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>
53 #include <sys/sysctl.h>
54 #include <sys/thread2.h>
55 #include <sys/in_cksum.h>
59 #include <net/netisr.h>
61 #include <net/route.h>
63 #include <netinet/in.h>
64 #include <netinet/in_systm.h>
65 #include <netinet/ip.h>
66 #include <netinet/in_pcb.h>
67 #include <netinet/in_var.h>
68 #include <netinet/ip_var.h>
70 #include <netproto/mpls/mpls_var.h>
72 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
75 #include <netinet6/ipsec.h>
76 #include <netproto/key/key.h>
78 #include <netproto/key/key_debug.h>
80 #define KEYDEBUG(lev,arg)
85 #include <netproto/ipsec/ipsec.h>
86 #include <netproto/ipsec/xform.h>
87 #include <netproto/ipsec/key.h>
90 #include <net/ipfw/ip_fw.h>
91 #include <net/dummynet/ip_dummynet.h>
93 #define print_ip(x, a, y) kprintf("%s %d.%d.%d.%d%s",\
94 x, (ntohl(a.s_addr)>>24)&0xFF,\
95 (ntohl(a.s_addr)>>16)&0xFF,\
96 (ntohl(a.s_addr)>>8)&0xFF,\
97 (ntohl(a.s_addr))&0xFF, y);
101 #ifdef MBUF_STRESS_TEST
102 int mbuf_frag_size = 0;
103 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
104 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
107 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
108 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
109 static void ip_mloopback
110 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
111 static int ip_getmoptions
112 (struct sockopt *, struct ip_moptions *);
113 static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
114 static int ip_setmoptions
115 (struct sockopt *, struct ip_moptions **);
117 int ip_optcopy(struct ip *, struct ip *);
119 extern int route_assert_owner_access;
121 extern struct protosw inetsw[];
124 ip_localforward(struct mbuf *m, const struct sockaddr_in *dst, int hlen)
126 struct in_ifaddr_container *iac;
129 * We need to figure out if we have been forwarded to a local
130 * socket. If so, then we should somehow "loop back" to
131 * ip_input(), and get directed to the PCB as if we had received
132 * this packet. This is because it may be difficult to identify
133 * the packets you want to forward until they are being output
134 * and have selected an interface (e.g. locally initiated
135 * packets). If we used the loopback inteface, we would not be
136 * able to control what happens as the packet runs through
137 * ip_input() as it is done through a ISR.
139 LIST_FOREACH(iac, INADDR_HASH(dst->sin_addr.s_addr), ia_hash) {
141 * If the addr to forward to is one of ours, we pretend
142 * to be the destination for this packet.
144 if (IA_SIN(iac->ia)->sin_addr.s_addr == dst->sin_addr.s_addr)
150 if (m->m_pkthdr.rcvif == NULL)
151 m->m_pkthdr.rcvif = ifunit("lo0");
152 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
153 m->m_pkthdr.csum_flags |= CSUM_DATA_VALID |
155 m->m_pkthdr.csum_data = 0xffff;
157 m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID;
160 * Make sure that the IP header is in one mbuf,
161 * required by ip_input
163 if (m->m_len < hlen) {
164 m = m_pullup(m, hlen);
166 /* The packet was freed; we are done */
170 ip = mtod(m, struct ip *);
172 ip->ip_len = htons(ip->ip_len);
173 ip->ip_off = htons(ip->ip_off);
176 return 1; /* The packet gets forwarded locally */
182 * IP output. The packet in mbuf chain m contains a skeletal IP
183 * header (with len, off, ttl, proto, tos, src, dst).
184 * The mbuf chain containing the packet will be freed.
185 * The mbuf opt, if present, will not be freed.
188 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
189 int flags, struct ip_moptions *imo, struct inpcb *inp)
192 struct ifnet *ifp = NULL; /* keep compiler happy */
194 int hlen = sizeof(struct ip);
196 struct sockaddr_in *dst = NULL; /* keep compiler happy */
197 struct in_ifaddr *ia = NULL;
198 int isbroadcast, sw_csum;
199 struct in_addr pkt_dst;
200 struct route iproute;
203 struct secpolicy *sp = NULL;
204 struct socket *so = inp ? inp->inp_socket : NULL;
207 struct secpolicy *sp = NULL;
208 struct tdb_ident *tdbi;
209 #endif /* FAST_IPSEC */
210 struct sockaddr_in *next_hop = NULL;
211 int src_was_INADDR_ANY = 0; /* as the name says... */
218 bzero(ro, sizeof *ro);
219 } else if (ro->ro_rt != NULL && ro->ro_rt->rt_cpuid != mycpuid) {
220 if (flags & IP_DEBUGROUTE) {
221 if (route_assert_owner_access) {
223 "rt rt_cpuid %d accessed on cpu %d\n",
224 ro->ro_rt->rt_cpuid, mycpuid);
226 kprintf("ip_output: "
227 "rt rt_cpuid %d accessed on cpu %d\n",
228 ro->ro_rt->rt_cpuid, mycpuid);
235 * If the cached rtentry's owner CPU is not the current CPU,
236 * then don't touch the cached rtentry (remote free is too
237 * expensive in this context); just relocate the route.
240 bzero(ro, sizeof *ro);
243 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
245 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
246 KKASSERT(mtag != NULL);
247 next_hop = m_tag_data(mtag);
250 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
251 struct dn_pkt *dn_pkt;
253 /* Extract info from dummynet tag */
254 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
255 KKASSERT(mtag != NULL);
256 dn_pkt = m_tag_data(mtag);
259 * The packet was already tagged, so part of the
260 * processing was already done, and we need to go down.
261 * Get the calculated parameters from the tag.
