struct sockopt *sopt, struct ip_moptions **imop)
1759 struct in_addr addr;
1760 struct ip_mreq mreq;
1762 struct ip_moptions *imo = *imop;
1767 * No multicast option buffer attached to the pcb;
1768 * allocate one and initialize to default values.
1770 imo = kmalloc(sizeof *imo, M_IPMOPTS, M_WAITOK);
1773 imo->imo_multicast_ifp = NULL;
1774 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1775 imo->imo_multicast_vif = -1;
1776 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1777 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1778 imo->imo_num_memberships = 0;
1780 switch (sopt->sopt_name) {
1781 /* store an index number for the vif you wanna use in the send */
1782 case IP_MULTICAST_VIF:
1783 if (legal_vif_num == 0) {
1787 error = soopt_to_kbuf(sopt, &i, sizeof i, sizeof i);
1790 if (!legal_vif_num(i) && (i != -1)) {
1794 imo->imo_multicast_vif = i;
1797 case IP_MULTICAST_IF:
1799 * Select the interface for outgoing multicast packets.
1801 error = soopt_to_kbuf(sopt, &addr, sizeof addr, sizeof addr);
1806 * INADDR_ANY is used to remove a previous selection.
1807 * When no interface is selected, a default one is
1808 * chosen every time a multicast packet is sent.
1810 if (addr.s_addr == INADDR_ANY) {
1811 imo->imo_multicast_ifp = NULL;
1815 * The selected interface is identified by its local
1816 * IP address. Find the interface and confirm that
1817 * it supports multicasting.
1820 ifp = ip_multicast_if(&addr, &ifindex);
1821 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1823 error = EADDRNOTAVAIL;
1826 imo->imo_multicast_ifp = ifp;
1828 imo->imo_multicast_addr = addr;
1830 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1834 case IP_MULTICAST_TTL:
1836 * Set the IP time-to-live for outgoing multicast packets.
1837 * The original multicast API required a char argument,
1838 * which is inconsistent with the rest of the socket API.
1839 * We allow either a char or an int.
1841 if (sopt->sopt_valsize == 1) {
1843 error = soopt_to_kbuf(sopt, &ttl, 1, 1);
1846 imo->imo_multicast_ttl = ttl;
1849 error = soopt_to_kbuf(sopt, &ttl, sizeof ttl, sizeof ttl);
1855 imo->imo_multicast_ttl = ttl;
1859 case IP_MULTICAST_LOOP:
1861 * Set the loopback flag for outgoing multicast packets.
1862 * Must be zero or one. The original multicast API required a
1863 * char argument, which is inconsistent with the rest
1864 * of the socket API. We allow either a char or an int.
1866 if (sopt->sopt_valsize == 1) {
1869 error = soopt_to_kbuf(sopt, &loop, 1, 1);
1872 imo->imo_multicast_loop = !!loop;
1876 error = soopt_to_kbuf(sopt, &loop, sizeof loop,
1880 imo->imo_multicast_loop = !!loop;
1884 case IP_ADD_MEMBERSHIP:
1886 * Add a multicast group membership.
1887 * Group must be a valid IP multicast address.
1889 error = soopt_to_kbuf(sopt, &mreq, sizeof mreq, sizeof mreq);
1893 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1899 * If no interface address was provided, use the interface of
1900 * the route to the given multicast address.
1902 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1903 struct sockaddr_in dst;
1906 bzero(&dst, sizeof(struct sockaddr_in));
1907 dst.sin_len = sizeof(struct sockaddr_in);
1908 dst.sin_family = AF_INET;
1909 dst.sin_addr = mreq.imr_multiaddr;
1910 rt = rtlookup((struct sockaddr *)&dst);
1912 error = EADDRNOTAVAIL;
1919 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1923 * See if we found an interface, and confirm that it
1924 * supports multicast.
1926 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1927 error = EADDRNOTAVAIL;
1932 * See if the membership already exists or if all the
1933 * membership slots are full.
1935 for (i = 0; i < imo->imo_num_memberships; ++i) {
1936 if (imo->imo_membership[i]->inm_ifp == ifp &&
1937 imo->imo_membership[i]->inm_addr.s_addr
1938 == mreq.imr_multiaddr.s_addr)
1941 if (i < imo->imo_num_memberships) {
1946 if (i == IP_MAX_MEMBERSHIPS) {
1947 error = ETOOMANYREFS;
1952 * Everything looks good; add a new record to the multicast
1953 * address list for the given interface.
1955 if ((imo->imo_membership[i] =
1956 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
1961 ++imo->imo_num_memberships;
1965 case IP_DROP_MEMBERSHIP:
1967 * Drop a multicast group membership.
1968 * Group must be a valid IP multicast address.
1970 error = soopt_to_kbuf(sopt, &mreq, sizeof mreq, sizeof mreq);
1974 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1981 * If an interface address was specified, get a pointer
1982 * to its ifnet structure.
