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 1.44 2008/06/09 11:24:24 sephe Exp $
38 #include "opt_ipdivert.h"
39 #include "opt_ipfilter.h"
40 #include "opt_ipsec.h"
41 #include "opt_mbuf_stress_test.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
46 #include <sys/malloc.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
52 #include <sys/sysctl.h>
53 #include <sys/thread2.h>
54 #include <sys/in_cksum.h>
57 #include <net/netisr.h>
59 #include <net/route.h>
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/ip.h>
64 #include <netinet/in_pcb.h>
65 #include <netinet/in_var.h>
66 #include <netinet/ip_var.h>
68 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
71 #include <netinet6/ipsec.h>
72 #include <netproto/key/key.h>
74 #include <netproto/key/key_debug.h>
76 #define KEYDEBUG(lev,arg)
81 #include <netproto/ipsec/ipsec.h>
82 #include <netproto/ipsec/xform.h>
83 #include <netproto/ipsec/key.h>
86 #include <net/ipfw/ip_fw.h>
87 #include <net/dummynet/ip_dummynet.h>
89 #define print_ip(x, a, y) kprintf("%s %d.%d.%d.%d%s",\
90 x, (ntohl(a.s_addr)>>24)&0xFF,\
91 (ntohl(a.s_addr)>>16)&0xFF,\
92 (ntohl(a.s_addr)>>8)&0xFF,\
93 (ntohl(a.s_addr))&0xFF, y);
97 #ifdef MBUF_STRESS_TEST
98 int mbuf_frag_size = 0;
99 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
100 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
103 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
104 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
105 static void ip_mloopback
106 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
107 static int ip_getmoptions
108 (struct sockopt *, struct ip_moptions *);
109 static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
110 static int ip_setmoptions
111 (struct sockopt *, struct ip_moptions **);
113 int ip_optcopy(struct ip *, struct ip *);
116 extern struct protosw inetsw[];
119 * IP output. The packet in mbuf chain m contains a skeletal IP
120 * header (with len, off, ttl, proto, tos, src, dst).
121 * The mbuf chain containing the packet will be freed.
122 * The mbuf opt, if present, will not be freed.
125 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
126 int flags, struct ip_moptions *imo, struct inpcb *inp)
129 struct ifnet *ifp = NULL; /* keep compiler happy */
131 int hlen = sizeof(struct ip);
132 int len, off, error = 0;
133 struct sockaddr_in *dst = NULL; /* keep compiler happy */
134 struct in_ifaddr *ia = NULL;
135 int isbroadcast, sw_csum;
136 struct in_addr pkt_dst;
137 struct route iproute;
138 struct m_tag *dn_mtag = NULL;
140 struct secpolicy *sp = NULL;
141 struct socket *so = inp ? inp->inp_socket : NULL;
145 struct secpolicy *sp = NULL;
146 struct tdb_ident *tdbi;
147 #endif /* FAST_IPSEC */
148 struct ip_fw_args args;
149 int src_was_INADDR_ANY = 0; /* as the name says... */
153 args.next_hop = NULL;
155 /* Grab info from MT_TAG mbufs prepended to the chain. */
156 while (m0 != NULL && m0->m_type == MT_TAG) {
157 switch(m0->_m_tag_id) {
158 case PACKET_TAG_IPFORWARD:
159 args.next_hop = (struct sockaddr_in *)m0->m_data;
162 kprintf("ip_output: unrecognised MT_TAG tag %d\n",
171 /* Extract info from dummynet tag */
172 dn_mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
173 if (dn_mtag != NULL) {
174 struct dn_pkt *dn_pkt = m_tag_data(dn_mtag);
177 * The packet was already tagged, so part of the
178 * processing was already done, and we need to go down.
179 * Get parameters from the tag.
181 args.rule = dn_pkt->dn_priv;
185 dst = dn_pkt->dn_dst;
187 flags = dn_pkt->flags;
190 * Don't delete the dummynet tag here, just unlink it,
191 * since some local variables (like 'ro' and 'dst') are
192 * still referencing certain parts of it.
193 * The dummynet tag will be freed at the end of the
196 m_tag_unlink(m, dn_mtag);
201 bzero(ro, sizeof *ro);
204 if (args.rule != NULL) { /* dummynet already saw us */
205 ip = mtod(m, struct ip *);
206 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
208 ia = ifatoia(ro->ro_rt->rt_ifa);
214 m = ip_insertoptions(m, opt, &len);
218 ip = mtod(m, struct ip *);
219 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
224 if (!(flags & (IP_FORWARDING|IP_RAWOUTPUT))) {
225 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
227 ip->ip_id = ip_newid();
228 ipstat.ips_localout++;
230 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
233 dst = (struct sockaddr_in *)&ro->ro_dst;
235 * If there is a cached route,
236 * check that it is to the same destination
237 * and is still up. If not, free it and try again.
238 * The address family should also be checked in case of sharing the
242 (!(ro->ro_rt->rt_flags & RTF_UP) ||
243 dst->sin_family != AF_INET ||
244 dst->sin_addr.s_addr != pkt_dst.s_addr)) {
246 ro->ro_rt = (struct rtentry *)NULL;
248 if (ro->ro_rt == NULL) {
249 bzero(dst, sizeof *dst);
250 dst->sin_family = AF_INET;
251 dst->sin_len = sizeof *dst;
252 dst->sin_addr = pkt_dst;
255 * If routing to interface only,
256 * short circuit routing lookup.
258 if (flags & IP_ROUTETOIF) {
259 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
260 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) {
261 ipstat.ips_noroute++;
267 isbroadcast = in_broadcast(dst->sin_addr, ifp);
268 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
269 imo != NULL && imo->imo_multicast_ifp != NULL) {
271 * Bypass the normal routing lookup for multicast
272 * packets if the interface is specified.
274 ifp = imo->imo_multicast_ifp;
276 isbroadcast = 0; /* fool gcc */
279 * If this is the case, we probably don't want to allocate
280 * a protocol-cloned route since we didn't get one from the
281 * ULP. This lets TCP do its thing, while not burdening
282 * forwarding or ICMP with the overhead of cloning a route.