265 KKASSERT(ro == &iproute);
266 *ro = dn_pkt->ro; /* structure copy */
267 KKASSERT(ro->ro_rt == NULL || ro->ro_rt->rt_cpuid == mycpuid);
269 dst = dn_pkt->dn_dst;
270 if (dst == (struct sockaddr_in *)&(dn_pkt->ro.ro_dst)) {
271 /* If 'dst' points into dummynet tag, adjust it */
272 dst = (struct sockaddr_in *)&(ro->ro_dst);
275 ip = mtod(m, struct ip *);
276 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
278 ia = ifatoia(ro->ro_rt->rt_ifa);
284 m = ip_insertoptions(m, opt, &len);
288 ip = mtod(m, struct ip *);
293 if (!(flags & (IP_FORWARDING|IP_RAWOUTPUT))) {
294 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
296 ip->ip_id = ip_newid();
297 ipstat.ips_localout++;
299 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
303 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
306 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
309 * Multicast is not MPSAFE yet. Caller must hold
310 * BGL when output a multicast IP packet.
312 ASSERT_MP_LOCK_HELD(curthread);
316 dst = (struct sockaddr_in *)&ro->ro_dst;
318 * If there is a cached route,
319 * check that it is to the same destination
320 * and is still up. If not, free it and try again.
321 * The address family should also be checked in case of sharing the
325 (!(ro->ro_rt->rt_flags & RTF_UP) ||
326 dst->sin_family != AF_INET ||
327 dst->sin_addr.s_addr != pkt_dst.s_addr)) {
329 ro->ro_rt = (struct rtentry *)NULL;
331 if (ro->ro_rt == NULL) {
332 bzero(dst, sizeof *dst);
333 dst->sin_family = AF_INET;
334 dst->sin_len = sizeof *dst;
335 dst->sin_addr = pkt_dst;
338 * If routing to interface only,
339 * short circuit routing lookup.
341 if (flags & IP_ROUTETOIF) {
342 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
343 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) {
344 ipstat.ips_noroute++;
350 isbroadcast = in_broadcast(dst->sin_addr, ifp);
351 } else if (IN_MULTICAST(ntohl(pkt_dst.s_addr)) &&
352 imo != NULL && imo->imo_multicast_ifp != NULL) {
354 * Bypass the normal routing lookup for multicast
355 * packets if the interface is specified.
357 ifp = imo->imo_multicast_ifp;
359 isbroadcast = 0; /* fool gcc */
362 * If this is the case, we probably don't want to allocate
363 * a protocol-cloned route since we didn't get one from the
364 * ULP. This lets TCP do its thing, while not burdening
365 * forwarding or ICMP with the overhead of cloning a route.
366 * Of course, we still want to do any cloning requested by
367 * the link layer, as this is probably required in all cases
368 * for correct operation (as it is for ARP).
370 if (ro->ro_rt == NULL)
371 rtalloc_ign(ro, RTF_PRCLONING);
372 if (ro->ro_rt == NULL) {
373 ipstat.ips_noroute++;
374 error = EHOSTUNREACH;
377 ia = ifatoia(ro->ro_rt->rt_ifa);
378 ifp = ro->ro_rt->rt_ifp;
380 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
381 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
382 if (ro->ro_rt->rt_flags & RTF_HOST)
383 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
385 isbroadcast = in_broadcast(dst->sin_addr, ifp);
387 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
388 struct in_multi *inm;
390 m->m_flags |= M_MCAST;
392 * IP destination address is multicast. Make sure "dst"
393 * still points to the address in "ro". (It may have been
394 * changed to point to a gateway address, above.)
396 dst = (struct sockaddr_in *)&ro->ro_dst;
398 * See if the caller provided any multicast options
401 ip->ip_ttl = imo->imo_multicast_ttl;
402 if (imo->imo_multicast_vif != -1) {
405 ip_mcast_src(imo->imo_multicast_vif) :
409 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
412 * Confirm that the outgoing interface supports multicast.
414 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
415 if (!(ifp->if_flags & IFF_MULTICAST)) {
416 ipstat.ips_noroute++;
422 * If source address not specified yet, use address
423 * of outgoing interface.
425 if (ip->ip_src.s_addr == INADDR_ANY) {
426 /* Interface may have no addresses. */
428 ip->ip_src = IA_SIN(ia)->sin_addr;
431 IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
433 (imo == NULL || imo->imo_multicast_loop)) {
435 * If we belong to the destination multicast group
436 * on the outgoing interface, and the caller did not
437 * forbid loopback, loop back a copy.
439 ip_mloopback(ifp, m, dst, hlen);
442 * If we are acting as a multicast router, perform
443 * multicast forwarding as if the packet had just
444 * arrived on the interface to which we are about
445 * to send. The multicast forwarding function
446 * recursively calls this function, using the
447 * IP_FORWARDING flag to prevent infinite recursion.
449 * Multicasts that are looped back by ip_mloopback(),
450 * above, will be forwarded by the ip_input() routine,
453 if (ip_mrouter && !(flags & IP_FORWARDING)) {
455 * If rsvp daemon is not running, do not
456 * set ip_moptions. This ensures that the packet
457 * is multicast and not just sent down one link
458 * as prescribed by rsvpd.
463 ip_mforward(ip, ifp, m, imo) != 0) {
471 * Multicasts with a time-to-live of zero may be looped-
472 * back, above, but must not be transmitted on a network.
473 * Also, multicasts addressed to the loopback interface
474 * are not sent -- the above call to ip_mloopback() will
475 * loop back a copy if this host actually belongs to the
476 * destination group on the loopback interface.
478 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
485 m->m_flags &= ~M_MCAST;
489 * If the source address is not specified yet, use the address
490 * of the outoing interface. In case, keep note we did that, so
491 * if the the firewall changes the next-hop causing the output
492 * interface to change, we can fix that.
494 if (ip->ip_src.s_addr == INADDR_ANY || src_was_INADDR_ANY) {
495 /* Interface may have no addresses. */
497 ip->ip_src = IA_SIN(ia)->sin_addr;
498 src_was_INADDR_ANY = 1;
504 * Disable packet drop hack.
505 * Packetdrop should be done by queueing.
