2 * Copyright (c) 2000-2007 Apple Inc. All rights reserved.
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
29 * Copyright (c) 1982, 1986, 1988, 1990, 1993
30 * The Regents of the University of California. All rights reserved.
32 * Redistribution and use in source and binary forms, with or without
33 * modification, are permitted provided that the following conditions
35 * 1. Redistributions of source code must retain the above copyright
36 * notice, this list of conditions and the following disclaimer.
37 * 2. Redistributions in binary form must reproduce the above copyright
38 * notice, this list of conditions and the following disclaimer in the
39 * documentation and/or other materials provided with the distribution.
40 * 3. All advertising materials mentioning features or use of this software
41 * must display the following acknowledgement:
42 * This product includes software developed by the University of
43 * California, Berkeley and its contributors.
44 * 4. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
61 * $FreeBSD: src/sys/netinet/ip_output.c,v 1.99.2.16 2001/07/19 06:37:26 kris Exp $
64 * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
65 * support for mandatory and extensible security protections. This notice
66 * is included in support of clause 2.2 (b) of the Apple Public License,
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/kernel.h>
75 #include <sys/malloc.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <kern/locks.h>
81 #include <sys/sysctl.h>
84 #include <net/route.h>
86 #include <netinet/in.h>
87 #include <netinet/in_systm.h>
88 #include <netinet/ip.h>
89 #include <netinet/in_pcb.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip_var.h>
93 #include <netinet/kpi_ipfilter_var.h>
96 #include <security/mac_framework.h>
101 #include <net/dlil.h>
102 #include <sys/kdebug.h>
103 #include <libkern/OSAtomic.h>
105 #define DBG_LAYER_BEG NETDBG_CODE(DBG_NETIP, 1)
106 #define DBG_LAYER_END NETDBG_CODE(DBG_NETIP, 3)
107 #define DBG_FNC_IP_OUTPUT NETDBG_CODE(DBG_NETIP, (1 << 8) | 1)
108 #define DBG_FNC_IPSEC4_OUTPUT NETDBG_CODE(DBG_NETIP, (2 << 8) | 1)
110 #define SWAP16(v) ((((v) & 0xff) << 8) | ((v) >> 8))
113 #include <netinet6/ipsec.h>
114 #include <netkey/key.h>
116 #include <netkey/key_debug.h>
118 #define KEYDEBUG(lev,arg)
122 #include <netinet/ip_fw.h>
123 #include <netinet/ip_divert.h>
126 #include <netinet/ip_dummynet.h>
129 #if IPFIREWALL_FORWARD_DEBUG
130 #define print_ip(a) printf("%ld.%ld.%ld.%ld",(ntohl(a.s_addr)>>24)&0xFF,\
131 (ntohl(a.s_addr)>>16)&0xFF,\
132 (ntohl(a.s_addr)>>8)&0xFF,\
133 (ntohl(a.s_addr))&0xFF);
139 static struct mbuf
*ip_insertoptions(struct mbuf
*, struct mbuf
*, int *);
140 static struct ifnet
*ip_multicast_if(struct in_addr
*, int *);
141 static void ip_mloopback(struct ifnet
*, struct mbuf
*,
142 struct sockaddr_in
*, int);
143 static int ip_getmoptions(struct sockopt
*, struct ip_moptions
*);
144 static int ip_pcbopts(int, struct mbuf
**, struct mbuf
*);
145 static int ip_setmoptions(struct sockopt
*, struct ip_moptions
**);
147 static void ip_out_cksum_stats(int, u_int32_t
);
149 int ip_createmoptions(struct ip_moptions
**imop
);
150 int ip_addmembership(struct ip_moptions
*imo
, struct ip_mreq
*mreq
);
151 int ip_dropmembership(struct ip_moptions
*imo
, struct ip_mreq
*mreq
);
152 int ip_optcopy(struct ip
*, struct ip
*);
153 void in_delayed_cksum_offset(struct mbuf
*, int );
154 void in_cksum_offset(struct mbuf
* , size_t );
156 extern int (*fr_checkp
)(struct ip
*, int, struct ifnet
*, int, struct mbuf
**);
158 extern u_long route_generation
;
160 extern struct protosw inetsw
[];
162 extern struct ip_linklocal_stat ip_linklocal_stat
;
163 extern lck_mtx_t
*ip_mutex
;
165 /* temporary: for testing */
167 extern int ipsec_bypass
;
170 static int ip_maxchainsent
= 0;
171 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, maxchainsent
, CTLFLAG_RW
,
172 &ip_maxchainsent
, 0, "use dlil_output_list");
174 static int forge_ce
= 0;
175 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, forge_ce
, CTLFLAG_RW
,
176 &forge_ce
, 0, "Forge ECN CE");
179 * IP output. The packet in mbuf chain m contains a skeletal IP
180 * header (with len, off, ttl, proto, tos, src, dst).
181 * The mbuf chain containing the packet will be freed.
182 * The mbuf opt, if present, will not be freed.
190 struct ip_moptions
*imo
,
194 error
= ip_output_list(m0
, 0, opt
, ro
, flags
, imo
, ifp
);
207 * ipsec4_getpolicybyaddr:??? [IPSEC 4th argument, contents modified]
208 * ipsec4_getpolicybysock:??? [IPSEC 4th argument, contents modified]
209 * key_spdacquire:??? [IPSEC]
210 * ipsec4_output:??? [IPSEC]
211 * <fr_checkp>:??? [firewall]
212 * ip_dn_io_ptr:??? [dummynet]
213 * dlil_output:??? [DLIL]
214 * dlil_output_list:??? [DLIL]
216 * Notes: The ipsec4_getpolicyby{addr|sock} function error returns are
217 * only used as the error return from this function where one of
218 * these functions fails to return a policy.
227 struct ip_moptions
*imo
,
228 #if CONFIG_FORCE_OUT_IFP
229 struct ifnet
*pdp_ifp
231 __unused
struct ifnet
*unused_ifp
235 struct ip
*ip
, *mhip
;
236 struct ifnet
*ifp
= NULL
;
238 int hlen
= sizeof (struct ip
);
239 int len
= 0, off
, error
= 0;
240 struct sockaddr_in
*dst
= NULL
;
241 struct in_ifaddr
*ia
= NULL
;
242 int isbroadcast
, sw_csum
;
243 struct in_addr pkt_dst
;
245 struct route iproute
;
246 struct socket
*so
= NULL
;
247 struct secpolicy
*sp
= NULL
;
249 #if IPFIREWALL_FORWARD
250 int fwd_rewrite_src
= 0;
253 struct ip_fw_args args
;
256 ipfilter_t inject_filter_ref
= 0;
258 struct route saved_route
;
259 struct mbuf
* packetlist
;
263 KERNEL_DEBUG(DBG_FNC_IP_OUTPUT
| DBG_FUNC_START
, 0,0,0,0,0);
267 args
.next_hop
= NULL
;
270 args
.divert_rule
= 0; /* divert cookie */
272 /* Grab info from mtags prepended to the chain */
274 if ((tag
= m_tag_locate(m0
, KERNEL_MODULE_TAG_ID
, KERNEL_TAG_TYPE_DUMMYNET
, NULL
)) != NULL
) {
275 struct dn_pkt_tag
*dn_tag
;
277 dn_tag
= (struct dn_pkt_tag
*)(tag
+1);
278 args
.rule
= dn_tag
->rule
;
280 saved_route
= dn_tag
->ro
;
284 dst
= dn_tag
->dn_dst
;
286 flags
= dn_tag
->flags
;
288 m_tag_delete(m0
, tag
);
290 #endif /* DUMMYNET */
293 if ((tag
= m_tag_locate(m0
, KERNEL_MODULE_TAG_ID
, KERNEL_TAG_TYPE_DIVERT
, NULL
)) != NULL
) {
294 struct divert_tag
*div_tag
;
296 div_tag
= (struct divert_tag
*)(tag
+1);
297 args
.divert_rule
= div_tag
->cookie
;
299 m_tag_delete(m0
, tag
);
301 #endif /* IPDIVERT */
303 if ((tag
= m_tag_locate(m0
, KERNEL_MODULE_TAG_ID
, KERNEL_TAG_TYPE_IPFORWARD
, NULL
)) != NULL
) {
304 struct ip_fwd_tag
*ipfwd_tag
;
306 ipfwd_tag
= (struct ip_fwd_tag
*)(tag
+1);
307 args
.next_hop
= ipfwd_tag
->next_hop
;
309 m_tag_delete(m0
, tag
);
311 #endif /* IPFIREWALL */
316 if ( !m
|| (m
->m_flags
& M_PKTHDR
) != 0)
317 panic("ip_output no HDR");
319 panic("ip_output no route, proto = %d",
320 mtod(m
, struct ip
*)->ip_p
);
324 if (args
.rule
!= NULL
) { /* dummynet already saw us */
325 ip
= mtod(m
, struct ip
*);
326 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2 ;
327 lck_mtx_lock(rt_mtx
);
328 if (ro
->ro_rt
!= NULL
)
329 ia
= (struct in_ifaddr
*)ro
->ro_rt
->rt_ifa
;
332 lck_mtx_unlock(rt_mtx
);
334 if (ipsec_bypass
== 0 && (flags
& IP_NOIPSEC
) == 0) {
335 so
= ipsec_getsocket(m
);
336 (void)ipsec_setsocket(m
, NULL
);
341 #endif /* IPFIREWALL */
344 if (ipsec_bypass
== 0 && (flags
& IP_NOIPSEC
) == 0) {
345 so
= ipsec_getsocket(m
);
346 (void)ipsec_setsocket(m
, NULL
);
351 * No need to proccess packet twice if we've
354 inject_filter_ref
= ipf_get_inject_filter(m
);
357 m
= ip_insertoptions(m
, opt
, &len
);
360 ip
= mtod(m
, struct ip
*);
362 pkt_dst
= args
.next_hop
? args
.next_hop
->sin_addr
: ip
->ip_dst
;
364 pkt_dst
= ip
->ip_dst
;
370 if ((flags
& (IP_FORWARDING
|IP_RAWOUTPUT
)) == 0) {
371 ip
->ip_vhl
= IP_MAKE_VHL(IPVERSION
, hlen
>> 2);
374 ip
->ip_id
= ip_randomid();
376 ip
->ip_id
= htons(ip_id
++);
378 OSAddAtomic(1, (SInt32
*)&ipstat
.ips_localout
);
380 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
384 /* For debugging, we let the stack forge congestion */
386 ((ip
->ip_tos
& IPTOS_ECN_MASK
) == IPTOS_ECN_ECT1
||
387 (ip
->ip_tos
& IPTOS_ECN_MASK
) == IPTOS_ECN_ECT0
)) {
388 ip
->ip_tos
= (ip
->ip_tos
& ~IPTOS_ECN_MASK
) | IPTOS_ECN_CE
;
393 KERNEL_DEBUG(DBG_LAYER_BEG
, ip
->ip_dst
.s_addr
,
394 ip
->ip_src
.s_addr
, ip
->ip_p
, ip
->ip_off
, ip
->ip_len
);
396 dst
= (struct sockaddr_in
*)&ro
->ro_dst
;
399 * If there is a cached route,
400 * check that it is to the same destination
401 * and is still up. If not, free it and try again.