283 * Of course, we still want to do any cloning requested by
284 * the link layer, as this is probably required in all cases
285 * for correct operation (as it is for ARP).
287 if (ro->ro_rt == NULL)
288 rtalloc_ign(ro, RTF_PRCLONING);
289 if (ro->ro_rt == NULL) {
290 ipstat.ips_noroute++;
291 error = EHOSTUNREACH;
294 ia = ifatoia(ro->ro_rt->rt_ifa);
295 ifp = ro->ro_rt->rt_ifp;
297 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
298 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
299 if (ro->ro_rt->rt_flags & RTF_HOST)
300 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
302 isbroadcast = in_broadcast(dst->sin_addr, ifp);
304 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
305 struct in_multi *inm;
307 m->m_flags |= M_MCAST;
309 * IP destination address is multicast. Make sure "dst"
310 * still points to the address in "ro". (It may have been
311 * changed to point to a gateway address, above.)
313 dst = (struct sockaddr_in *)&ro->ro_dst;
315 * See if the caller provided any multicast options
318 ip->ip_ttl = imo->imo_multicast_ttl;
319 if (imo->imo_multicast_vif != -1)
322 ip_mcast_src(imo->imo_multicast_vif) :
325 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
327 * Confirm that the outgoing interface supports multicast.
329 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
330 if (!(ifp->if_flags & IFF_MULTICAST)) {
331 ipstat.ips_noroute++;
337 * If source address not specified yet, use address
338 * of outgoing interface.
340 if (ip->ip_src.s_addr == INADDR_ANY) {
341 /* Interface may have no addresses. */
343 ip->ip_src = IA_SIN(ia)->sin_addr;
346 IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
348 (imo == NULL || imo->imo_multicast_loop)) {
350 * If we belong to the destination multicast group
351 * on the outgoing interface, and the caller did not
352 * forbid loopback, loop back a copy.
354 ip_mloopback(ifp, m, dst, hlen);
358 * If we are acting as a multicast router, perform
359 * multicast forwarding as if the packet had just
360 * arrived on the interface to which we are about
361 * to send. The multicast forwarding function
362 * recursively calls this function, using the
363 * IP_FORWARDING flag to prevent infinite recursion.
365 * Multicasts that are looped back by ip_mloopback(),
366 * above, will be forwarded by the ip_input() routine,
369 if (ip_mrouter && !(flags & IP_FORWARDING)) {
371 * If rsvp daemon is not running, do not
372 * set ip_moptions. This ensures that the packet
373 * is multicast and not just sent down one link
374 * as prescribed by rsvpd.
379 ip_mforward(ip, ifp, m, imo) != 0) {
387 * Multicasts with a time-to-live of zero may be looped-
388 * back, above, but must not be transmitted on a network.
389 * Also, multicasts addressed to the loopback interface
390 * are not sent -- the above call to ip_mloopback() will
391 * loop back a copy if this host actually belongs to the
392 * destination group on the loopback interface.
394 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
403 * If the source address is not specified yet, use the address
404 * of the outoing interface. In case, keep note we did that, so
405 * if the the firewall changes the next-hop causing the output
406 * interface to change, we can fix that.
408 if (ip->ip_src.s_addr == INADDR_ANY) {
409 /* Interface may have no addresses. */
411 ip->ip_src = IA_SIN(ia)->sin_addr;
412 src_was_INADDR_ANY = 1;
418 * Disable packet drop hack.
419 * Packetdrop should be done by queueing.
423 * Verify that we have any chance at all of being able to queue
424 * the packet or packet fragments
426 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
427 ifp->if_snd.ifq_maxlen) {
429 ipstat.ips_odropped++;
435 * Look for broadcast address and
436 * verify user is allowed to send
440 if (!(ifp->if_flags & IFF_BROADCAST)) {
441 error = EADDRNOTAVAIL;
444 if (!(flags & IP_ALLOWBROADCAST)) {
448 /* don't allow broadcast messages to be fragmented */
449 if (ip->ip_len > ifp->if_mtu) {
453 m->m_flags |= M_BCAST;
455 m->m_flags &= ~M_BCAST;
460 /* get SP for this packet */
462 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
464 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
467 ipsecstat.out_inval++;
474 switch (sp->policy) {
475 case IPSEC_POLICY_DISCARD:
477 * This packet is just discarded.
479 ipsecstat.out_polvio++;
482 case IPSEC_POLICY_BYPASS:
483 case IPSEC_POLICY_NONE:
484 /* no need to do IPsec. */
487 case IPSEC_POLICY_IPSEC:
488 if (sp->req == NULL) {
489 /* acquire a policy */
490 error = key_spdacquire(sp);
495 case IPSEC_POLICY_ENTRUST:
497 kprintf("ip_output: Invalid policy found. %d\n", sp->policy);
500 struct ipsec_output_state state;
501 bzero(&state, sizeof state);
503 if (flags & IP_ROUTETOIF) {
505 bzero(&iproute, sizeof iproute);
508 state.dst = (struct sockaddr *)dst;
514 * delayed checksums are not currently compatible with IPsec
516 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
518 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
521 ip->ip_len = htons(ip->ip_len);
522 ip->ip_off = htons(ip->ip_off);
524 error = ipsec4_output(&state, sp, flags);
527 if (flags & IP_ROUTETOIF) {
529 * if we have tunnel mode SA, we may need to ignore
532 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
533 flags &= ~IP_ROUTETOIF;
538 dst = (struct sockaddr_in *)state.dst;
540 /* mbuf is already reclaimed in ipsec4_output. */
550 kprintf("ip4_output (ipsec): error code %d\n", error);
553 /* don't show these error codes to the user */
561 /* be sure to update variables that are affected by ipsec4_output() */
562 ip = mtod(m, struct ip *);
564 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
566 hlen = ip->ip_hl << 2;
568 if (ro->ro_rt == NULL) {
569 if (!(flags & IP_ROUTETOIF)) {
570 kprintf("ip_output: "
571 "can't update route after IPsec processing\n");
572 error = EHOSTUNREACH; /*XXX*/
576 ia = ifatoia(ro->ro_rt->rt_ifa);
577 ifp = ro->ro_rt->rt_ifp;
580 /* make it flipped, again. */
581 ip->ip_len = ntohs(ip->ip_len);
582 ip->ip_off = ntohs(ip->ip_off);
587 * Check the security policy (SP) for the packet and, if
588 * required, do IPsec-related processing. There are two
589 * cases here; the first time a packet is sent through
590 * it will be untagged and handled by ipsec4_checkpolicy.