1984 if (mreq.imr_interface.s_addr == INADDR_ANY)
1987 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1989 error = EADDRNOTAVAIL;
1995 * Find the membership in the membership array.
1997 for (i = 0; i < imo->imo_num_memberships; ++i) {
1999 imo->imo_membership[i]->inm_ifp == ifp) &&
2000 imo->imo_membership[i]->inm_addr.s_addr ==
2001 mreq.imr_multiaddr.s_addr)
2004 if (i == imo->imo_num_memberships) {
2005 error = EADDRNOTAVAIL;
2010 * Give up the multicast address record to which the
2011 * membership points.
2013 in_delmulti(imo->imo_membership[i]);
2015 * Remove the gap in the membership array.
2017 for (++i; i < imo->imo_num_memberships; ++i)
2018 imo->imo_membership[i-1] = imo->imo_membership[i];
2019 --imo->imo_num_memberships;
2029 * If all options have default values, no need to keep the mbuf.
2031 if (imo->imo_multicast_ifp == NULL &&
2032 imo->imo_multicast_vif == -1 &&
2033 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2034 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2035 imo->imo_num_memberships == 0) {
2036 kfree(*imop, M_IPMOPTS);
2044 * Return the IP multicast options in response to user getsockopt().
2047 ip_getmoptions(struct sockopt *sopt, struct ip_moptions *imo)
2049 struct in_addr addr;
2050 struct in_ifaddr *ia;
2055 switch (sopt->sopt_name) {
2056 case IP_MULTICAST_VIF:
2058 optval = imo->imo_multicast_vif;
2061 soopt_from_kbuf(sopt, &optval, sizeof optval);
2064 case IP_MULTICAST_IF:
2065 if (imo == NULL || imo->imo_multicast_ifp == NULL)
2066 addr.s_addr = INADDR_ANY;
2067 else if (imo->imo_multicast_addr.s_addr) {
2068 /* return the value user has set */
2069 addr = imo->imo_multicast_addr;
2071 ia = IFP_TO_IA(imo->imo_multicast_ifp);
2072 addr.s_addr = (ia == NULL) ? INADDR_ANY
2073 : IA_SIN(ia)->sin_addr.s_addr;
2075 soopt_from_kbuf(sopt, &addr, sizeof addr);
2078 case IP_MULTICAST_TTL:
2080 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
2082 optval = coptval = imo->imo_multicast_ttl;
2083 if (sopt->sopt_valsize == 1)
2084 soopt_from_kbuf(sopt, &coptval, 1);
2086 soopt_from_kbuf(sopt, &optval, sizeof optval);
2089 case IP_MULTICAST_LOOP:
2091 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
2093 optval = coptval = imo->imo_multicast_loop;
2094 if (sopt->sopt_valsize == 1)
2095 soopt_from_kbuf(sopt, &coptval, 1);
2097 soopt_from_kbuf(sopt, &optval, sizeof optval);
2101 error = ENOPROTOOPT;
2108 * Discard the IP multicast options.
2111 ip_freemoptions(struct ip_moptions *imo)
2116 for (i = 0; i < imo->imo_num_memberships; ++i)
2117 in_delmulti(imo->imo_membership[i]);
2118 kfree(imo, M_IPMOPTS);
2123 * Routine called from ip_output() to loop back a copy of an IP multicast
2124 * packet to the input queue of a specified interface. Note that this
2125 * calls the output routine of the loopback "driver", but with an interface
2126 * pointer that might NOT be a loopback interface -- evil, but easier than
2127 * replicating that code here.
2130 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
2136 copym = m_copypacket(m, MB_DONTWAIT);
2137 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2138 copym = m_pullup(copym, hlen);
2139 if (copym != NULL) {
2141 * if the checksum hasn't been computed, mark it as valid
2143 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2144 in_delayed_cksum(copym);
2145 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2146 copym->m_pkthdr.csum_flags |=
2147 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2148 copym->m_pkthdr.csum_data = 0xffff;
2151 * We don't bother to fragment if the IP length is greater
2152 * than the interface's MTU. Can this possibly matter?
2154 ip = mtod(copym, struct ip *);
2155 ip->ip_len = htons(ip->ip_len);
2156 ip->ip_off = htons(ip->ip_off);
2158 if (ip->ip_vhl == IP_VHL_BORING) {
2159 ip->ip_sum = in_cksum_hdr(ip);
2161 ip->ip_sum = in_cksum(copym, hlen);
2165 * It's not clear whether there are any lingering
2166 * reentrancy problems in other areas which might
2167 * be exposed by using ip_input directly (in
2168 * particular, everything which modifies the packet
2169 * in-place). Yet another option is using the
2170 * protosw directly to deliver the looped back
2171 * packet. For the moment, we'll err on the side
2172 * of safety by using if_simloop().
2175 if (dst->sin_family != AF_INET) {
2176 kprintf("ip_mloopback: bad address family %d\n",
2178 dst->sin_family = AF_INET;
2181 if_simloop(ifp, copym, dst->sin_family, 0);