402 * The address family should also be checked in case of sharing the
406 lck_mtx_lock(rt_mtx
);
407 if (ro
->ro_rt
!= NULL
) {
408 if (ro
->ro_rt
->generation_id
!= route_generation
&&
409 ((flags
& (IP_ROUTETOIF
| IP_FORWARDING
)) == 0) &&
410 (ip
->ip_src
.s_addr
!= INADDR_ANY
) &&
411 (ifa_foraddr(ip
->ip_src
.s_addr
) == 0)) {
412 error
= EADDRNOTAVAIL
;
413 lck_mtx_unlock(rt_mtx
);
416 if ((ro
->ro_rt
->rt_flags
& RTF_UP
) == 0 ||
417 dst
->sin_family
!= AF_INET
||
418 dst
->sin_addr
.s_addr
!= pkt_dst
.s_addr
) {
419 rtfree_locked(ro
->ro_rt
);
422 if (ro
->ro_rt
&& ro
->ro_rt
->generation_id
!= route_generation
)
423 ro
->ro_rt
->generation_id
= route_generation
;
425 if (ro
->ro_rt
== NULL
) {
426 bzero(dst
, sizeof(*dst
));
427 dst
->sin_family
= AF_INET
;
428 dst
->sin_len
= sizeof(*dst
);
429 dst
->sin_addr
= pkt_dst
;
432 * If routing to interface only,
433 * short circuit routing lookup.
435 #define ifatoia(ifa) ((struct in_ifaddr *)(ifa))
436 #define sintosa(sin) ((struct sockaddr *)(sin))
437 if (flags
& IP_ROUTETOIF
) {
439 ifafree(&ia
->ia_ifa
);
440 if ((ia
= ifatoia(ifa_ifwithdstaddr(sintosa(dst
)))) == 0) {
441 if ((ia
= ifatoia(ifa_ifwithnet(sintosa(dst
)))) == 0) {
442 OSAddAtomic(1, (SInt32
*)&ipstat
.ips_noroute
);
444 lck_mtx_unlock(rt_mtx
);
450 isbroadcast
= in_broadcast(dst
->sin_addr
, ifp
);
453 #if CONFIG_FORCE_OUT_IFP
454 /* Check if this packet should be forced out a specific interface */
455 if (ro
->ro_rt
== 0 && pdp_ifp
!= NULL
) {
456 pdp_context_route_locked(pdp_ifp
, ro
);
458 if (ro
->ro_rt
== NULL
) {
459 OSAddAtomic(1, (UInt32
*)&ipstat
.ips_noroute
);
460 error
= EHOSTUNREACH
;
461 lck_mtx_unlock(rt_mtx
);
468 * If this is the case, we probably don't want to allocate
469 * a protocol-cloned route since we didn't get one from the
470 * ULP. This lets TCP do its thing, while not burdening
471 * forwarding or ICMP with the overhead of cloning a route.
472 * Of course, we still want to do any cloning requested by
473 * the link layer, as this is probably required in all cases
474 * for correct operation (as it is for ARP).
477 if (ro
->ro_rt
== 0) {
478 unsigned long ign
= RTF_PRCLONING
;
480 * We make an exception here: if the destination
481 * address is INADDR_BROADCAST, allocate a protocol-
482 * cloned host route so that we end up with a route
483 * marked with the RTF_BROADCAST flag. Otherwise,
484 * we would end up referring to the default route,
485 * instead of creating a cloned host route entry.
486 * That would introduce inconsistencies between ULPs
487 * that allocate a route and those that don't. The
488 * RTF_BROADCAST route is important since we'd want
489 * to send out undirected IP broadcast packets using
490 * link-level broadcast address.
492 * This exception will no longer be necessary when
493 * the RTF_PRCLONING scheme is no longer present.
495 if (dst
->sin_addr
.s_addr
== INADDR_BROADCAST
)
496 ign
&= ~RTF_PRCLONING
;
498 rtalloc_ign_locked(ro
, ign
);
500 if (ro
->ro_rt
== 0) {
501 OSAddAtomic(1, (SInt32
*)&ipstat
.ips_noroute
);
502 error
= EHOSTUNREACH
;
503 lck_mtx_unlock(rt_mtx
);
508 ifafree(&ia
->ia_ifa
);
509 ia
= ifatoia(ro
->ro_rt
->rt_ifa
);
512 ifp
= ro
->ro_rt
->rt_ifp
;
514 if (ro
->ro_rt
->rt_flags
& RTF_GATEWAY
)
515 dst
= (struct sockaddr_in
*)ro
->ro_rt
->rt_gateway
;
516 if (ro
->ro_rt
->rt_flags
& RTF_HOST
)
517 isbroadcast
= (ro
->ro_rt
->rt_flags
& RTF_BROADCAST
);
519 isbroadcast
= in_broadcast(dst
->sin_addr
, ifp
);
521 lck_mtx_unlock(rt_mtx
);
522 if (IN_MULTICAST(ntohl(pkt_dst
.s_addr
))) {
523 struct in_multi
*inm
;
525 m
->m_flags
|= M_MCAST
;
527 * IP destination address is multicast. Make sure "dst"
528 * still points to the address in "ro". (It may have been
529 * changed to point to a gateway address, above.)
531 dst
= (struct sockaddr_in
*)&ro
->ro_dst
;
533 * See if the caller provided any multicast options
536 if ((flags
& IP_RAWOUTPUT
) == 0) ip
->ip_ttl
= imo
->imo_multicast_ttl
;
537 if (imo
->imo_multicast_ifp
!= NULL
) {
538 ifp
= imo
->imo_multicast_ifp
;
541 if (imo
->imo_multicast_vif
!= -1 &&
542 ((flags
& IP_RAWOUTPUT
) == 0 || ip
->ip_src
.s_addr
== INADDR_ANY
))
544 ip_mcast_src(imo
->imo_multicast_vif
);
545 #endif /* MROUTING */
547 if ((flags
& IP_RAWOUTPUT
) == 0) ip
->ip_ttl
= IP_DEFAULT_MULTICAST_TTL
;
549 * Confirm that the outgoing interface supports multicast.
551 if ((imo
== NULL
) || (imo
->imo_multicast_vif
== -1)) {
552 if ((ifp
->if_flags
& IFF_MULTICAST
) == 0) {
553 OSAddAtomic(1, (SInt32
*)&ipstat
.ips_noroute
);
559 * If source address not specified yet, use address
560 * of outgoing interface.
562 if (ip
->ip_src
.s_addr
== INADDR_ANY
) {
563 register struct in_ifaddr
*ia1
;
564 lck_mtx_lock(rt_mtx
);
565 TAILQ_FOREACH(ia1
, &in_ifaddrhead
, ia_link
)
566 if (ia1
->ia_ifp
== ifp
) {
567 ip
->ip_src
= IA_SIN(ia1
)->sin_addr
;
571 lck_mtx_unlock(rt_mtx
);
572 if (ip
->ip_src
.s_addr
== INADDR_ANY
) {
578 ifnet_lock_shared(ifp
);
579 IN_LOOKUP_MULTI(pkt_dst
, ifp
, inm
);
580 ifnet_lock_done(ifp
);
582 (imo
== NULL
|| imo
->imo_multicast_loop
)) {
584 * If we belong to the destination multicast group
585 * on the outgoing interface, and the caller did not
586 * forbid loopback, loop back a copy.