509 * Verify that we have any chance at all of being able to queue
510 * the packet or packet fragments
512 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
513 ifp->if_snd.ifq_maxlen) {
515 ipstat.ips_odropped++;
521 * Look for broadcast address and
522 * verify user is allowed to send
526 if (!(ifp->if_flags & IFF_BROADCAST)) {
527 error = EADDRNOTAVAIL;
530 if (!(flags & IP_ALLOWBROADCAST)) {
534 /* don't allow broadcast messages to be fragmented */
535 if (ip->ip_len > ifp->if_mtu) {
539 m->m_flags |= M_BCAST;
541 m->m_flags &= ~M_BCAST;
546 /* get SP for this packet */
548 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
550 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
553 ipsecstat.out_inval++;
560 switch (sp->policy) {
561 case IPSEC_POLICY_DISCARD:
563 * This packet is just discarded.
565 ipsecstat.out_polvio++;
568 case IPSEC_POLICY_BYPASS:
569 case IPSEC_POLICY_NONE:
570 /* no need to do IPsec. */
573 case IPSEC_POLICY_IPSEC:
574 if (sp->req == NULL) {
575 /* acquire a policy */
576 error = key_spdacquire(sp);
581 case IPSEC_POLICY_ENTRUST:
583 kprintf("ip_output: Invalid policy found. %d\n", sp->policy);
586 struct ipsec_output_state state;
587 bzero(&state, sizeof state);
589 if (flags & IP_ROUTETOIF) {
591 bzero(&iproute, sizeof iproute);
594 state.dst = (struct sockaddr *)dst;
600 * delayed checksums are not currently compatible with IPsec
602 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
604 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
607 ip->ip_len = htons(ip->ip_len);
608 ip->ip_off = htons(ip->ip_off);
610 error = ipsec4_output(&state, sp, flags);
613 if (flags & IP_ROUTETOIF) {
615 * if we have tunnel mode SA, we may need to ignore
618 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
619 flags &= ~IP_ROUTETOIF;
624 dst = (struct sockaddr_in *)state.dst;
626 /* mbuf is already reclaimed in ipsec4_output. */
636 kprintf("ip4_output (ipsec): error code %d\n", error);
639 /* don't show these error codes to the user */
647 /* be sure to update variables that are affected by ipsec4_output() */
648 ip = mtod(m, struct ip *);
650 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
652 hlen = ip->ip_hl << 2;
654 if (ro->ro_rt == NULL) {
655 if (!(flags & IP_ROUTETOIF)) {
656 kprintf("ip_output: "
657 "can't update route after IPsec processing\n");
658 error = EHOSTUNREACH; /*XXX*/
662 ia = ifatoia(ro->ro_rt->rt_ifa);
663 ifp = ro->ro_rt->rt_ifp;
666 /* make it flipped, again. */
667 ip->ip_len = ntohs(ip->ip_len);
668 ip->ip_off = ntohs(ip->ip_off);
673 * Check the security policy (SP) for the packet and, if
674 * required, do IPsec-related processing. There are two
675 * cases here; the first time a packet is sent through
676 * it will be untagged and handled by ipsec4_checkpolicy.
677 * If the packet is resubmitted to ip_output (e.g. after
678 * AH, ESP, etc. processing), there will be a tag to bypass
679 * the lookup and related policy checking.
681 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
684 tdbi = (struct tdb_ident *)m_tag_data(mtag);
685 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
687 error = -EINVAL; /* force silent drop */
688 m_tag_delete(m, mtag);
690 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
694 * There are four return cases:
695 * sp != NULL apply IPsec policy
696 * sp == NULL, error == 0 no IPsec handling needed
697 * sp == NULL, error == -EINVAL discard packet w/o error
698 * sp == NULL, error != 0 discard packet, report error
701 /* Loop detection, check if ipsec processing already done */
702 KASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
703 for (mtag = m_tag_first(m); mtag != NULL;
704 mtag = m_tag_next(m, mtag)) {
705 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
707 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
708 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
711 * Check if policy has an SA associated with it.
712 * This can happen when an SP has yet to acquire
713 * an SA; e.g. on first reference. If it occurs,
714 * then we let ipsec4_process_packet do its thing.
716 if (sp->req->sav == NULL)
718 tdbi = (struct tdb_ident *)m_tag_data(mtag);
719 if (tdbi->spi == sp->req->sav->spi &&
720 tdbi->proto == sp->req->sav->sah->saidx.proto &&
721 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
722 sizeof(union sockaddr_union)) == 0) {
724 * No IPsec processing is needed, free
727 * NB: null pointer to avoid free at
730 KEY_FREESP(&sp), sp = NULL;
737 * Do delayed checksums now because we send before
738 * this is done in the normal processing path.
740 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
742 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
745 ip->ip_len = htons(ip->ip_len);
746 ip->ip_off = htons(ip->ip_off);
748 /* NB: callee frees mbuf */
749 error = ipsec4_process_packet(m, sp->req, flags, 0);
751 * Preserve KAME behaviour: ENOENT can be returned
752 * when an SA acquire is in progress. Don't propagate
753 * this to user-level; it confuses applications.
755 * XXX this will go away when the SADB is redone.
766 * Hack: -EINVAL is used to signal that a packet
767 * should be silently discarded. This is typically
768 * because we asked key management for an SA and
769 * it was delayed (e.g. kicked up to IKE).
771 if (error == -EINVAL)
775 /* No IPsec processing for this packet. */
779 * If deferred crypto processing is needed, check that
780 * the interface supports it.
782 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
783 if (mtag != NULL && !(ifp->if_capenable & IFCAP_IPSEC)) {
784 /* notify IPsec to do its own crypto */
785 ipsp_skipcrypto_unmark((struct tdb_ident *)m_tag_data(mtag));
786 error = EHOSTUNREACH;
792 #endif /* FAST_IPSEC */
794 /* We are already being fwd'd from a firewall. */
795 if (next_hop != NULL)
799 if (!pfil_has_hooks(&inet_pfil_hook)) {
800 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
802 * Strip dummynet tags from stranded packets
804 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
805 KKASSERT(mtag != NULL);
806 m_tag_delete(m, mtag);
807 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
814 * - Xlate: translate packet's addr/port (NAT).