591 * If the packet is resubmitted to ip_output (e.g. after
592 * AH, ESP, etc. processing), there will be a tag to bypass
593 * the lookup and related policy checking.
595 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
598 tdbi = (struct tdb_ident *)m_tag_data(mtag);
599 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
601 error = -EINVAL; /* force silent drop */
602 m_tag_delete(m, mtag);
604 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
608 * There are four return cases:
609 * sp != NULL apply IPsec policy
610 * sp == NULL, error == 0 no IPsec handling needed
611 * sp == NULL, error == -EINVAL discard packet w/o error
612 * sp == NULL, error != 0 discard packet, report error
615 /* Loop detection, check if ipsec processing already done */
616 KASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
617 for (mtag = m_tag_first(m); mtag != NULL;
618 mtag = m_tag_next(m, mtag)) {
619 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
621 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
622 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
625 * Check if policy has an SA associated with it.
626 * This can happen when an SP has yet to acquire
627 * an SA; e.g. on first reference. If it occurs,
628 * then we let ipsec4_process_packet do its thing.
630 if (sp->req->sav == NULL)
632 tdbi = (struct tdb_ident *)m_tag_data(mtag);
633 if (tdbi->spi == sp->req->sav->spi &&
634 tdbi->proto == sp->req->sav->sah->saidx.proto &&
635 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
636 sizeof(union sockaddr_union)) == 0) {
638 * No IPsec processing is needed, free
641 * NB: null pointer to avoid free at
644 KEY_FREESP(&sp), sp = NULL;
651 * Do delayed checksums now because we send before
652 * this is done in the normal processing path.
654 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
656 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
659 ip->ip_len = htons(ip->ip_len);
660 ip->ip_off = htons(ip->ip_off);
662 /* NB: callee frees mbuf */
663 error = ipsec4_process_packet(m, sp->req, flags, 0);
665 * Preserve KAME behaviour: ENOENT can be returned
666 * when an SA acquire is in progress. Don't propagate
667 * this to user-level; it confuses applications.
669 * XXX this will go away when the SADB is redone.
680 * Hack: -EINVAL is used to signal that a packet
681 * should be silently discarded. This is typically
682 * because we asked key management for an SA and
683 * it was delayed (e.g. kicked up to IKE).
685 if (error == -EINVAL)
689 /* No IPsec processing for this packet. */
693 * If deferred crypto processing is needed, check that
694 * the interface supports it.
696 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
697 if (mtag != NULL && !(ifp->if_capenable & IFCAP_IPSEC)) {
698 /* notify IPsec to do its own crypto */
699 ipsp_skipcrypto_unmark((struct tdb_ident *)m_tag_data(mtag));
700 error = EHOSTUNREACH;
706 #endif /* FAST_IPSEC */
709 * - Xlate: translate packet's addr/port (NAT).
710 * - Firewall: deny/allow/etc.
711 * - Wrap: fake packet's addr/port <unimpl.>
712 * - Encapsulate: put it in another IP and send out. <unimp.>
716 * Run through list of hooks for output packets.
718 if (pfil_has_hooks(&inet_pfil_hook)) {
719 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT);
720 if (error != 0 || m == NULL)
722 ip = mtod(m, struct ip *);
726 * Check with the firewall...
727 * but not if we are already being fwd'd from a firewall.
729 if (fw_enable && IPFW_LOADED && !args.next_hop) {
730 struct sockaddr_in *old = dst;
735 off = ip_fw_chk_ptr(&args);
740 * On return we must do the following:
741 * m == NULL -> drop the pkt (old interface, deprecated)
742 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface)
743 * 1<=off<= 0xffff -> DIVERT
744 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe
745 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet
746 * dst != old -> IPFIREWALL_FORWARD
747 * off==0, dst==old -> accept
748 * If some of the above modules are not compiled in, then
749 * we should't have to check the corresponding condition
750 * (because the ipfw control socket should not accept
751 * unsupported rules), but better play safe and drop
752 * packets in case of doubt.
754 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
760 ip = mtod(m, struct ip *);
761 if (off == 0 && dst == old) /* common case */
763 if (off & IP_FW_PORT_DYNT_FLAG) {
765 * pass the pkt to dummynet. Need to include
766 * pipe number, m, ifp, ro, dst because these are
767 * not recomputed in the next pass.
768 * All other parameters have been already used and
769 * so they are not needed anymore.
770 * XXX note: if the ifp or ro entry are deleted
771 * while a pkt is in dummynet, we are in trouble!
778 ip_fw_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT, &args);
782 if (off != 0 && !(off & IP_FW_PORT_DYNT_FLAG)) {
783 struct mbuf *clone = NULL;
785 /* Clone packet if we're doing a 'tee' */
786 if ((off & IP_FW_PORT_TEE_FLAG))
787 clone = m_dup(m, MB_DONTWAIT);
791 * delayed checksums are not currently compatible
792 * with divert sockets.
794 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
796 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
799 /* Restore packet header fields to original values */
800 ip->ip_len = htons(ip->ip_len);
801 ip->ip_off = htons(ip->ip_off);
803 /* Deliver packet to divert input routine */
804 divert_packet(m, 0, off & 0xffff);
806 /* If 'tee', continue with original packet */
809 ip = mtod(m, struct ip *);
816 /* IPFIREWALL_FORWARD */
818 * Check dst to make sure it is directly reachable on the
819 * interface we previously thought it was.
820 * If it isn't (which may be likely in some situations) we have
821 * to re-route it (ie, find a route for the next-hop and the
822 * associated interface) and set them here. This is nested
823 * forwarding which in most cases is undesirable, except where
824 * such control is nigh impossible. So we do it here.