588 if (!TAILQ_EMPTY(&ipv4_filters
)) {
589 struct ipfilter
*filter
;
590 int seen
= (inject_filter_ref
== 0);
591 struct ipf_pktopts
*ippo
= 0, ipf_pktopts
;
595 ipf_pktopts
.ippo_mcast_ifnet
= imo
->imo_multicast_ifp
;
596 ipf_pktopts
.ippo_mcast_ttl
= imo
->imo_multicast_ttl
;
597 ipf_pktopts
.ippo_mcast_loop
= imo
->imo_multicast_loop
;
602 /* 4135317 - always pass network byte order to filter */
606 TAILQ_FOREACH(filter
, &ipv4_filters
, ipf_link
) {
608 if ((struct ipfilter
*)inject_filter_ref
== filter
)
610 } else if (filter
->ipf_filter
.ipf_output
) {
612 result
= filter
->ipf_filter
.ipf_output(filter
->ipf_filter
.cookie
, (mbuf_t
*)&m
, ippo
);
613 if (result
== EJUSTRETURN
) {
624 /* set back to host byte order */
625 ip
= mtod(m
, struct ip
*);
632 ip_mloopback(ifp
, m
, dst
, hlen
);
637 * If we are acting as a multicast router, perform
638 * multicast forwarding as if the packet had just
639 * arrived on the interface to which we are about
640 * to send. The multicast forwarding function
641 * recursively calls this function, using the
642 * IP_FORWARDING flag to prevent infinite recursion.
644 * Multicasts that are looped back by ip_mloopback(),
645 * above, will be forwarded by the ip_input() routine,
648 if (ip_mrouter
&& (flags
& IP_FORWARDING
) == 0) {
650 * Check if rsvp daemon is running. If not, don't
651 * set ip_moptions. This ensures that the packet
652 * is multicast and not just sent down one link
653 * as prescribed by rsvpd.
657 if (ip_mforward(ip
, ifp
, m
, imo
) != 0) {
663 #endif /* MROUTING */
666 * Multicasts with a time-to-live of zero may be looped-
667 * back, above, but must not be transmitted on a network.
668 * Also, multicasts addressed to the loopback interface
669 * are not sent -- the above call to ip_mloopback() will
670 * loop back a copy if this host actually belongs to the
671 * destination group on the loopback interface.
673 if (ip
->ip_ttl
== 0 || ifp
->if_flags
& IFF_LOOPBACK
) {
682 * If source address not specified yet, use address
683 * of outgoing interface.
685 if (ip
->ip_src
.s_addr
== INADDR_ANY
) {
686 ip
->ip_src
= IA_SIN(ia
)->sin_addr
;
687 #if IPFIREWALL_FORWARD
688 /* Keep note that we did this - if the firewall changes
689 * the next-hop, our interface may change, changing the
690 * default source IP. It's a shame so much effort happens
694 #endif /* IPFIREWALL_FORWARD */
699 * Look for broadcast address and
700 * and verify user is allowed to send
704 if ((ifp
->if_flags
& IFF_BROADCAST
) == 0) {
705 error
= EADDRNOTAVAIL
;
708 if ((flags
& IP_ALLOWBROADCAST
) == 0) {
712 /* don't allow broadcast messages to be fragmented */
713 if ((u_short
)ip
->ip_len
> ifp
->if_mtu
) {
717 m
->m_flags
|= M_BCAST
;
719 m
->m_flags
&= ~M_BCAST
;
724 * Force IP TTL to 255 following draft-ietf-zeroconf-ipv4-linklocal.txt
726 if (IN_LINKLOCAL(ntohl(ip
->ip_src
.s_addr
)) || IN_LINKLOCAL(ntohl(ip
->ip_dst
.s_addr
))) {
727 ip_linklocal_stat
.iplls_out_total
++;
728 if (ip
->ip_ttl
!= MAXTTL
) {
729 ip_linklocal_stat
.iplls_out_badttl
++;
734 if (!didfilter
&& !TAILQ_EMPTY(&ipv4_filters
)) {
735 struct ipfilter
*filter
;
736 int seen
= (inject_filter_ref
== 0);
740 /* 4135317 - always pass network byte order to filter */
744 TAILQ_FOREACH(filter
, &ipv4_filters
, ipf_link
) {
746 if ((struct ipfilter
*)inject_filter_ref
== filter
)
748 } else if (filter
->ipf_filter
.ipf_output
) {
750 result
= filter
->ipf_filter
.ipf_output(filter
->ipf_filter
.cookie
, (mbuf_t
*)&m
, 0);
751 if (result
== EJUSTRETURN
) {
762 /* set back to host byte order */
763 ip
= mtod(m
, struct ip
*);
771 /* temporary for testing only: bypass ipsec alltogether */
773 if (ipsec_bypass
!= 0 || (flags
& IP_NOIPSEC
) != 0)
776 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT
| DBG_FUNC_START
, 0,0,0,0,0);
779 /* get SP for this packet */
781 sp
= ipsec4_getpolicybyaddr(m
, IPSEC_DIR_OUTBOUND
, flags
, &error
);
783 sp
= ipsec4_getpolicybysock(m
, IPSEC_DIR_OUTBOUND
, so
, &error
);
786 IPSEC_STAT_INCREMENT(ipsecstat
.out_inval
);
787 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT
| DBG_FUNC_END
, 0,0,0,0,0);
794 switch (sp
->policy
) {
795 case IPSEC_POLICY_DISCARD
:
796 case IPSEC_POLICY_GENERATE
:
798 * This packet is just discarded.
800 IPSEC_STAT_INCREMENT(ipsecstat
.out_polvio
);
801 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT
| DBG_FUNC_END
, 1,0,0,0,0);
804 case IPSEC_POLICY_BYPASS
:
805 case IPSEC_POLICY_NONE
:
806 /* no need to do IPsec. */
807 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT
| DBG_FUNC_END
, 2,0,0,0,0);
810 case IPSEC_POLICY_IPSEC
:
811 if (sp
->req
== NULL
) {
812 /* acquire a policy */
813 error
= key_spdacquire(sp
);
814 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT
| DBG_FUNC_END
, 3,0,0,0,0);
819 case IPSEC_POLICY_ENTRUST
:
821 printf("ip_output: Invalid policy found. %d\n", sp
->policy
);
824 struct ipsec_output_state state
;
825 bzero(&state
, sizeof(state
));
827 if (flags
& IP_ROUTETOIF
) {
829 bzero(&iproute
, sizeof(iproute
));
832 state
.dst
= (struct sockaddr
*)dst
;
838 * delayed checksums are not currently compatible with IPsec
840 if (m
->m_pkthdr
.csum_flags
& CSUM_DELAY_DATA
) {
842 m
->m_pkthdr
.csum_flags
&= ~CSUM_DELAY_DATA
;
848 error
= ipsec4_output(&state
, sp
, flags
);
852 if (flags
& IP_ROUTETOIF
) {
854 * if we have tunnel mode SA, we may need to ignore
857 if (state
.ro
!= &iproute
|| state
.ro
->ro_rt
!= NULL
) {
858 flags
&= ~IP_ROUTETOIF
;
864 dst
= (struct sockaddr_in
*)state
.dst
;
866 /* mbuf is already reclaimed in ipsec4_output. */
876 printf("ip4_output (ipsec): error code %d\n", error
);
879 /* don't show these error codes to the user */
883 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT
| DBG_FUNC_END
, 4,0,0,0,0);
888 /* be sure to update variables that are affected by ipsec4_output() */
889 ip
= mtod(m
, struct ip
*);
892 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
894 hlen
= ip
->ip_hl
<< 2;
896 /* Check that there wasn't a route change and src is still valid */
898 lck_mtx_lock(rt_mtx
);
899 if (ro
->ro_rt
&& ro
->ro_rt
->generation_id
!= route_generation
) {
900 if (ifa_foraddr(ip
->ip_src
.s_addr
) == 0 && ((flags
& (IP_ROUTETOIF
| IP_FORWARDING
)) == 0)) {
901 error
= EADDRNOTAVAIL
;
902 lck_mtx_unlock(rt_mtx
);
903 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT
| DBG_FUNC_END
, 5,0,0,0,0);
906 rtfree_locked(ro
->ro_rt
);
910 if (ro
->ro_rt
== NULL
) {
911 if ((flags
& IP_ROUTETOIF
) == 0) {
913 "can't update route after IPsec processing\n");
914 error
= EHOSTUNREACH
; /*XXX*/
915 lck_mtx_unlock(rt_mtx
);
916 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT
| DBG_FUNC_END
, 6,0,0,0,0);
921 ifafree(&ia
->ia_ifa
);
922 ia
= ifatoia(ro
->ro_rt
->rt_ifa
);
925 ifp
= ro
->ro_rt
->rt_ifp
;
927 lck_mtx_unlock(rt_mtx
);
929 /* make it flipped, again. */
932 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT
| DBG_FUNC_END
, 7,0xff,0xff,0xff,0xff);
934 /* Pass to filters again */
935 if (!TAILQ_EMPTY(&ipv4_filters
)) {
936 struct ipfilter
*filter
;
940 /* 4135317 - always pass network byte order to filter */
944 TAILQ_FOREACH(filter
, &ipv4_filters
, ipf_link
) {
945 if (filter
->ipf_filter
.ipf_output
) {
947 result
= filter
->ipf_filter
.ipf_output(filter
->ipf_filter
.cookie
, (mbuf_t
*)&m
, 0);
948 if (result
== EJUSTRETURN
) {
959 /* set back to host byte order */
960 ip
= mtod(m
, struct ip
*);
972 * - Xlate: translate packet's addr/port (NAT).