815 * - Firewall: deny/allow/etc.
816 * - Wrap: fake packet's addr/port <unimpl.>
817 * - Encapsulate: put it in another IP and send out. <unimp.>
821 * Run through list of hooks for output packets.
823 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT);
824 if (error != 0 || m == NULL)
826 ip = mtod(m, struct ip *);
828 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
830 * Check dst to make sure it is directly reachable on the
831 * interface we previously thought it was.
832 * If it isn't (which may be likely in some situations) we have
833 * to re-route it (ie, find a route for the next-hop and the
834 * associated interface) and set them here. This is nested
835 * forwarding which in most cases is undesirable, except where
836 * such control is nigh impossible. So we do it here.
839 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
840 KKASSERT(mtag != NULL);
841 next_hop = m_tag_data(mtag);
844 * Try local forwarding first
846 if (ip_localforward(m, next_hop, hlen))
850 * Relocate the route based on next_hop.
851 * If the current route is inp's cache, keep it untouched.
853 if (ro == &iproute && ro->ro_rt != NULL) {
858 bzero(ro, sizeof *ro);
861 * Forwarding to broadcast address is not allowed.
862 * XXX Should we follow IP_ROUTETOIF?
864 flags &= ~(IP_ALLOWBROADCAST | IP_ROUTETOIF);
866 /* We are doing forwarding now */
867 flags |= IP_FORWARDING;
872 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
873 struct dn_pkt *dn_pkt;
875 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
876 KKASSERT(mtag != NULL);
877 dn_pkt = m_tag_data(mtag);
880 * Under certain cases it is not possible to recalculate
881 * 'ro' and 'dst', let alone 'flags', so just save them in
882 * dummynet tag and avoid the possible wrong reculcalation
883 * when we come back to ip_output() again.
885 * All other parameters have been already used and so they
886 * are not needed anymore.
887 * XXX if the ifp is deleted while a pkt is in dummynet,
888 * we are in trouble! (TODO use ifnet_detach_event)
890 * We need to copy *ro because for ICMP pkts (and maybe
891 * others) the caller passed a pointer into the stack;
892 * dst might also be a pointer into *ro so it needs to
897 ro->ro_rt->rt_refcnt++;
898 if (dst == (struct sockaddr_in *)&ro->ro_dst) {
899 /* 'dst' points into 'ro' */
900 dst = (struct sockaddr_in *)&(dn_pkt->ro.ro_dst);
902 dn_pkt->dn_dst = dst;
903 dn_pkt->flags = flags;
909 /* 127/8 must not appear on wire - RFC1122. */
910 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
911 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
912 if (!(ifp->if_flags & IFF_LOOPBACK)) {
913 ipstat.ips_badaddr++;
914 error = EADDRNOTAVAIL;
919 m->m_pkthdr.csum_flags |= CSUM_IP;
920 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
921 if (sw_csum & CSUM_DELAY_DATA) {
923 sw_csum &= ~CSUM_DELAY_DATA;
925 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
928 * If small enough for interface, or the interface will take
929 * care of the fragmentation for us, can just send directly.
931 if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
932 !(ip->ip_off & IP_DF))) {
933 ip->ip_len = htons(ip->ip_len);
934 ip->ip_off = htons(ip->ip_off);
936 if (sw_csum & CSUM_DELAY_IP) {
937 if (ip->ip_vhl == IP_VHL_BORING)
938 ip->ip_sum = in_cksum_hdr(ip);
940 ip->ip_sum = in_cksum(m, hlen);
943 /* Record statistics for this interface address. */
944 if (!(flags & IP_FORWARDING) && ia) {
945 ia->ia_ifa.if_opackets++;
946 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
950 /* clean ipsec history once it goes out of the node */
954 #ifdef MBUF_STRESS_TEST
955 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) {
956 struct mbuf *m1, *m2;
959 tmp = length = m->m_pkthdr.len;
961 while ((length -= mbuf_frag_size) >= 1) {
962 m1 = m_split(m, length, MB_DONTWAIT);
966 while (m2->m_next != NULL)
970 m->m_pkthdr.len = tmp;
975 if (!mpls_output_process(m, ro->ro_rt))
978 error = ifp->if_output(ifp, m, (struct sockaddr *)dst,
983 if (ip->ip_off & IP_DF) {
986 * This case can happen if the user changed the MTU
987 * of an interface after enabling IP on it. Because
988 * most netifs don't keep track of routes pointing to
989 * them, there is no way for one to update all its
990 * routes when the MTU is changed.
992 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
993 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
994 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
995 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
997 ipstat.ips_cantfrag++;
1002 * Too large for interface; fragment if possible. If successful,
1003 * on return, m will point to a list of packets to be sent.
1005 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
1010 m->m_nextpkt = NULL;
1012 /* clean ipsec history once it goes out of the node */
1016 /* Record statistics for this interface address. */
1018 ia->ia_ifa.if_opackets++;
1019 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1022 if (!mpls_output_process(m, ro->ro_rt))
1025 error = ifp->if_output(ifp, m, (struct sockaddr *)dst,
1033 ipstat.ips_fragmented++;
1036 if (ro == &iproute && ro->ro_rt != NULL) {
1042 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1043 kprintf("DP ip_output call free SP:%p\n", sp));
1058 * Create a chain of fragments which fit the given mtu. m_frag points to the
1059 * mbuf to be fragmented; on return it points to the chain with the fragments.
1060 * Return 0 if no error. If error, m_frag may contain a partially built
1061 * chain of fragments that should be freed by the caller.
1063 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
1064 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
1067 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
1068 u_long if_hwassist_flags, int sw_csum)
1071 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1072 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
1074 struct mbuf *m0 = *m_frag; /* the original packet */
1076 struct mbuf **mnext;
1079 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
1080 ipstat.ips_cantfrag++;
1085 * Must be able to put at least 8 bytes per fragment.
1091 * If the interface will not calculate checksums on
1092 * fragmented packets, then do it here.