827 if (off == 0 && old != dst) { /* FORWARD, dst has changed */
830 * XXX To improve readability, this block should be
831 * changed into a function call as below:
833 error = ip_ipforward(&m, &dst, &ifp);
836 if (m == NULL) /* ip_input consumed the mbuf */
839 struct in_ifaddr *ia;
840 struct in_ifaddr_container *iac;
843 * XXX sro_fwd below is static, and a pointer
844 * to it gets passed to routines downstream.
845 * This could have surprisingly bad results in
846 * practice, because its content is overwritten
847 * by subsequent packets.
849 /* There must be a better way to do this next line... */
850 static struct route sro_fwd;
851 struct route *ro_fwd = &sro_fwd;
854 print_ip("IPFIREWALL_FORWARD: New dst ip: ",
855 dst->sin_addr, "\n");
859 * We need to figure out if we have been forwarded
860 * to a local socket. If so, then we should somehow
861 * "loop back" to ip_input, and get directed to the
862 * PCB as if we had received this packet. This is
863 * because it may be dificult to identify the packets
864 * you want to forward until they are being output
865 * and have selected an interface. (e.g. locally
866 * initiated packets) If we used the loopback inteface,
867 * we would not be able to control what happens
868 * as the packet runs through ip_input() as
869 * it is done through a ISR.
872 LIST_FOREACH(iac, INADDR_HASH(dst->sin_addr.s_addr),
875 * If the addr to forward to is one
876 * of ours, we pretend to
877 * be the destination for this packet.
879 if (IA_SIN(iac->ia)->sin_addr.s_addr ==
880 dst->sin_addr.s_addr) {
885 if (ia != NULL) { /* tell ip_input "dont filter" */
888 tag.mh_type = MT_TAG;
889 tag.mh_flags = PACKET_TAG_IPFORWARD;
890 tag.mh_data = (caddr_t)args.next_hop;
893 if (m->m_pkthdr.rcvif == NULL)
894 m->m_pkthdr.rcvif = ifunit("lo0");
895 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
896 m->m_pkthdr.csum_flags |=
897 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
898 m->m_pkthdr.csum_data = 0xffff;
900 m->m_pkthdr.csum_flags |=
901 CSUM_IP_CHECKED | CSUM_IP_VALID;
902 ip->ip_len = htons(ip->ip_len);
903 ip->ip_off = htons(ip->ip_off);
904 ip_input((struct mbuf *)&tag);
907 /* Some of the logic for this was nicked from above.
909 * This rewrites the cached route in a local PCB.
910 * Is this what we want to do?
912 bcopy(dst, &ro_fwd->ro_dst, sizeof *dst);
913 ro_fwd->ro_rt = NULL;
915 rtalloc_ign(ro_fwd, RTF_PRCLONING);
916 if (ro_fwd->ro_rt == NULL) {
917 ipstat.ips_noroute++;
918 error = EHOSTUNREACH;
922 ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
923 ifp = ro_fwd->ro_rt->rt_ifp;
924 ro_fwd->ro_rt->rt_use++;
925 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
926 dst = (struct sockaddr_in *)
927 ro_fwd->ro_rt->rt_gateway;
928 if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
930 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
932 isbroadcast = in_broadcast(dst->sin_addr, ifp);
933 if (ro->ro_rt != NULL)
935 ro->ro_rt = ro_fwd->ro_rt;
936 dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
938 #endif /* ... block to be put into a function */
940 * If we added a default src ip earlier,
941 * which would have been gotten from the-then
942 * interface, do it again, from the new one.
944 if (src_was_INADDR_ANY)
945 ip->ip_src = IA_SIN(ia)->sin_addr;
950 * if we get here, none of the above matches, and
951 * we have to drop the pkt
954 error = EACCES; /* not sure this is the right error msg */
959 /* 127/8 must not appear on wire - RFC1122. */
960 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
961 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
962 if (!(ifp->if_flags & IFF_LOOPBACK)) {
963 ipstat.ips_badaddr++;
964 error = EADDRNOTAVAIL;
969 m->m_pkthdr.csum_flags |= CSUM_IP;
970 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
971 if (sw_csum & CSUM_DELAY_DATA) {
973 sw_csum &= ~CSUM_DELAY_DATA;
975 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
978 * If small enough for interface, or the interface will take
979 * care of the fragmentation for us, can just send directly.
981 if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
982 !(ip->ip_off & IP_DF))) {
983 ip->ip_len = htons(ip->ip_len);
984 ip->ip_off = htons(ip->ip_off);
986 if (sw_csum & CSUM_DELAY_IP) {
987 if (ip->ip_vhl == IP_VHL_BORING) {
988 ip->ip_sum = in_cksum_hdr(ip);
990 ip->ip_sum = in_cksum(m, hlen);
994 /* Record statistics for this interface address. */
995 if (!(flags & IP_FORWARDING) && ia) {
996 ia->ia_ifa.if_opackets++;
997 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1001 /* clean ipsec history once it goes out of the node */
1005 #ifdef MBUF_STRESS_TEST
1006 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) {
1007 struct mbuf *m1, *m2;
1010 tmp = length = m->m_pkthdr.len;
1012 while ((length -= mbuf_frag_size) >= 1) {
1013 m1 = m_split(m, length, MB_DONTWAIT);
1017 while (m2->m_next != NULL)
1021 m->m_pkthdr.len = tmp;
1024 error = ifp->if_output(ifp, m, (struct sockaddr *)dst,
1029 if (ip->ip_off & IP_DF) {
1032 * This case can happen if the user changed the MTU
1033 * of an interface after enabling IP on it. Because
1034 * most netifs don't keep track of routes pointing to
1035 * them, there is no way for one to update all its
1036 * routes when the MTU is changed.
1038 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
1039 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
1040 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
1041 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
1043 ipstat.ips_cantfrag++;
1048 * Too large for interface; fragment if possible. If successful,
1049 * on return, m will point to a list of packets to be sent.