973 * - Firewall: deny/allow/etc.
974 * - Wrap: fake packet's addr/port <unimpl.>
975 * - Encapsulate: put it in another IP and send out. <unimp.>
980 if ((error
= (*fr_checkp
)(ip
, hlen
, ifp
, 1, &m1
)) || !m1
) {
983 ip
= mtod(m0
= m
= m1
, struct ip
*);
987 * Check with the firewall...
988 * but not if we are already being fwd'd from a firewall.
990 if (fw_enable
&& IPFW_LOADED
&& !args
.next_hop
) {
991 struct sockaddr_in
*old
= dst
;
996 off
= ip_fw_chk_ptr(&args
);
1001 * On return we must do the following:
1002 * IP_FW_PORT_DENY_FLAG -> drop the pkt (XXX new)
1003 * 1<=off<= 0xffff -> DIVERT
1004 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe
1005 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet
1006 * dst != old -> IPFIREWALL_FORWARD
1007 * off==0, dst==old -> accept
1008 * If some of the above modules is not compiled in, then
1009 * we should't have to check the corresponding condition
1010 * (because the ipfw control socket should not accept
1011 * unsupported rules), but better play safe and drop
1012 * packets in case of doubt.
1015 if ( (off
& IP_FW_PORT_DENY_FLAG
) || m
== NULL
) {
1021 ip
= mtod(m
, struct ip
*);
1023 if (off
== 0 && dst
== old
) {/* common case */
1027 if (DUMMYNET_LOADED
&& (off
& IP_FW_PORT_DYNT_FLAG
) != 0) {
1029 * pass the pkt to dummynet. Need to include
1030 * pipe number, m, ifp, ro, dst because these are
1031 * not recomputed in the next pass.
1032 * All other parameters have been already used and
1033 * so they are not needed anymore.
1034 * XXX note: if the ifp or ro entry are deleted
1035 * while a pkt is in dummynet, we are in trouble!
1041 error
= ip_dn_io_ptr(m
, off
& 0xffff, DN_TO_IP_OUT
,
1045 #endif /* DUMMYNET */
1047 if (off
!= 0 && (off
& IP_FW_PORT_DYNT_FLAG
) == 0) {
1048 struct mbuf
*clone
= NULL
;
1050 /* Clone packet if we're doing a 'tee' */
1051 if ((off
& IP_FW_PORT_TEE_FLAG
) != 0)
1052 clone
= m_dup(m
, M_DONTWAIT
);
1055 * delayed checksums are not currently compatible
1056 * with divert sockets.
1058 if (m
->m_pkthdr
.csum_flags
& CSUM_DELAY_DATA
) {
1059 in_delayed_cksum(m
);
1060 m
->m_pkthdr
.csum_flags
&= ~CSUM_DELAY_DATA
;
1063 /* Restore packet header fields to original values */
1067 /* Deliver packet to divert input routine */
1068 divert_packet(m
, 0, off
& 0xffff, args
.divert_rule
);
1070 /* If 'tee', continue with original packet */
1071 if (clone
!= NULL
) {
1073 ip
= mtod(m
, struct ip
*);
1080 #if IPFIREWALL_FORWARD
1081 /* Here we check dst to make sure it's directly reachable on the
1082 * interface we previously thought it was.
1083 * If it isn't (which may be likely in some situations) we have
1084 * to re-route it (ie, find a route for the next-hop and the
1085 * associated interface) and set them here. This is nested
1086 * forwarding which in most cases is undesirable, except where
1087 * such control is nigh impossible. So we do it here.
1090 if (off
== 0 && old
!= dst
) {
1091 struct in_ifaddr
*ia_fw
;
1093 /* It's changed... */
1094 /* There must be a better way to do this next line... */
1095 static struct route sro_fwd
, *ro_fwd
= &sro_fwd
;
1096 #if IPFIREWALL_FORWARD_DEBUG
1097 printf("IPFIREWALL_FORWARD: New dst ip: ");
1098 print_ip(dst
->sin_addr
);
1102 * We need to figure out if we have been forwarded
1103 * to a local socket. If so then we should somehow
1104 * "loop back" to ip_input, and get directed to the
1105 * PCB as if we had received this packet. This is
1106 * because it may be dificult to identify the packets
1107 * you want to forward until they are being output
1108 * and have selected an interface. (e.g. locally
1109 * initiated packets) If we used the loopback inteface,
1110 * we would not be able to control what happens
1111 * as the packet runs through ip_input() as
1112 * it is done through a ISR.
1114 TAILQ_FOREACH(ia_fw
, &in_ifaddrhead
, ia_link
) {
1116 * If the addr to forward to is one
1117 * of ours, we pretend to
1118 * be the destination for this packet.
1120 if (IA_SIN(ia_fw
)->sin_addr
.s_addr
==
1121 dst
->sin_addr
.s_addr
)
1125 /* tell ip_input "dont filter" */
1126 struct m_tag
*fwd_tag
;
1127 struct ip_fwd_tag
*ipfwd_tag
;
1129 fwd_tag
= m_tag_alloc(KERNEL_MODULE_TAG_ID
, KERNEL_TAG_TYPE_IPFORWARD
,
1130 sizeof(struct sockaddr_in
), M_NOWAIT
);
1131 if (fwd_tag
== NULL
) {
1136 ipfwd_tag
= (struct ip_fwd_tag
*)(fwd_tag
+1);
1137 ipfwd_tag
->next_hop
= args
.next_hop
;
1139 m_tag_prepend(m
, fwd_tag
);
1141 if (m
->m_pkthdr
.rcvif
== NULL
)
1142 m
->m_pkthdr
.rcvif
= ifunit("lo0");
1143 if ((~IF_HWASSIST_CSUM_FLAGS(m
->m_pkthdr
.rcvif
->if_hwassist
) &
1144 m
->m_pkthdr
.csum_flags
) == 0) {
1145 if (m
->m_pkthdr
.csum_flags
& CSUM_DELAY_DATA
) {
1146 m
->m_pkthdr
.csum_flags
&= ~CSUM_DELAY_DATA
;
1147 m
->m_pkthdr
.csum_flags
|=
1148 CSUM_DATA_VALID
| CSUM_PSEUDO_HDR
;
1149 m
->m_pkthdr
.csum_data
= 0xffff;
1151 m
->m_pkthdr
.csum_flags
|=
1152 CSUM_IP_CHECKED
| CSUM_IP_VALID
;
1154 else if (m
->m_pkthdr
.csum_flags
& CSUM_DELAY_DATA
) {
1155 in_delayed_cksum(m
);
1156 m
->m_pkthdr
.csum_flags
&= ~CSUM_DELAY_DATA
;
1157 ip
->ip_sum
= in_cksum(m
, hlen
);
1163 /* we need to call dlil_output to run filters
1164 * and resync to avoid recursion loops.
1167 dlil_output(lo_ifp
, PF_INET
, m
, 0, (struct sockaddr
*)dst
, 0);
1170 printf("ip_output: no loopback ifp for forwarding!!!\n");
1174 /* Some of the logic for this was
1175 * nicked from above.
1177 * This rewrites the cached route in a local PCB.
1178 * Is this what we want to do?
1180 bcopy(dst
, &ro_fwd
->ro_dst
, sizeof(*dst
));
1183 lck_mtx_lock(rt_mtx
);
1184 rtalloc_ign_locked(ro_fwd
, RTF_PRCLONING
);
1186 if (ro_fwd
->ro_rt
== 0) {
1187 OSAddAtomic(1, (SInt32
*)&ipstat
.ips_noroute
);
1188 error
= EHOSTUNREACH
;
1189 lck_mtx_unlock(rt_mtx
);
1193 ia_fw
= ifatoia(ro_fwd
->ro_rt
->rt_ifa
);
1194 ifp
= ro_fwd
->ro_rt
->rt_ifp
;
1195 ro_fwd
->ro_rt
->rt_use
++;
1196 if (ro_fwd
->ro_rt
->rt_flags
& RTF_GATEWAY
)
1197 dst
= (struct sockaddr_in
*)ro_fwd
->ro_rt
->rt_gateway
;
1198 if (ro_fwd
->ro_rt
->rt_flags
& RTF_HOST
)
1200 (ro_fwd
->ro_rt
->rt_flags
& RTF_BROADCAST
);
1202 isbroadcast
= in_broadcast(dst
->sin_addr
, ifp
);
1203 rtfree_locked(ro
->ro_rt
);
1204 ro
->ro_rt
= ro_fwd
->ro_rt
;
1205 dst
= (struct sockaddr_in
*)&ro_fwd
->ro_dst
;
1206 lck_mtx_unlock(rt_mtx
);
1209 * If we added a default src ip earlier,
1210 * which would have been gotten from the-then
1211 * interface, do it again, from the new one.