1094 if ((m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) &&
1095 !(if_hwassist_flags & CSUM_IP_FRAGS)) {
1096 in_delayed_cksum(m0);
1097 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1100 if (len > PAGE_SIZE) {
1102 * Fragment large datagrams such that each segment
1103 * contains a multiple of PAGE_SIZE amount of data,
1104 * plus headers. This enables a receiver to perform
1105 * page-flipping zero-copy optimizations.
1107 * XXX When does this help given that sender and receiver
1108 * could have different page sizes, and also mtu could
1109 * be less than the receiver's page size ?
1114 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
1118 * firstlen (off - hlen) must be aligned on an
1122 goto smart_frag_failure;
1123 off = ((off - hlen) & ~7) + hlen;
1124 newlen = (~PAGE_MASK) & mtu;
1125 if ((newlen + sizeof(struct ip)) > mtu) {
1126 /* we failed, go back the default */
1137 firstlen = off - hlen;
1138 mnext = &m0->m_nextpkt; /* pointer to next packet */
1141 * Loop through length of segment after first fragment,
1142 * make new header and copy data of each part and link onto chain.
1143 * Here, m0 is the original packet, m is the fragment being created.
1144 * The fragments are linked off the m_nextpkt of the original
1145 * packet, which after processing serves as the first fragment.
1147 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
1148 struct ip *mhip; /* ip header on the fragment */
1150 int mhlen = sizeof(struct ip);
1152 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1155 ipstat.ips_odropped++;
1158 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
1160 * In the first mbuf, leave room for the link header, then
1161 * copy the original IP header including options. The payload
1162 * goes into an additional mbuf chain returned by m_copy().
1164 m->m_data += max_linkhdr;
1165 mhip = mtod(m, struct ip *);
1167 if (hlen > sizeof(struct ip)) {
1168 mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip);
1169 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
1172 /* XXX do we need to add ip->ip_off below ? */
1173 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
1174 if (off + len >= ip->ip_len) { /* last fragment */
1175 len = ip->ip_len - off;
1176 m->m_flags |= M_LASTFRAG;
1178 mhip->ip_off |= IP_MF;
1179 mhip->ip_len = htons((u_short)(len + mhlen));
1180 m->m_next = m_copy(m0, off, len);
1181 if (m->m_next == NULL) { /* copy failed */
1183 error = ENOBUFS; /* ??? */
1184 ipstat.ips_odropped++;
1187 m->m_pkthdr.len = mhlen + len;
1188 m->m_pkthdr.rcvif = (struct ifnet *)NULL;
1189 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1190 mhip->ip_off = htons(mhip->ip_off);
1192 if (sw_csum & CSUM_DELAY_IP)
1193 mhip->ip_sum = in_cksum(m, mhlen);
1195 mnext = &m->m_nextpkt;
1197 ipstat.ips_ofragments += nfrags;
1199 /* set first marker for fragment chain */
1200 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1201 m0->m_pkthdr.csum_data = nfrags;
1204 * Update first fragment by trimming what's been copied out
1205 * and updating header.
1207 m_adj(m0, hlen + firstlen - ip->ip_len);
1208 m0->m_pkthdr.len = hlen + firstlen;
1209 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1210 ip->ip_off |= IP_MF;
1211 ip->ip_off = htons(ip->ip_off);
1213 if (sw_csum & CSUM_DELAY_IP)
1214 ip->ip_sum = in_cksum(m0, hlen);
1222 in_delayed_cksum(struct mbuf *m)
1225 u_short csum, offset;
1227 ip = mtod(m, struct ip *);
1228 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1229 csum = in_cksum_skip(m, ip->ip_len, offset);
1230 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1232 offset += m->m_pkthdr.csum_data; /* checksum offset */
1234 if (offset + sizeof(u_short) > m->m_len) {
1235 kprintf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1236 m->m_len, offset, ip->ip_p);
1239 * this shouldn't happen, but if it does, the
1240 * correct behavior may be to insert the checksum
1241 * in the existing chain instead of rearranging it.
1243 m = m_pullup(m, offset + sizeof(u_short));
1245 *(u_short *)(m->m_data + offset) = csum;
1249 * Insert IP options into preformed packet.
1250 * Adjust IP destination as required for IP source routing,
1251 * as indicated by a non-zero in_addr at the start of the options.
1253 * XXX This routine assumes that the packet has no options in place.
1255 static struct mbuf *
1256 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
1258 struct ipoption *p = mtod(opt, struct ipoption *);
1260 struct ip *ip = mtod(m, struct ip *);
1263 optlen = opt->m_len - sizeof p->ipopt_dst;
1264 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
1266 return (m); /* XXX should fail */
1268 if (p->ipopt_dst.s_addr)
1269 ip->ip_dst = p->ipopt_dst;
1270 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1271 MGETHDR(n, MB_DONTWAIT, MT_HEADER);
1276 n->m_pkthdr.rcvif = (struct ifnet *)NULL;
1277 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1278 m->m_len -= sizeof(struct ip);
1279 m->m_data += sizeof(struct ip);
1282 m->m_len = optlen + sizeof(struct ip);
1283 m->m_data += max_linkhdr;
1284 memcpy(mtod(m, void *), ip, sizeof(struct ip));
1286 m->m_data -= optlen;
1288 m->m_pkthdr.len += optlen;
1289 ovbcopy(ip, mtod(m, caddr_t), sizeof(struct ip));
1291 ip = mtod(m, struct ip *);
1292 bcopy(p->ipopt_list, ip + 1, optlen);
1293 *phlen = sizeof(struct ip) + optlen;
1294 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1295 ip->ip_len += optlen;
1300 * Copy options from ip to jp,
1301 * omitting those not copied during fragmentation.