1051 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
1056 m->m_nextpkt = NULL;
1058 /* clean ipsec history once it goes out of the node */
1062 /* Record statistics for this interface address. */
1064 ia->ia_ifa.if_opackets++;
1065 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1067 error = ifp->if_output(ifp, m, (struct sockaddr *)dst,
1075 ipstat.ips_fragmented++;
1078 if (ro == &iproute && ro->ro_rt != NULL) {
1084 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1085 kprintf("DP ip_output call free SP:%p\n", sp));
1093 if (dn_mtag != NULL)
1094 m_tag_free(dn_mtag);
1103 * Create a chain of fragments which fit the given mtu. m_frag points to the
1104 * mbuf to be fragmented; on return it points to the chain with the fragments.
1105 * Return 0 if no error. If error, m_frag may contain a partially built
1106 * chain of fragments that should be freed by the caller.
1108 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
1109 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
1112 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
1113 u_long if_hwassist_flags, int sw_csum)
1116 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1117 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
1119 struct mbuf *m0 = *m_frag; /* the original packet */
1121 struct mbuf **mnext;
1124 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
1125 ipstat.ips_cantfrag++;
1130 * Must be able to put at least 8 bytes per fragment.
1136 * If the interface will not calculate checksums on
1137 * fragmented packets, then do it here.
1139 if ((m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) &&
1140 !(if_hwassist_flags & CSUM_IP_FRAGS)) {
1141 in_delayed_cksum(m0);
1142 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1145 if (len > PAGE_SIZE) {
1147 * Fragment large datagrams such that each segment
1148 * contains a multiple of PAGE_SIZE amount of data,
1149 * plus headers. This enables a receiver to perform
1150 * page-flipping zero-copy optimizations.
1152 * XXX When does this help given that sender and receiver
1153 * could have different page sizes, and also mtu could
1154 * be less than the receiver's page size ?
1159 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
1163 * firstlen (off - hlen) must be aligned on an
1167 goto smart_frag_failure;
1168 off = ((off - hlen) & ~7) + hlen;
1169 newlen = (~PAGE_MASK) & mtu;
1170 if ((newlen + sizeof(struct ip)) > mtu) {
1171 /* we failed, go back the default */
1182 firstlen = off - hlen;
1183 mnext = &m0->m_nextpkt; /* pointer to next packet */
1186 * Loop through length of segment after first fragment,
1187 * make new header and copy data of each part and link onto chain.
1188 * Here, m0 is the original packet, m is the fragment being created.
1189 * The fragments are linked off the m_nextpkt of the original
1190 * packet, which after processing serves as the first fragment.
1192 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
1193 struct ip *mhip; /* ip header on the fragment */
1195 int mhlen = sizeof(struct ip);
1197 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1200 ipstat.ips_odropped++;
1203 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
1205 * In the first mbuf, leave room for the link header, then
1206 * copy the original IP header including options. The payload
1207 * goes into an additional mbuf chain returned by m_copy().
1209 m->m_data += max_linkhdr;
1210 mhip = mtod(m, struct ip *);
1212 if (hlen > sizeof(struct ip)) {
1213 mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip);
1214 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
1217 /* XXX do we need to add ip->ip_off below ? */
1218 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
1219 if (off + len >= ip->ip_len) { /* last fragment */
1220 len = ip->ip_len - off;
1221 m->m_flags |= M_LASTFRAG;
1223 mhip->ip_off |= IP_MF;
1224 mhip->ip_len = htons((u_short)(len + mhlen));
1225 m->m_next = m_copy(m0, off, len);
1226 if (m->m_next == NULL) { /* copy failed */
1228 error = ENOBUFS; /* ??? */
1229 ipstat.ips_odropped++;
1232 m->m_pkthdr.len = mhlen + len;
1233 m->m_pkthdr.rcvif = (struct ifnet *)NULL;
1234 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1235 mhip->ip_off = htons(mhip->ip_off);
1237 if (sw_csum & CSUM_DELAY_IP)
1238 mhip->ip_sum = in_cksum(m, mhlen);
1240 mnext = &m->m_nextpkt;
1242 ipstat.ips_ofragments += nfrags;
1244 /* set first marker for fragment chain */
1245 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1246 m0->m_pkthdr.csum_data = nfrags;
1249 * Update first fragment by trimming what's been copied out
1250 * and updating header.
1252 m_adj(m0, hlen + firstlen - ip->ip_len);
1253 m0->m_pkthdr.len = hlen + firstlen;
1254 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1255 ip->ip_off |= IP_MF;
1256 ip->ip_off = htons(ip->ip_off);
1258 if (sw_csum & CSUM_DELAY_IP)
1259 ip->ip_sum = in_cksum(m0, hlen);
1267 in_delayed_cksum(struct mbuf *m)
1270 u_short csum, offset;
1272 ip = mtod(m, struct ip *);
1273 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1274 csum = in_cksum_skip(m, ip->ip_len, offset);
1275 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1277 offset += m->m_pkthdr.csum_data; /* checksum offset */
1279 if (offset + sizeof(u_short) > m->m_len) {
1280 kprintf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1281 m->m_len, offset, ip->ip_p);
1284 * this shouldn't happen, but if it does, the
1285 * correct behavior may be to insert the checksum
1286 * in the existing chain instead of rearranging it.
1288 m = m_pullup(m, offset + sizeof(u_short));
1290 *(u_short *)(m->m_data + offset) = csum;
1294 * Insert IP options into preformed packet.
1295 * Adjust IP destination as required for IP source routing,
1296 * as indicated by a non-zero in_addr at the start of the options.
1298 * XXX This routine assumes that the packet has no options in place.
1300 static struct mbuf *
1301 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
1303 struct ipoption *p = mtod(opt, struct ipoption *);
1305 struct ip *ip = mtod(m, struct ip *);
1308 optlen = opt->m_len - sizeof p->ipopt_dst;
1309 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
1311 return (m); /* XXX should fail */
1313 if (p->ipopt_dst.s_addr)
1314 ip->ip_dst = p->ipopt_dst;
1315 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1316 MGETHDR(n, MB_DONTWAIT, MT_HEADER);
1321 n->m_pkthdr.rcvif = (struct ifnet *)NULL;
1322 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1323 m->m_len -= sizeof(struct ip);
1324 m->m_data += sizeof(struct ip);
1327 m->m_len = optlen + sizeof(struct ip);
1328 m->m_data += max_linkhdr;
1329 memcpy(mtod(m, void *), ip, sizeof(struct ip));
1331 m->m_data -= optlen;
1333 m->m_pkthdr.len += optlen;
1334 ovbcopy(ip, mtod(m, caddr_t), sizeof(struct ip));
1336 ip = mtod(m, struct ip *);
1337 bcopy(p->ipopt_list, ip + 1, optlen);
1338 *phlen = sizeof(struct ip) + optlen;
1339 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1340 ip->ip_len += optlen;
1345 * Copy options from ip to jp,
1346 * omitting those not copied during fragmentation.