1213 if (fwd_rewrite_src
)
1214 ip
->ip_src
= IA_SIN(ia_fw
)->sin_addr
;
1217 #endif /* IPFIREWALL_FORWARD */
1219 * if we get here, none of the above matches, and
1220 * we have to drop the pkt
1223 error
= EACCES
; /* not sure this is the right error msg */
1226 #endif /* IPFIREWALL */
1230 /* Do not allow loopback address to wind up on a wire */
1231 if ((ifp
->if_flags
& IFF_LOOPBACK
) == 0 &&
1232 ((ntohl(ip
->ip_src
.s_addr
) >> IN_CLASSA_NSHIFT
) == IN_LOOPBACKNET
||
1233 (ntohl(ip
->ip_dst
.s_addr
) >> IN_CLASSA_NSHIFT
) == IN_LOOPBACKNET
)) {
1234 OSAddAtomic(1, (SInt32
*)&ipstat
.ips_badaddr
);
1237 * Do not simply drop the packet just like a firewall -- we want the
1238 * the application to feel the pain.
1239 * Return ENETUNREACH like ip6_output does in some similar cases.
1240 * This can startle the otherwise clueless process that specifies
1241 * loopback as the source address.
1243 error
= ENETUNREACH
;
1247 m
->m_pkthdr
.csum_flags
|= CSUM_IP
;
1248 sw_csum
= m
->m_pkthdr
.csum_flags
1249 & ~IF_HWASSIST_CSUM_FLAGS(ifp
->if_hwassist
);
1251 if ((ifp
->if_hwassist
& CSUM_TCP_SUM16
) != 0) {
1253 * Special case code for GMACE
1254 * frames that can be checksumed by GMACE SUM16 HW:
1255 * frame >64, no fragments, no UDP
1257 if (apple_hwcksum_tx
&& (m
->m_pkthdr
.csum_flags
& CSUM_TCP
)
1258 && (ip
->ip_len
> 50) && (ip
->ip_len
<= ifp
->if_mtu
)) {
1259 /* Apple GMAC HW, expects STUFF_OFFSET << 16 | START_OFFSET */
1260 u_short offset
= (IP_VHL_HL(ip
->ip_vhl
) << 2) +14 ; /* IP+Enet header length */
1261 u_short csumprev
= m
->m_pkthdr
.csum_data
& 0xFFFF;
1262 m
->m_pkthdr
.csum_flags
= CSUM_DATA_VALID
| CSUM_TCP_SUM16
; /* for GMAC */
1263 m
->m_pkthdr
.csum_data
= (csumprev
+ offset
) << 16 ;
1264 m
->m_pkthdr
.csum_data
+= offset
;
1265 sw_csum
= CSUM_DELAY_IP
; /* do IP hdr chksum in software */
1268 /* let the software handle any UDP or TCP checksums */
1269 sw_csum
|= (CSUM_DELAY_DATA
& m
->m_pkthdr
.csum_flags
);
1271 } else if (apple_hwcksum_tx
== 0) {
1272 sw_csum
|= (CSUM_DELAY_DATA
| CSUM_DELAY_IP
) &
1273 m
->m_pkthdr
.csum_flags
;
1276 if (sw_csum
& CSUM_DELAY_DATA
) {
1277 in_delayed_cksum(m
);
1278 sw_csum
&= ~CSUM_DELAY_DATA
;
1279 m
->m_pkthdr
.csum_flags
&= ~CSUM_DELAY_DATA
;
1282 if (apple_hwcksum_tx
!= 0) {
1283 m
->m_pkthdr
.csum_flags
&=
1284 IF_HWASSIST_CSUM_FLAGS(ifp
->if_hwassist
);
1286 m
->m_pkthdr
.csum_flags
= 0;
1290 * If small enough for interface, or the interface will take
1291 * care of the fragmentation for us, can just send directly.
1293 if ((u_short
)ip
->ip_len
<= ifp
->if_mtu
||
1294 ifp
->if_hwassist
& CSUM_FRAGMENT
) {
1295 struct rtentry
*rte
;
1300 if (sw_csum
& CSUM_DELAY_IP
) {
1301 ip
->ip_sum
= in_cksum(m
, hlen
);
1305 /* Record statistics for this interface address. */
1306 if (!(flags
& IP_FORWARDING
) && ia
!= NULL
) {
1307 ia
->ia_ifa
.if_opackets
++;
1308 ia
->ia_ifa
.if_obytes
+= m
->m_pkthdr
.len
;
1313 /* clean ipsec history once it goes out of the node */
1314 if (ipsec_bypass
== 0 && (flags
& IP_NOIPSEC
) == 0)
1317 if (packetchain
== 0) {
1318 lck_mtx_lock(rt_mtx
);
1319 if ((rte
= ro
->ro_rt
) != NULL
)
1321 lck_mtx_unlock(rt_mtx
);
1322 error
= ifnet_output(ifp
, PF_INET
, m
, rte
,
1323 (struct sockaddr
*)dst
);
1328 else { /* packet chaining allows us to reuse the route for all packets */
1331 if (pktcnt
> ip_maxchainsent
)
1332 ip_maxchainsent
= pktcnt
;
1333 lck_mtx_lock(rt_mtx
);
1334 if ((rte
= ro
->ro_rt
) != NULL
)
1336 lck_mtx_unlock(rt_mtx
);
1338 error
= ifnet_output(ifp
, PF_INET
, packetlist
,
1339 rte
, (struct sockaddr
*)dst
);
1352 * Too large for interface; fragment if possible.
1353 * Must be able to put at least 8 bytes per fragment.
1355 if (ip
->ip_off
& IP_DF
) {
1358 * This case can happen if the user changed the MTU
1359 * of an interface after enabling IP on it. Because
1360 * most netifs don't keep track of routes pointing to
1361 * them, there is no way for one to update all its
1362 * routes when the MTU is changed.
1365 lck_mtx_lock(rt_mtx
);
1366 if (ro
->ro_rt
&& (ro
->ro_rt
->rt_flags
& (RTF_UP
| RTF_HOST
))
1367 && !(ro
->ro_rt
->rt_rmx
.rmx_locks
& RTV_MTU
)
1368 && (ro
->ro_rt
->rt_rmx
.rmx_mtu
> ifp
->if_mtu
)) {
1369 ro
->ro_rt
->rt_rmx
.rmx_mtu
= ifp
->if_mtu
;
1371 lck_mtx_unlock(rt_mtx
);
1372 OSAddAtomic(1, (SInt32
*)&ipstat
.ips_cantfrag
);
1375 len
= (ifp
->if_mtu
- hlen
) &~ 7;
1382 * if the interface will not calculate checksums on
1383 * fragmented packets, then do it here.
1385 if (m
->m_pkthdr
.csum_flags
& CSUM_DELAY_DATA
&&
1386 (ifp
->if_hwassist
& CSUM_IP_FRAGS
) == 0) {
1387 in_delayed_cksum(m
);
1388 m
->m_pkthdr
.csum_flags
&= ~CSUM_DELAY_DATA
;
1393 int mhlen
, firstlen
= len
;
1394 struct mbuf
**mnext
= &m
->m_nextpkt
;
1398 * Loop through length of segment after first fragment,
1399 * make new header and copy data of each part and link onto chain.
1402 mhlen
= sizeof (struct ip
);
1403 for (off
= hlen
+ len
; off
< (u_short
)ip
->ip_len
; off
+= len
) {
1404 MGETHDR(m
, M_DONTWAIT
, MT_HEADER
); /* MAC-OK */
1407 OSAddAtomic(1, (SInt32
*)&ipstat
.ips_odropped
);
1410 m
->m_flags
|= (m0
->m_flags
& M_MCAST
) | M_FRAG
;
1411 m
->m_data
+= max_linkhdr
;
1412 mhip
= mtod(m
, struct ip
*);
1414 if (hlen
> sizeof (struct ip
)) {
1415 mhlen
= ip_optcopy(ip
, mhip
) + sizeof (struct ip
);
1416 mhip
->ip_vhl
= IP_MAKE_VHL(IPVERSION
, mhlen
>> 2);
1419 mhip
->ip_off
= ((off
- hlen
) >> 3) + (ip
->ip_off
& ~IP_MF
);
1420 if (ip
->ip_off
& IP_MF
)
1421 mhip
->ip_off
|= IP_MF
;
1422 if (off
+ len
>= (u_short
)ip
->ip_len
)
1423 len
= (u_short
)ip
->ip_len
- off
;
1425 mhip
->ip_off
|= IP_MF
;
1426 mhip
->ip_len
= htons((u_short
)(len
+ mhlen
));
1427 m
->m_next
= m_copy(m0
, off
, len
);
1428 if (m
->m_next
== 0) {
1430 error
= ENOBUFS
; /* ??? */
1431 OSAddAtomic(1, (SInt32
*)&ipstat
.ips_odropped
);
1434 m
->m_pkthdr
.len
= mhlen
+ len
;
1435 m
->m_pkthdr
.rcvif
= 0;
1436 m
->m_pkthdr
.csum_flags
= m0
->m_pkthdr
.csum_flags
;
1437 m
->m_pkthdr
.socket_id
= m0
->m_pkthdr
.socket_id
;
1439 mac_netinet_fragment(m0
, m
);
1441 HTONS(mhip
->ip_off
);
1443 if (sw_csum
& CSUM_DELAY_IP
) {
1444 mhip
->ip_sum
= in_cksum(m
, mhlen
);
1447 mnext
= &m
->m_nextpkt
;
1450 OSAddAtomic(nfrags
, (SInt32
*)&ipstat
.ips_ofragments
);
1452 /* set first/last markers for fragment chain */
1453 m
->m_flags
|= M_LASTFRAG
;
1454 m0
->m_flags
|= M_FIRSTFRAG
| M_FRAG
;
1455 m0
->m_pkthdr
.csum_data
= nfrags
;
1458 * Update first fragment by trimming what's been copied out
1459 * and updating header, then send each fragment (in order).