1304 ip_optcopy(struct ip *ip, struct ip *jp)
1307 int opt, optlen, cnt;
1309 cp = (u_char *)(ip + 1);
1310 dp = (u_char *)(jp + 1);
1311 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1312 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1314 if (opt == IPOPT_EOL)
1316 if (opt == IPOPT_NOP) {
1317 /* Preserve for IP mcast tunnel's LSRR alignment. */
1323 KASSERT(cnt >= IPOPT_OLEN + sizeof *cp,
1324 ("ip_optcopy: malformed ipv4 option"));
1325 optlen = cp[IPOPT_OLEN];
1326 KASSERT(optlen >= IPOPT_OLEN + sizeof *cp && optlen <= cnt,
1327 ("ip_optcopy: malformed ipv4 option"));
1329 /* bogus lengths should have been caught by ip_dooptions */
1332 if (IPOPT_COPIED(opt)) {
1333 bcopy(cp, dp, optlen);
1337 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1343 * IP socket option processing.
1346 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1348 struct inpcb *inp = so->so_pcb;
1352 if (sopt->sopt_level != IPPROTO_IP) {
1356 switch (sopt->sopt_dir) {
1358 switch (sopt->sopt_name) {
1365 if (sopt->sopt_valsize > MLEN) {
1369 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1374 m->m_len = sopt->sopt_valsize;
1375 error = soopt_to_kbuf(sopt, mtod(m, void *), m->m_len,
1377 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1385 case IP_RECVRETOPTS:
1386 case IP_RECVDSTADDR:
1390 error = soopt_to_kbuf(sopt, &optval, sizeof optval,
1394 switch (sopt->sopt_name) {
1396 inp->inp_ip_tos = optval;
1400 inp->inp_ip_ttl = optval;
1403 if (optval > 0 && optval <= MAXTTL)
1404 inp->inp_ip_minttl = optval;
1408 #define OPTSET(bit) \
1410 inp->inp_flags |= bit; \
1412 inp->inp_flags &= ~bit;
1415 OPTSET(INP_RECVOPTS);
1418 case IP_RECVRETOPTS:
1419 OPTSET(INP_RECVRETOPTS);
1422 case IP_RECVDSTADDR:
1423 OPTSET(INP_RECVDSTADDR);
1431 OPTSET(INP_RECVTTL);
1441 case IP_MULTICAST_IF:
1442 case IP_MULTICAST_VIF:
1443 case IP_MULTICAST_TTL:
1444 case IP_MULTICAST_LOOP:
1445 case IP_ADD_MEMBERSHIP:
1446 case IP_DROP_MEMBERSHIP:
1447 error = ip_setmoptions(sopt, &inp->inp_moptions);
1451 error = soopt_to_kbuf(sopt, &optval, sizeof optval,
1457 case IP_PORTRANGE_DEFAULT:
1458 inp->inp_flags &= ~(INP_LOWPORT);
1459 inp->inp_flags &= ~(INP_HIGHPORT);
1462 case IP_PORTRANGE_HIGH:
1463 inp->inp_flags &= ~(INP_LOWPORT);
1464 inp->inp_flags |= INP_HIGHPORT;
1467 case IP_PORTRANGE_LOW:
1468 inp->inp_flags &= ~(INP_HIGHPORT);
1469 inp->inp_flags |= INP_LOWPORT;
1478 #if defined(IPSEC) || defined(FAST_IPSEC)
1479 case IP_IPSEC_POLICY:
1487 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1489 soopt_to_mbuf(sopt, m);
1490 priv = (sopt->sopt_td != NULL &&
1491 suser(sopt->sopt_td) != 0) ? 0 : 1;
1492 req = mtod(m, caddr_t);
1494 optname = sopt->sopt_name;
1495 error = ipsec4_set_policy(inp, optname, req, len, priv);
1502 error = ENOPROTOOPT;
1508 switch (sopt->sopt_name) {
1511 if (inp->inp_options)
1512 soopt_from_kbuf(sopt, mtod(inp->inp_options,
1514 inp->inp_options->m_len);
1516 sopt->sopt_valsize = 0;
1523 case IP_RECVRETOPTS:
1524 case IP_RECVDSTADDR:
1529 switch (sopt->sopt_name) {
1532 optval = inp->inp_ip_tos;
1536 optval = inp->inp_ip_ttl;
1539 optval = inp->inp_ip_minttl;
1542 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1545 optval = OPTBIT(INP_RECVOPTS);
1548 case IP_RECVRETOPTS:
1549 optval = OPTBIT(INP_RECVRETOPTS);
1552 case IP_RECVDSTADDR:
1553 optval = OPTBIT(INP_RECVDSTADDR);
1557 optval = OPTBIT(INP_RECVTTL);
1561 optval = OPTBIT(INP_RECVIF);
1565 if (inp->inp_flags & INP_HIGHPORT)
1566 optval = IP_PORTRANGE_HIGH;
1567 else if (inp->inp_flags & INP_LOWPORT)
1568 optval = IP_PORTRANGE_LOW;
1574 optval = OPTBIT(INP_FAITH);
1577 soopt_from_kbuf(sopt, &optval, sizeof optval);
1580 case IP_MULTICAST_IF:
1581 case IP_MULTICAST_VIF:
1582 case IP_MULTICAST_TTL:
1583 case IP_MULTICAST_LOOP:
1584 case IP_ADD_MEMBERSHIP:
1585 case IP_DROP_MEMBERSHIP:
1586 error = ip_getmoptions(sopt, inp->inp_moptions);
1589 #if defined(IPSEC) || defined(FAST_IPSEC)
1590 case IP_IPSEC_POLICY:
1592 struct mbuf *m = NULL;
1597 req = mtod(m, caddr_t);
1600 error = ipsec4_get_policy(so->so_pcb, req, len, &m);
1602 error = soopt_from_mbuf(sopt, m); /* XXX */
1610 error = ENOPROTOOPT;
1619 * Set up IP options in pcb for insertion in output packets.
1620 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1621 * with destination address if source routed.
1624 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
1630 /* turn off any old options */
1634 if (m == NULL || m->m_len == 0) {
1636 * Only turning off any previous options.
1643 if (m->m_len % sizeof(int32_t))
1646 * IP first-hop destination address will be stored before
1647 * actual options; move other options back
1648 * and clear it when none present.