1349 ip_optcopy(struct ip *ip, struct ip *jp)
1352 int opt, optlen, cnt;
1354 cp = (u_char *)(ip + 1);
1355 dp = (u_char *)(jp + 1);
1356 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1357 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1359 if (opt == IPOPT_EOL)
1361 if (opt == IPOPT_NOP) {
1362 /* Preserve for IP mcast tunnel's LSRR alignment. */
1368 KASSERT(cnt >= IPOPT_OLEN + sizeof *cp,
1369 ("ip_optcopy: malformed ipv4 option"));
1370 optlen = cp[IPOPT_OLEN];
1371 KASSERT(optlen >= IPOPT_OLEN + sizeof *cp && optlen <= cnt,
1372 ("ip_optcopy: malformed ipv4 option"));
1374 /* bogus lengths should have been caught by ip_dooptions */
1377 if (IPOPT_COPIED(opt)) {
1378 bcopy(cp, dp, optlen);
1382 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1388 * IP socket option processing.
1391 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1393 struct inpcb *inp = so->so_pcb;
1397 if (sopt->sopt_level != IPPROTO_IP) {
1401 switch (sopt->sopt_dir) {
1403 switch (sopt->sopt_name) {
1410 if (sopt->sopt_valsize > MLEN) {
1414 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1419 m->m_len = sopt->sopt_valsize;
1420 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1423 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1431 case IP_RECVRETOPTS:
1432 case IP_RECVDSTADDR:
1436 error = sooptcopyin(sopt, &optval, sizeof optval,
1441 switch (sopt->sopt_name) {
1443 inp->inp_ip_tos = optval;
1447 inp->inp_ip_ttl = optval;
1450 if (optval > 0 && optval <= MAXTTL)
1451 inp->inp_ip_minttl = optval;
1455 #define OPTSET(bit) \
1457 inp->inp_flags |= bit; \
1459 inp->inp_flags &= ~bit;
1462 OPTSET(INP_RECVOPTS);
1465 case IP_RECVRETOPTS:
1466 OPTSET(INP_RECVRETOPTS);
1469 case IP_RECVDSTADDR:
1470 OPTSET(INP_RECVDSTADDR);
1478 OPTSET(INP_RECVTTL);
1488 case IP_MULTICAST_IF:
1489 case IP_MULTICAST_VIF:
1490 case IP_MULTICAST_TTL:
1491 case IP_MULTICAST_LOOP:
1492 case IP_ADD_MEMBERSHIP:
1493 case IP_DROP_MEMBERSHIP:
1494 error = ip_setmoptions(sopt, &inp->inp_moptions);
1498 error = sooptcopyin(sopt, &optval, sizeof optval,
1504 case IP_PORTRANGE_DEFAULT:
1505 inp->inp_flags &= ~(INP_LOWPORT);
1506 inp->inp_flags &= ~(INP_HIGHPORT);
1509 case IP_PORTRANGE_HIGH:
1510 inp->inp_flags &= ~(INP_LOWPORT);
1511 inp->inp_flags |= INP_HIGHPORT;
1514 case IP_PORTRANGE_LOW:
1515 inp->inp_flags &= ~(INP_HIGHPORT);
1516 inp->inp_flags |= INP_LOWPORT;
1525 #if defined(IPSEC) || defined(FAST_IPSEC)
1526 case IP_IPSEC_POLICY:
1534 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1536 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1538 priv = (sopt->sopt_td != NULL &&
1539 suser(sopt->sopt_td) != 0) ? 0 : 1;
1540 req = mtod(m, caddr_t);
1542 optname = sopt->sopt_name;
1543 error = ipsec4_set_policy(inp, optname, req, len, priv);
1550 error = ENOPROTOOPT;
1556 switch (sopt->sopt_name) {
1559 if (inp->inp_options)
1560 error = sooptcopyout(sopt,
1561 mtod(inp->inp_options,
1563 inp->inp_options->m_len);
1565 sopt->sopt_valsize = 0;
1572 case IP_RECVRETOPTS:
1573 case IP_RECVDSTADDR:
1578 switch (sopt->sopt_name) {
1581 optval = inp->inp_ip_tos;
1585 optval = inp->inp_ip_ttl;
1588 optval = inp->inp_ip_minttl;
1591 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1594 optval = OPTBIT(INP_RECVOPTS);
1597 case IP_RECVRETOPTS:
1598 optval = OPTBIT(INP_RECVRETOPTS);
1601 case IP_RECVDSTADDR:
1602 optval = OPTBIT(INP_RECVDSTADDR);
1606 optval = OPTBIT(INP_RECVTTL);
1610 optval = OPTBIT(INP_RECVIF);
1614 if (inp->inp_flags & INP_HIGHPORT)
1615 optval = IP_PORTRANGE_HIGH;
1616 else if (inp->inp_flags & INP_LOWPORT)
1617 optval = IP_PORTRANGE_LOW;
1623 optval = OPTBIT(INP_FAITH);
1626 error = sooptcopyout(sopt, &optval, sizeof optval);
1629 case IP_MULTICAST_IF:
1630 case IP_MULTICAST_VIF:
1631 case IP_MULTICAST_TTL:
1632 case IP_MULTICAST_LOOP:
1633 case IP_ADD_MEMBERSHIP:
1634 case IP_DROP_MEMBERSHIP:
1635 error = ip_getmoptions(sopt, inp->inp_moptions);
1638 #if defined(IPSEC) || defined(FAST_IPSEC)
1639 case IP_IPSEC_POLICY:
1641 struct mbuf *m = NULL;
1646 req = mtod(m, caddr_t);
1649 error = ipsec4_get_policy(so->so_pcb, req, len, &m);
1651 error = soopt_mcopyout(sopt, m); /* XXX */
1659 error = ENOPROTOOPT;
1668 * Set up IP options in pcb for insertion in output packets.