1462 m_adj(m
, hlen
+ firstlen
- (u_short
)ip
->ip_len
);
1463 m
->m_pkthdr
.len
= hlen
+ firstlen
;
1464 ip
->ip_len
= htons((u_short
)m
->m_pkthdr
.len
);
1465 ip
->ip_off
|= IP_MF
;
1468 if (sw_csum
& CSUM_DELAY_IP
) {
1469 ip
->ip_sum
= in_cksum(m
, hlen
);
1473 KERNEL_DEBUG(DBG_LAYER_END
, ip
->ip_dst
.s_addr
,
1474 ip
->ip_src
.s_addr
, ip
->ip_p
, ip
->ip_off
, ip
->ip_len
);
1476 for (m
= m0
; m
; m
= m0
) {
1480 /* clean ipsec history once it goes out of the node */
1481 if (ipsec_bypass
== 0 && (flags
& IP_NOIPSEC
) == 0)
1485 struct rtentry
*rte
;
1487 /* Record statistics for this interface address. */
1489 ia
->ia_ifa
.if_opackets
++;
1490 ia
->ia_ifa
.if_obytes
+= m
->m_pkthdr
.len
;
1493 if ((packetchain
!= 0) && (pktcnt
> 0))
1494 panic("ip_output: mix of packet in packetlist is wrong=%p", packetlist
);
1495 lck_mtx_lock(rt_mtx
);
1496 if ((rte
= ro
->ro_rt
) != NULL
)
1498 lck_mtx_unlock(rt_mtx
);
1499 error
= ifnet_output(ifp
, PF_INET
, m
, rte
,
1500 (struct sockaddr
*)dst
);
1508 OSAddAtomic(1, (SInt32
*)&ipstat
.ips_fragmented
);
1512 ifafree(&ia
->ia_ifa
);
1516 if (ipsec_bypass
== 0 && (flags
& IP_NOIPSEC
) == 0) {
1517 if (ro
== &iproute
&& ro
->ro_rt
) {
1522 KEYDEBUG(KEYDEBUG_IPSEC_STAMP
,
1523 printf("DP ip_output call free SP:%x\n", sp
));
1524 key_freesp(sp
, KEY_SADB_UNLOCKED
);
1529 KERNEL_DEBUG(DBG_FNC_IP_OUTPUT
| DBG_FUNC_END
, error
,0,0,0,0);
1537 ip_out_cksum_stats(int proto
, u_int32_t len
)
1541 tcp_out_cksum_stats(len
);
1544 udp_out_cksum_stats(len
);
1547 /* keep only TCP or UDP stats for now */
1553 in_delayed_cksum_offset(struct mbuf
*m0
, int ip_offset
)
1556 unsigned char buf
[sizeof(struct ip
)];
1557 u_short csum
, offset
, ip_len
;
1558 struct mbuf
*m
= m0
;
1560 while (ip_offset
>= m
->m_len
) {
1561 ip_offset
-= m
->m_len
;
1564 printf("in_delayed_cksum_withoffset failed - ip_offset wasn't in the packet\n");
1569 /* Sometimes the IP header is not contiguous, yes this can happen! */
1570 if (ip_offset
+ sizeof(struct ip
) > m
->m_len
) {
1572 printf("delayed m_pullup, m->len: %ld off: %d\n",
1573 m
->m_len
, ip_offset
);
1575 m_copydata(m
, ip_offset
, sizeof(struct ip
), (caddr_t
) buf
);
1577 ip
= (struct ip
*)buf
;
1579 ip
= (struct ip
*)(m
->m_data
+ ip_offset
);
1584 m
->m_len
-= ip_offset
;
1585 m
->m_data
+= ip_offset
;
1588 offset
= IP_VHL_HL(ip
->ip_vhl
) << 2 ;
1591 * We could be in the context of an IP or interface filter; in the
1592 * former case, ip_len would be in host (correct) order while for
1593 * the latter it would be in network order. Because of this, we
1594 * attempt to interpret the length field by comparing it against
1595 * the actual packet length. If the comparison fails, byte swap
1596 * the length and check again. If it still fails, then the packet
1597 * is bogus and we give up.
1599 ip_len
= ip
->ip_len
;
1600 if (ip_len
!= (m0
->m_pkthdr
.len
- ip_offset
)) {
1601 ip_len
= SWAP16(ip_len
);
1602 if (ip_len
!= (m0
->m_pkthdr
.len
- ip_offset
)) {
1603 printf("in_delayed_cksum_offset: ip_len %d (%d) "
1604 "doesn't match actual length %d\n", ip
->ip_len
,
1605 ip_len
, (m0
->m_pkthdr
.len
- ip_offset
));
1610 csum
= in_cksum_skip(m
, ip_len
, offset
);
1613 ip_out_cksum_stats(ip
->ip_p
, ip_len
- offset
);
1615 if (m0
->m_pkthdr
.csum_flags
& CSUM_UDP
&& csum
== 0)
1617 offset
+= m0
->m_pkthdr
.csum_data
& 0xFFFF; /* checksum offset */
1621 if (M_LEADINGSPACE(m
) < ip_offset
)
1622 panic("in_delayed_cksum_offset - chain modified!\n");
1623 m
->m_len
+= ip_offset
;
1624 m
->m_data
-= ip_offset
;
1627 if (offset
> ip_len
) /* bogus offset */
1630 /* Insert the checksum in the existing chain */
1631 if (offset
+ ip_offset
+ sizeof(u_short
) > m
->m_len
) {
1635 printf("delayed m_copyback, m->len: %ld off: %d p: %d\n",
1636 m
->m_len
, offset
+ ip_offset
, ip
->ip_p
);
1638 *(u_short
*)tmp
= csum
;
1639 m_copyback(m
, offset
+ ip_offset
, 2, tmp
);
1641 *(u_short
*)(m
->m_data
+ offset
+ ip_offset
) = csum
;
1645 in_delayed_cksum(struct mbuf
*m
)
1647 in_delayed_cksum_offset(m
, 0);
1651 in_cksum_offset(struct mbuf
* m
, size_t ip_offset
)
1653 struct ip
* ip
= NULL
;
1655 unsigned char buf
[sizeof(struct ip
)];
1658 while (ip_offset
>= m
->m_len
) {
1659 ip_offset
-= m
->m_len
;
1662 printf("in_cksum_offset failed - ip_offset wasn't in the packet\n");
1667 /* Sometimes the IP header is not contiguous, yes this can happen! */
1668 if (ip_offset
+ sizeof(struct ip
) > m
->m_len
) {
1671 printf("in_cksum_offset - delayed m_pullup, m->len: %ld off: %lu\n",
1672 m
->m_len
, ip_offset
);
1674 m_copydata(m
, ip_offset
, sizeof(struct ip
), (caddr_t
) buf
);
1676 ip
= (struct ip
*)buf
;
1678 m_copyback(m
, ip_offset
+ offsetof(struct ip
, ip_sum
), 2, (caddr_t
)&ip
->ip_sum
);
1680 ip
= (struct ip
*)(m
->m_data
+ ip_offset
);
1686 m
->m_len
-= ip_offset
;
1687 m
->m_data
+= ip_offset
;
1691 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
1693 hlen
= ip
->ip_hl
<< 2;
1696 * We could be in the context of an IP or interface filter; in the
1697 * former case, ip_len would be in host order while for the latter
1698 * it would be in network (correct) order. Because of this, we
1699 * attempt to interpret the length field by comparing it against
1700 * the actual packet length. If the comparison fails, byte swap
1701 * the length and check again. If it still fails, then the packet
1702 * is bogus and we give up.