1650 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1653 m->m_len += sizeof(struct in_addr);
1654 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1655 ovbcopy(mtod(m, caddr_t), cp, cnt);
1656 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1658 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1659 opt = cp[IPOPT_OPTVAL];
1660 if (opt == IPOPT_EOL)
1662 if (opt == IPOPT_NOP)
1665 if (cnt < IPOPT_OLEN + sizeof *cp)
1667 optlen = cp[IPOPT_OLEN];
1668 if (optlen < IPOPT_OLEN + sizeof *cp || optlen > cnt)
1679 * user process specifies route as:
1681 * D must be our final destination (but we can't
1682 * check that since we may not have connected yet).
1683 * A is first hop destination, which doesn't appear in
1684 * actual IP option, but is stored before the options.
1686 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1688 m->m_len -= sizeof(struct in_addr);
1689 cnt -= sizeof(struct in_addr);
1690 optlen -= sizeof(struct in_addr);
1691 cp[IPOPT_OLEN] = optlen;
1693 * Move first hop before start of options.
1695 bcopy(&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1696 sizeof(struct in_addr));
1698 * Then copy rest of options back
1699 * to close up the deleted entry.
1701 ovbcopy(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr),
1702 &cp[IPOPT_OFFSET+1],
1703 cnt - (IPOPT_MINOFF - 1));
1707 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1719 * The whole multicast option thing needs to be re-thought.
1720 * Several of these options are equally applicable to non-multicast
1721 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1722 * standard option (IP_TTL).
1726 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1728 static struct ifnet *
1729 ip_multicast_if(struct in_addr *a, int *ifindexp)
1736 if (ntohl(a->s_addr) >> 24 == 0) {
1737 ifindex = ntohl(a->s_addr) & 0xffffff;
1738 if (ifindex < 0 || if_index < ifindex)
1740 ifp = ifindex2ifnet[ifindex];
1742 *ifindexp = ifindex;
1744 ifp = INADDR_TO_IFP(a);
1750 * Set the IP multicast options in response to user setsockopt().
1753 ip_setmoptions(struct sockopt *sopt, struct ip_moptions **imop)
1757 struct in_addr addr;
1758 struct ip_mreq mreq;
1760 struct ip_moptions *imo = *imop;
1765 * No multicast option buffer attached to the pcb;
1766 * allocate one and initialize to default values.
1768 imo = kmalloc(sizeof *imo, M_IPMOPTS, M_WAITOK);
1771 imo->imo_multicast_ifp = NULL;
1772 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1773 imo->imo_multicast_vif = -1;
1774 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1775 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1776 imo->imo_num_memberships = 0;
1778 switch (sopt->sopt_name) {
1779 /* store an index number for the vif you wanna use in the send */
1780 case IP_MULTICAST_VIF:
1781 if (legal_vif_num == 0) {
1785 error = soopt_to_kbuf(sopt, &i, sizeof i, sizeof i);
1788 if (!legal_vif_num(i) && (i != -1)) {
1792 imo->imo_multicast_vif = i;
1795 case IP_MULTICAST_IF:
1797 * Select the interface for outgoing multicast packets.
1799 error = soopt_to_kbuf(sopt, &addr, sizeof addr, sizeof addr);
1804 * INADDR_ANY is used to remove a previous selection.
1805 * When no interface is selected, a default one is
1806 * chosen every time a multicast packet is sent.
1808 if (addr.s_addr == INADDR_ANY) {
1809 imo->imo_multicast_ifp = NULL;
1813 * The selected interface is identified by its local
1814 * IP address. Find the interface and confirm that
1815 * it supports multicasting.
1818 ifp = ip_multicast_if(&addr, &ifindex);
1819 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1821 error = EADDRNOTAVAIL;
1824 imo->imo_multicast_ifp = ifp;
1826 imo->imo_multicast_addr = addr;
1828 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1832 case IP_MULTICAST_TTL:
1834 * Set the IP time-to-live for outgoing multicast packets.
1835 * The original multicast API required a char argument,
1836 * which is inconsistent with the rest of the socket API.
1837 * We allow either a char or an int.
1839 if (sopt->sopt_valsize == 1) {
1841 error = soopt_to_kbuf(sopt, &ttl, 1, 1);
1844 imo->imo_multicast_ttl = ttl;
1847 error = soopt_to_kbuf(sopt, &ttl, sizeof ttl, sizeof ttl);
1853 imo->imo_multicast_ttl = ttl;
1857 case IP_MULTICAST_LOOP:
1859 * Set the loopback flag for outgoing multicast packets.
1860 * Must be zero or one. The original multicast API required a
1861 * char argument, which is inconsistent with the rest
1862 * of the socket API. We allow either a char or an int.
1864 if (sopt->sopt_valsize == 1) {
1867 error = soopt_to_kbuf(sopt, &loop, 1, 1);
1870 imo->imo_multicast_loop = !!loop;
1874 error = soopt_to_kbuf(sopt, &loop, sizeof loop,
1878 imo->imo_multicast_loop = !!loop;
1882 case IP_ADD_MEMBERSHIP:
1884 * Add a multicast group membership.
1885 * Group must be a valid IP multicast address.
1887 error = soopt_to_kbuf(sopt, &mreq, sizeof mreq, sizeof mreq);
1891 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1897 * If no interface address was provided, use the interface of
1898 * the route to the given multicast address.
1900 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1901 struct sockaddr_in dst;
1904 bzero(&dst, sizeof(struct sockaddr_in));
1905 dst.sin_len = sizeof(struct sockaddr_in);
1906 dst.sin_family = AF_INET;
1907 dst.sin_addr = mreq.imr_multiaddr;
1908 rt = rtlookup((struct sockaddr *)&dst);
1910 error = EADDRNOTAVAIL;
1917 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1921 * See if we found an interface, and confirm that it
1922 * supports multicast.
1924 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1925 error = EADDRNOTAVAIL;
1930 * See if the membership already exists or if all the
1931 * membership slots are full.