1669 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1670 * with destination address if source routed.
1673 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
1679 /* turn off any old options */
1683 if (m == NULL || m->m_len == 0) {
1685 * Only turning off any previous options.
1692 if (m->m_len % sizeof(int32_t))
1695 * IP first-hop destination address will be stored before
1696 * actual options; move other options back
1697 * and clear it when none present.
1699 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1702 m->m_len += sizeof(struct in_addr);
1703 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1704 ovbcopy(mtod(m, caddr_t), cp, cnt);
1705 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1707 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1708 opt = cp[IPOPT_OPTVAL];
1709 if (opt == IPOPT_EOL)
1711 if (opt == IPOPT_NOP)
1714 if (cnt < IPOPT_OLEN + sizeof *cp)
1716 optlen = cp[IPOPT_OLEN];
1717 if (optlen < IPOPT_OLEN + sizeof *cp || optlen > cnt)
1728 * user process specifies route as:
1730 * D must be our final destination (but we can't
1731 * check that since we may not have connected yet).
1732 * A is first hop destination, which doesn't appear in
1733 * actual IP option, but is stored before the options.
1735 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1737 m->m_len -= sizeof(struct in_addr);
1738 cnt -= sizeof(struct in_addr);
1739 optlen -= sizeof(struct in_addr);
1740 cp[IPOPT_OLEN] = optlen;
1742 * Move first hop before start of options.
1744 bcopy(&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1745 sizeof(struct in_addr));
1747 * Then copy rest of options back
1748 * to close up the deleted entry.
1750 ovbcopy(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr),
1751 &cp[IPOPT_OFFSET+1],
1752 cnt - (IPOPT_MINOFF - 1));
1756 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1768 * The whole multicast option thing needs to be re-thought.
1769 * Several of these options are equally applicable to non-multicast
1770 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1771 * standard option (IP_TTL).
1775 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1777 static struct ifnet *
1778 ip_multicast_if(struct in_addr *a, int *ifindexp)
1785 if (ntohl(a->s_addr) >> 24 == 0) {
1786 ifindex = ntohl(a->s_addr) & 0xffffff;
1787 if (ifindex < 0 || if_index < ifindex)
1789 ifp = ifindex2ifnet[ifindex];
1791 *ifindexp = ifindex;
1793 ifp = INADDR_TO_IFP(a);
1799 * Set the IP multicast options in response to user setsockopt().
1802 ip_setmoptions(struct sockopt *sopt, struct ip_moptions **imop)
1806 struct in_addr addr;
1807 struct ip_mreq mreq;
1809 struct ip_moptions *imo = *imop;
1814 * No multicast option buffer attached to the pcb;
1815 * allocate one and initialize to default values.
1817 imo = kmalloc(sizeof *imo, M_IPMOPTS, M_WAITOK);
1820 imo->imo_multicast_ifp = NULL;
1821 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1822 imo->imo_multicast_vif = -1;
1823 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1824 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1825 imo->imo_num_memberships = 0;
1828 switch (sopt->sopt_name) {
1829 /* store an index number for the vif you wanna use in the send */
1830 case IP_MULTICAST_VIF:
1831 if (legal_vif_num == 0) {
1835 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
1838 if (!legal_vif_num(i) && (i != -1)) {
1842 imo->imo_multicast_vif = i;
1845 case IP_MULTICAST_IF:
1847 * Select the interface for outgoing multicast packets.
1849 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr);
1853 * INADDR_ANY is used to remove a previous selection.
1854 * When no interface is selected, a default one is
1855 * chosen every time a multicast packet is sent.
1857 if (addr.s_addr == INADDR_ANY) {
1858 imo->imo_multicast_ifp = NULL;
1862 * The selected interface is identified by its local
1863 * IP address. Find the interface and confirm that
1864 * it supports multicasting.
1867 ifp = ip_multicast_if(&addr, &ifindex);
1868 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1870 error = EADDRNOTAVAIL;
1873 imo->imo_multicast_ifp = ifp;
1875 imo->imo_multicast_addr = addr;
1877 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1881 case IP_MULTICAST_TTL:
1883 * Set the IP time-to-live for outgoing multicast packets.
1884 * The original multicast API required a char argument,
1885 * which is inconsistent with the rest of the socket API.
1886 * We allow either a char or an int.
1888 if (sopt->sopt_valsize == 1) {
1890 error = sooptcopyin(sopt, &ttl, 1, 1);
1893 imo->imo_multicast_ttl = ttl;
1896 error = sooptcopyin(sopt, &ttl, sizeof ttl, sizeof ttl);
1902 imo->imo_multicast_ttl = ttl;
1906 case IP_MULTICAST_LOOP:
1908 * Set the loopback flag for outgoing multicast packets.
1909 * Must be zero or one. The original multicast API required a
1910 * char argument, which is inconsistent with the rest
1911 * of the socket API. We allow either a char or an int.
1913 if (sopt->sopt_valsize == 1) {
1916 error = sooptcopyin(sopt, &loop, 1, 1);
1919 imo->imo_multicast_loop = !!loop;
1923 error = sooptcopyin(sopt, &loop, sizeof loop,
1927 imo->imo_multicast_loop = !!loop;
1931 case IP_ADD_MEMBERSHIP:
1933 * Add a multicast group membership.
1934 * Group must be a valid IP multicast address.
1936 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1940 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1946 * If no interface address was provided, use the interface of
1947 * the route to the given multicast address.
1949 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1950 struct sockaddr_in dst;
1953 bzero(&dst, sizeof(struct sockaddr_in));
1954 dst.sin_len = sizeof(struct sockaddr_in);
1955 dst.sin_family = AF_INET;
1956 dst.sin_addr = mreq.imr_multiaddr;
1957 rt = rtlookup((struct sockaddr *)&dst);
1959 error = EADDRNOTAVAIL;
1966 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1970 * See if we found an interface, and confirm that it
1971 * supports multicast.