1704 if (ntohs(ip
->ip_len
) != (m
->m_pkthdr
.len
- ip_offset
)) {
1705 ip
->ip_len
= SWAP16(ip
->ip_len
);
1707 if (ntohs(ip
->ip_len
) != (m
->m_pkthdr
.len
- ip_offset
)) {
1708 ip
->ip_len
= SWAP16(ip
->ip_len
);
1709 printf("in_cksum_offset: ip_len %d (%d) "
1710 "doesn't match actual length %lu\n",
1711 ip
->ip_len
, SWAP16(ip
->ip_len
),
1712 (m
->m_pkthdr
.len
- ip_offset
));
1718 ip
->ip_sum
= in_cksum(m
, hlen
);
1720 ip
->ip_len
= SWAP16(ip
->ip_len
);
1724 if (M_LEADINGSPACE(m
) < ip_offset
)
1725 panic("in_cksum_offset - chain modified!\n");
1726 m
->m_len
+= ip_offset
;
1727 m
->m_data
-= ip_offset
;
1730 /* Insert the checksum in the existing chain if IP header not contiguous */
1731 if (ip_offset
+ sizeof(struct ip
) > m
->m_len
) {
1735 printf("in_cksum_offset m_copyback, m->len: %lu off: %lu p: %d\n",
1736 m
->m_len
, ip_offset
+ offsetof(struct ip
, ip_sum
), ip
->ip_p
);
1738 *(u_short
*)tmp
= ip
->ip_sum
;
1739 m_copyback(m
, ip_offset
+ offsetof(struct ip
, ip_sum
), 2, tmp
);
1744 * Insert IP options into preformed packet.
1745 * Adjust IP destination as required for IP source routing,
1746 * as indicated by a non-zero in_addr at the start of the options.
1748 * XXX This routine assumes that the packet has no options in place.
1750 static struct mbuf
*
1751 ip_insertoptions(m
, opt
, phlen
)
1752 register struct mbuf
*m
;
1756 register struct ipoption
*p
= mtod(opt
, struct ipoption
*);
1758 register struct ip
*ip
= mtod(m
, struct ip
*);
1761 optlen
= opt
->m_len
- sizeof(p
->ipopt_dst
);
1762 if (optlen
+ (u_short
)ip
->ip_len
> IP_MAXPACKET
)
1763 return (m
); /* XXX should fail */
1764 if (p
->ipopt_dst
.s_addr
)
1765 ip
->ip_dst
= p
->ipopt_dst
;
1766 if (m
->m_flags
& M_EXT
|| m
->m_data
- optlen
< m
->m_pktdat
) {
1767 MGETHDR(n
, M_DONTWAIT
, MT_HEADER
); /* MAC-OK */
1770 n
->m_pkthdr
.rcvif
= 0;
1772 mac_mbuf_label_copy(m
, n
);
1774 n
->m_pkthdr
.len
= m
->m_pkthdr
.len
+ optlen
;
1775 m
->m_len
-= sizeof(struct ip
);
1776 m
->m_data
+= sizeof(struct ip
);
1779 m
->m_len
= optlen
+ sizeof(struct ip
);
1780 m
->m_data
+= max_linkhdr
;
1781 (void)memcpy(mtod(m
, void *), ip
, sizeof(struct ip
));
1783 m
->m_data
-= optlen
;
1785 m
->m_pkthdr
.len
+= optlen
;
1786 ovbcopy((caddr_t
)ip
, mtod(m
, caddr_t
), sizeof(struct ip
));
1788 ip
= mtod(m
, struct ip
*);
1789 bcopy(p
->ipopt_list
, ip
+ 1, optlen
);
1790 *phlen
= sizeof(struct ip
) + optlen
;
1791 ip
->ip_vhl
= IP_MAKE_VHL(IPVERSION
, *phlen
>> 2);
1792 ip
->ip_len
+= optlen
;
1797 * Copy options from ip to jp,
1798 * omitting those not copied during fragmentation.
1804 register u_char
*cp
, *dp
;
1805 int opt
, optlen
, cnt
;
1807 cp
= (u_char
*)(ip
+ 1);
1808 dp
= (u_char
*)(jp
+ 1);
1809 cnt
= (IP_VHL_HL(ip
->ip_vhl
) << 2) - sizeof (struct ip
);
1810 for (; cnt
> 0; cnt
-= optlen
, cp
+= optlen
) {
1812 if (opt
== IPOPT_EOL
)
1814 if (opt
== IPOPT_NOP
) {
1815 /* Preserve for IP mcast tunnel's LSRR alignment. */
1821 if (cnt
< IPOPT_OLEN
+ sizeof(*cp
))
1822 panic("malformed IPv4 option passed to ip_optcopy");
1824 optlen
= cp
[IPOPT_OLEN
];
1826 if (optlen
< IPOPT_OLEN
+ sizeof(*cp
) || optlen
> cnt
)
1827 panic("malformed IPv4 option passed to ip_optcopy");
1829 /* bogus lengths should have been caught by ip_dooptions */
1832 if (IPOPT_COPIED(opt
)) {
1833 bcopy(cp
, dp
, optlen
);
1837 for (optlen
= dp
- (u_char
*)(jp
+1); optlen
& 0x3; optlen
++)
1843 * IP socket option processing.
1846 ip_ctloutput(so
, sopt
)
1848 struct sockopt
*sopt
;
1850 struct inpcb
*inp
= sotoinpcb(so
);
1854 if (sopt
->sopt_level
!= IPPROTO_IP
) {
1858 switch (sopt
->sopt_dir
) {
1860 switch (sopt
->sopt_name
) {
1867 if (sopt
->sopt_valsize
> MLEN
) {
1871 MGET(m
, sopt
->sopt_p
? M_WAIT
: M_DONTWAIT
, MT_HEADER
);
1876 m
->m_len
= sopt
->sopt_valsize
;
1877 error
= sooptcopyin(sopt
, mtod(m
, char *), m
->m_len
,
1882 return (ip_pcbopts(sopt
->sopt_name
, &inp
->inp_options
,
1889 case IP_RECVRETOPTS
:
1890 case IP_RECVDSTADDR
:
1893 #if defined(NFAITH) && NFAITH > 0
1896 error
= sooptcopyin(sopt
, &optval
, sizeof optval
,
1901 switch (sopt
->sopt_name
) {
1903 inp
->inp_ip_tos
= optval
;
1907 inp
->inp_ip_ttl
= optval
;
1909 #define OPTSET(bit) \
1911 inp->inp_flags |= bit; \
1913 inp->inp_flags &= ~bit;
1916 OPTSET(INP_RECVOPTS
);
1919 case IP_RECVRETOPTS
:
1920 OPTSET(INP_RECVRETOPTS
);
1923 case IP_RECVDSTADDR
:
1924 OPTSET(INP_RECVDSTADDR
);
1932 OPTSET(INP_RECVTTL
);
1935 #if defined(NFAITH) && NFAITH > 0
1944 #if CONFIG_FORCE_OUT_IFP
1945 case IP_FORCE_OUT_IFP
: {
1946 char ifname
[IFNAMSIZ
];
1949 /* Verify interface name parameter is sane */
1950 if (sopt
->sopt_valsize
> sizeof(ifname
)) {
1955 /* Copy the interface name */
1956 if (sopt
->sopt_valsize
!= 0) {
1957 error
= sooptcopyin(sopt
, ifname
, sizeof(ifname
), sopt
->sopt_valsize
);
1962 if (sopt
->sopt_valsize
== 0 || ifname
[0] == 0) {
1963 // Set pdp_ifp to NULL
1964 inp
->pdp_ifp
= NULL
;
1967 if (inp
->inp_route
.ro_rt
) {
1968 rtfree(inp
->inp_route
.ro_rt
);
1969 inp
->inp_route
.ro_rt
= NULL
;
1975 /* Verify name is NULL terminated */
1976 if (ifname
[sopt
->sopt_valsize
- 1] != 0) {
1981 if (ifnet_find_by_name(ifname
, &ifp
) != 0) {
1986 /* Won't actually free. Since we don't release this later, we should do it now. */
1989 /* This only works for point-to-point interfaces */
1990 if ((ifp
->if_flags
& IFF_POINTOPOINT
) == 0) {
1999 case IP_MULTICAST_IF
:
2000 case IP_MULTICAST_VIF
:
2001 case IP_MULTICAST_TTL
:
2002 case IP_MULTICAST_LOOP
:
2003 case IP_ADD_MEMBERSHIP
:
2004 case IP_DROP_MEMBERSHIP
:
2005 error
= ip_setmoptions(sopt
, &inp
->inp_moptions
);
2009 error
= sooptcopyin(sopt
, &optval
, sizeof optval
,
2015 case IP_PORTRANGE_DEFAULT
:
2016 inp
->inp_flags
&= ~(INP_LOWPORT
);
2017 inp
->inp_flags
&= ~(INP_HIGHPORT
);
2020 case IP_PORTRANGE_HIGH
:
2021 inp
->inp_flags
&= ~(INP_LOWPORT
);
2022 inp
->inp_flags
|= INP_HIGHPORT
;
2025 case IP_PORTRANGE_LOW
:
2026 inp
->inp_flags
&= ~(INP_HIGHPORT
);
2027 inp
->inp_flags
|= INP_LOWPORT
;
2037 case IP_IPSEC_POLICY
:
2045 if (sopt
->sopt_valsize
> MCLBYTES
) {
2049 if ((error
= soopt_getm(sopt
, &m
)) != 0) /* XXX */
2051 if ((error
= soopt_mcopyin(sopt
, m
)) != 0) /* XXX */
2053 priv
= (sopt
->sopt_p
!= NULL
&&
2054 proc_suser(sopt
->sopt_p
) != 0) ? 0 : 1;
2056 req
= mtod(m
, caddr_t
);
2059 optname
= sopt
->sopt_name
;
2060 error
= ipsec4_set_policy(inp
, optname
, req
, len
, priv
);
2067 case IP_TRAFFIC_MGT_BACKGROUND
:
2069 unsigned background
= 0;
2070 error
= sooptcopyin(sopt
, &background
, sizeof(background