1933 for (i = 0; i < imo->imo_num_memberships; ++i) {
1934 if (imo->imo_membership[i]->inm_ifp == ifp &&
1935 imo->imo_membership[i]->inm_addr.s_addr
1936 == mreq.imr_multiaddr.s_addr)
1939 if (i < imo->imo_num_memberships) {
1944 if (i == IP_MAX_MEMBERSHIPS) {
1945 error = ETOOMANYREFS;
1950 * Everything looks good; add a new record to the multicast
1951 * address list for the given interface.
1953 if ((imo->imo_membership[i] =
1954 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
1959 ++imo->imo_num_memberships;
1963 case IP_DROP_MEMBERSHIP:
1965 * Drop a multicast group membership.
1966 * Group must be a valid IP multicast address.
1968 error = soopt_to_kbuf(sopt, &mreq, sizeof mreq, sizeof mreq);
1972 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1979 * If an interface address was specified, get a pointer
1980 * to its ifnet structure.
1982 if (mreq.imr_interface.s_addr == INADDR_ANY)
1985 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1987 error = EADDRNOTAVAIL;
1993 * Find the membership in the membership array.
1995 for (i = 0; i < imo->imo_num_memberships; ++i) {
1997 imo->imo_membership[i]->inm_ifp == ifp) &&
1998 imo->imo_membership[i]->inm_addr.s_addr ==
1999 mreq.imr_multiaddr.s_addr)
2002 if (i == imo->imo_num_memberships) {
2003 error = EADDRNOTAVAIL;
2008 * Give up the multicast address record to which the
2009 * membership points.
2011 in_delmulti(imo->imo_membership[i]);
2013 * Remove the gap in the membership array.
2015 for (++i; i < imo->imo_num_memberships; ++i)
2016 imo->imo_membership[i-1] = imo->imo_membership[i];
2017 --imo->imo_num_memberships;
2027 * If all options have default values, no need to keep the mbuf.
2029 if (imo->imo_multicast_ifp == NULL &&
2030 imo->imo_multicast_vif == -1 &&
2031 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2032 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2033 imo->imo_num_memberships == 0) {
2034 kfree(*imop, M_IPMOPTS);
2042 * Return the IP multicast options in response to user getsockopt().
2045 ip_getmoptions(struct sockopt *sopt, struct ip_moptions *imo)
2047 struct in_addr addr;
2048 struct in_ifaddr *ia;
2053 switch (sopt->sopt_name) {
2054 case IP_MULTICAST_VIF:
2056 optval = imo->imo_multicast_vif;
2059 soopt_from_kbuf(sopt, &optval, sizeof optval);
2062 case IP_MULTICAST_IF:
2063 if (imo == NULL || imo->imo_multicast_ifp == NULL)
2064 addr.s_addr = INADDR_ANY;
2065 else if (imo->imo_multicast_addr.s_addr) {
2066 /* return the value user has set */
2067 addr = imo->imo_multicast_addr;
2069 ia = IFP_TO_IA(imo->imo_multicast_ifp);
2070 addr.s_addr = (ia == NULL) ? INADDR_ANY
2071 : IA_SIN(ia)->sin_addr.s_addr;
2073 soopt_from_kbuf(sopt, &addr, sizeof addr);
2076 case IP_MULTICAST_TTL:
2078 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
2080 optval = coptval = imo->imo_multicast_ttl;
2081 if (sopt->sopt_valsize == 1)
2082 soopt_from_kbuf(sopt, &coptval, 1);
2084 soopt_from_kbuf(sopt, &optval, sizeof optval);
2087 case IP_MULTICAST_LOOP:
2089 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
2091 optval = coptval = imo->imo_multicast_loop;
2092 if (sopt->sopt_valsize == 1)
2093 soopt_from_kbuf(sopt, &coptval, 1);
2095 soopt_from_kbuf(sopt, &optval, sizeof optval);
2099 error = ENOPROTOOPT;
2106 * Discard the IP multicast options.
2109 ip_freemoptions(struct ip_moptions *imo)
2114 for (i = 0; i < imo->imo_num_memberships; ++i)
2115 in_delmulti(imo->imo_membership[i]);
2116 kfree(imo, M_IPMOPTS);
2121 * Routine called from ip_output() to loop back a copy of an IP multicast
2122 * packet to the input queue of a specified interface. Note that this
2123 * calls the output routine of the loopback "driver", but with an interface
2124 * pointer that might NOT be a loopback interface -- evil, but easier than
2125 * replicating that code here.
2128 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
2134 copym = m_copypacket(m, MB_DONTWAIT);
2135 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2136 copym = m_pullup(copym, hlen);
2137 if (copym != NULL) {
2139 * if the checksum hasn't been computed, mark it as valid
2141 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2142 in_delayed_cksum(copym);
2143 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2144 copym->m_pkthdr.csum_flags |=
2145 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2146 copym->m_pkthdr.csum_data = 0xffff;
2149 * We don't bother to fragment if the IP length is greater
2150 * than the interface's MTU. Can this possibly matter?
2152 ip = mtod(copym, struct ip *);
2153 ip->ip_len = htons(ip->ip_len);
2154 ip->ip_off = htons(ip->ip_off);
2156 if (ip->ip_vhl == IP_VHL_BORING) {
2157 ip->ip_sum = in_cksum_hdr(ip);
2159 ip->ip_sum = in_cksum(copym, hlen);
2163 * It's not clear whether there are any lingering
2164 * reentrancy problems in other areas which might
2165 * be exposed by using ip_input directly (in
2166 * particular, everything which modifies the packet
2167 * in-place). Yet another option is using the
2168 * protosw directly to deliver the looped back
2169 * packet. For the moment, we'll err on the side
2170 * of safety by using if_simloop().
2173 if (dst->sin_family != AF_INET) {
2174 kprintf("ip_mloopback: bad address family %d\n",
2176 dst->sin_family = AF_INET;
2179 if_simloop(ifp, copym, dst->sin_family, 0);