1973 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1974 error = EADDRNOTAVAIL;
1979 * See if the membership already exists or if all the
1980 * membership slots are full.
1982 for (i = 0; i < imo->imo_num_memberships; ++i) {
1983 if (imo->imo_membership[i]->inm_ifp == ifp &&
1984 imo->imo_membership[i]->inm_addr.s_addr
1985 == mreq.imr_multiaddr.s_addr)
1988 if (i < imo->imo_num_memberships) {
1993 if (i == IP_MAX_MEMBERSHIPS) {
1994 error = ETOOMANYREFS;
1999 * Everything looks good; add a new record to the multicast
2000 * address list for the given interface.
2002 if ((imo->imo_membership[i] =
2003 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
2008 ++imo->imo_num_memberships;
2012 case IP_DROP_MEMBERSHIP:
2014 * Drop a multicast group membership.
2015 * Group must be a valid IP multicast address.
2017 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
2021 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
2028 * If an interface address was specified, get a pointer
2029 * to its ifnet structure.
2031 if (mreq.imr_interface.s_addr == INADDR_ANY)
2034 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
2036 error = EADDRNOTAVAIL;
2042 * Find the membership in the membership array.
2044 for (i = 0; i < imo->imo_num_memberships; ++i) {
2046 imo->imo_membership[i]->inm_ifp == ifp) &&
2047 imo->imo_membership[i]->inm_addr.s_addr ==
2048 mreq.imr_multiaddr.s_addr)
2051 if (i == imo->imo_num_memberships) {
2052 error = EADDRNOTAVAIL;
2057 * Give up the multicast address record to which the
2058 * membership points.
2060 in_delmulti(imo->imo_membership[i]);
2062 * Remove the gap in the membership array.
2064 for (++i; i < imo->imo_num_memberships; ++i)
2065 imo->imo_membership[i-1] = imo->imo_membership[i];
2066 --imo->imo_num_memberships;
2076 * If all options have default values, no need to keep the mbuf.
2078 if (imo->imo_multicast_ifp == NULL &&
2079 imo->imo_multicast_vif == -1 &&
2080 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2081 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2082 imo->imo_num_memberships == 0) {
2083 kfree(*imop, M_IPMOPTS);
2091 * Return the IP multicast options in response to user getsockopt().
2094 ip_getmoptions(struct sockopt *sopt, struct ip_moptions *imo)
2096 struct in_addr addr;
2097 struct in_ifaddr *ia;
2102 switch (sopt->sopt_name) {
2103 case IP_MULTICAST_VIF:
2105 optval = imo->imo_multicast_vif;
2108 error = sooptcopyout(sopt, &optval, sizeof optval);
2111 case IP_MULTICAST_IF:
2112 if (imo == NULL || imo->imo_multicast_ifp == NULL)
2113 addr.s_addr = INADDR_ANY;
2114 else if (imo->imo_multicast_addr.s_addr) {
2115 /* return the value user has set */
2116 addr = imo->imo_multicast_addr;
2118 ia = IFP_TO_IA(imo->imo_multicast_ifp);
2119 addr.s_addr = (ia == NULL) ? INADDR_ANY
2120 : IA_SIN(ia)->sin_addr.s_addr;
2122 error = sooptcopyout(sopt, &addr, sizeof addr);
2125 case IP_MULTICAST_TTL:
2127 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
2129 optval = coptval = imo->imo_multicast_ttl;
2130 if (sopt->sopt_valsize == 1)
2131 error = sooptcopyout(sopt, &coptval, 1);
2133 error = sooptcopyout(sopt, &optval, sizeof optval);
2136 case IP_MULTICAST_LOOP:
2138 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
2140 optval = coptval = imo->imo_multicast_loop;
2141 if (sopt->sopt_valsize == 1)
2142 error = sooptcopyout(sopt, &coptval, 1);
2144 error = sooptcopyout(sopt, &optval, sizeof optval);
2148 error = ENOPROTOOPT;
2155 * Discard the IP multicast options.
2158 ip_freemoptions(struct ip_moptions *imo)
2163 for (i = 0; i < imo->imo_num_memberships; ++i)
2164 in_delmulti(imo->imo_membership[i]);
2165 kfree(imo, M_IPMOPTS);
2170 * Routine called from ip_output() to loop back a copy of an IP multicast
2171 * packet to the input queue of a specified interface. Note that this
2172 * calls the output routine of the loopback "driver", but with an interface
2173 * pointer that might NOT be a loopback interface -- evil, but easier than
2174 * replicating that code here.
2177 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
2183 copym = m_copypacket(m, MB_DONTWAIT);
2184 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2185 copym = m_pullup(copym, hlen);
2186 if (copym != NULL) {
2188 * if the checksum hasn't been computed, mark it as valid
2190 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2191 in_delayed_cksum(copym);
2192 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2193 copym->m_pkthdr.csum_flags |=
2194 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2195 copym->m_pkthdr.csum_data = 0xffff;
2198 * We don't bother to fragment if the IP length is greater
2199 * than the interface's MTU. Can this possibly matter?
2201 ip = mtod(copym, struct ip *);
2202 ip->ip_len = htons(ip->ip_len);
2203 ip->ip_off = htons(ip->ip_off);
2205 if (ip->ip_vhl == IP_VHL_BORING) {
2206 ip->ip_sum = in_cksum_hdr(ip);
2208 ip->ip_sum = in_cksum(copym, hlen);
2212 * It's not clear whether there are any lingering
2213 * reentrancy problems in other areas which might
2214 * be exposed by using ip_input directly (in
2215 * particular, everything which modifies the packet
2216 * in-place). Yet another option is using the
2217 * protosw directly to deliver the looped back
2218 * packet. For the moment, we'll err on the side
2219 * of safety by using if_simloop().
2222 if (dst->sin_family != AF_INET) {
2223 kprintf("ip_mloopback: bad address family %d\n",
2225 dst->sin_family = AF_INET;
2230 copym->m_pkthdr.rcvif = ifp;
2233 if_simloop(ifp, copym, dst->sin_family, 0);