), sizeof(background
));
2075 so
->so_traffic_mgt_flags
|= TRAFFIC_MGT_SO_BACKGROUND
;
2077 so
->so_traffic_mgt_flags
&= ~TRAFFIC_MGT_SO_BACKGROUND
;
2081 #endif /* TRAFFIC_MGT */
2084 error
= ENOPROTOOPT
;
2090 switch (sopt
->sopt_name
) {
2093 if (inp
->inp_options
)
2094 error
= sooptcopyout(sopt
,
2095 mtod(inp
->inp_options
,
2097 inp
->inp_options
->m_len
);
2099 sopt
->sopt_valsize
= 0;
2105 case IP_RECVRETOPTS
:
2106 case IP_RECVDSTADDR
:
2110 #if defined(NFAITH) && NFAITH > 0
2113 switch (sopt
->sopt_name
) {
2116 optval
= inp
->inp_ip_tos
;
2120 optval
= inp
->inp_ip_ttl
;
2123 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
2126 optval
= OPTBIT(INP_RECVOPTS
);
2129 case IP_RECVRETOPTS
:
2130 optval
= OPTBIT(INP_RECVRETOPTS
);
2133 case IP_RECVDSTADDR
:
2134 optval
= OPTBIT(INP_RECVDSTADDR
);
2138 optval
= OPTBIT(INP_RECVIF
);
2142 optval
= OPTBIT(INP_RECVTTL
);
2146 if (inp
->inp_flags
& INP_HIGHPORT
)
2147 optval
= IP_PORTRANGE_HIGH
;
2148 else if (inp
->inp_flags
& INP_LOWPORT
)
2149 optval
= IP_PORTRANGE_LOW
;
2154 #if defined(NFAITH) && NFAITH > 0
2156 optval
= OPTBIT(INP_FAITH
);
2160 error
= sooptcopyout(sopt
, &optval
, sizeof optval
);
2163 case IP_MULTICAST_IF
:
2164 case IP_MULTICAST_VIF
:
2165 case IP_MULTICAST_TTL
:
2166 case IP_MULTICAST_LOOP
:
2167 case IP_ADD_MEMBERSHIP
:
2168 case IP_DROP_MEMBERSHIP
:
2169 error
= ip_getmoptions(sopt
, inp
->inp_moptions
);
2173 case IP_IPSEC_POLICY
:
2175 struct mbuf
*m
= NULL
;
2180 req
= mtod(m
, caddr_t
);
2183 error
= ipsec4_get_policy(sotoinpcb(so
), req
, len
, &m
);
2185 error
= soopt_mcopyout(sopt
, m
); /* XXX */
2193 case IP_TRAFFIC_MGT_BACKGROUND
:
2195 unsigned background
= so
->so_traffic_mgt_flags
;
2196 return (sooptcopyout(sopt
, &background
, sizeof(background
)));
2199 #endif /* TRAFFIC_MGT */
2202 error
= ENOPROTOOPT
;
2211 * Set up IP options in pcb for insertion in output packets.
2212 * Store in mbuf with pointer in pcbopt, adding pseudo-option
2213 * with destination address if source routed.
2217 __unused
int optname
,
2218 struct mbuf
**pcbopt
,
2219 register struct mbuf
*m
)
2221 register int cnt
, optlen
;
2222 register u_char
*cp
;
2225 /* turn off any old options */
2227 (void)m_free(*pcbopt
);
2229 if (m
== (struct mbuf
*)0 || m
->m_len
== 0) {
2231 * Only turning off any previous options.
2239 if (m
->m_len
% sizeof(int32_t))
2243 * IP first-hop destination address will be stored before
2244 * actual options; move other options back
2245 * and clear it when none present.
2247 if (m
->m_data
+ m
->m_len
+ sizeof(struct in_addr
) >= &m
->m_dat
[MLEN
])
2250 m
->m_len
+= sizeof(struct in_addr
);
2251 cp
= mtod(m
, u_char
*) + sizeof(struct in_addr
);
2252 ovbcopy(mtod(m
, caddr_t
), (caddr_t
)cp
, (unsigned)cnt
);
2253 bzero(mtod(m
, caddr_t
), sizeof(struct in_addr
));
2255 for (; cnt
> 0; cnt
-= optlen
, cp
+= optlen
) {
2256 opt
= cp
[IPOPT_OPTVAL
];
2257 if (opt
== IPOPT_EOL
)
2259 if (opt
== IPOPT_NOP
)
2262 if (cnt
< IPOPT_OLEN
+ sizeof(*cp
))
2264 optlen
= cp
[IPOPT_OLEN
];
2265 if (optlen
< IPOPT_OLEN
+ sizeof(*cp
) || optlen
> cnt
)
2276 * user process specifies route as:
2278 * D must be our final destination (but we can't
2279 * check that since we may not have connected yet).
2280 * A is first hop destination, which doesn't appear in
2281 * actual IP option, but is stored before the options.
2283 if (optlen
< IPOPT_MINOFF
- 1 + sizeof(struct in_addr
))
2285 m
->m_len
-= sizeof(struct in_addr
);
2286 cnt
-= sizeof(struct in_addr
);
2287 optlen
-= sizeof(struct in_addr
);
2288 cp
[IPOPT_OLEN
] = optlen
;
2290 * Move first hop before start of options.
2292 bcopy((caddr_t
)&cp
[IPOPT_OFFSET
+1], mtod(m
, caddr_t
),
2293 sizeof(struct in_addr
));
2295 * Then copy rest of options back
2296 * to close up the deleted entry.
2298 ovbcopy((caddr_t
)(&cp
[IPOPT_OFFSET
+1] +
2299 sizeof(struct in_addr
)),
2300 (caddr_t
)&cp
[IPOPT_OFFSET
+1],
2301 (unsigned)cnt
+ sizeof(struct in_addr
));
2305 if (m
->m_len
> MAX_IPOPTLEN
+ sizeof(struct in_addr
))
2317 * The whole multicast option thing needs to be re-thought.
2318 * Several of these options are equally applicable to non-multicast
2319 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
2320 * standard option (IP_TTL).
2324 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
2326 static struct ifnet
*
2327 ip_multicast_if(a
, ifindexp
)
2336 if (ntohl(a
->s_addr
) >> 24 == 0) {
2337 ifindex
= ntohl(a
->s_addr
) & 0xffffff;
2338 ifnet_head_lock_shared();
2339 if (ifindex
< 0 || if_index
< ifindex
) {
2343 ifp
= ifindex2ifnet
[ifindex
];
2346 *ifindexp
= ifindex
;
2348 INADDR_TO_IFP(*a
, ifp
);
2354 * Set the IP multicast options in response to user setsockopt().
2357 ip_setmoptions(sopt
, imop
)
2358 struct sockopt
*sopt
;
2359 struct ip_moptions
**imop
;
2363 struct in_addr addr
;
2364 struct ip_mreq mreq
;
2365 struct ifnet
*ifp
= NULL
;
2366 struct ip_moptions
*imo
= *imop
;
2371 * No multicast option buffer attached to the pcb;
2372 * allocate one and initialize to default values.
2374 error
= ip_createmoptions(imop
);
2380 switch (sopt
->sopt_name
) {
2381 /* store an index number for the vif you wanna use in the send */
2383 case IP_MULTICAST_VIF
:
2384 if (legal_vif_num
== 0) {
2388 error
= sooptcopyin(sopt
, &i
, sizeof i
, sizeof i
);
2391 if (!legal_vif_num(i
) && (i
!= -1)) {
2395 imo
->imo_multicast_vif
= i
;
2397 #endif /* MROUTING */
2399 case IP_MULTICAST_IF
:
2401 * Select the interface for outgoing multicast packets.
2403 error
= sooptcopyin(sopt
, &addr
, sizeof addr
, sizeof addr
);
2407 * INADDR_ANY is used to remove a previous selection.
2408 * When no interface is selected, a default one is
2409 * chosen every time a multicast packet is sent.
2411 if (addr
.s_addr
== INADDR_ANY
) {
2412 imo
->imo_multicast_ifp
= NULL
;
2416 * The selected interface is identified by its local
2417 * IP address. Find the interface and confirm that
2418 * it supports multicasting.
2420 ifp
= ip_multicast_if(&addr
, &ifindex
);
2421 if (ifp
== NULL
|| (ifp
->if_flags
& IFF_MULTICAST
) == 0) {
2422 error
= EADDRNOTAVAIL
;
2425 imo
->imo_multicast_ifp
= ifp
;
2427 imo
->imo_multicast_addr
= addr
;
2429 imo
->imo_multicast_addr
.s_addr
= INADDR_ANY
;
2432 case IP_MULTICAST_TTL
:
2434 * Set the IP time-to-live for outgoing multicast packets.
2435 * The original multicast API required a char argument,
2436 * which is inconsistent with the rest of the socket API.
2437 * We allow either a char or an int.
2439 if (sopt
->sopt_valsize
== 1) {
2441 error
= sooptcopyin(sopt
, &ttl
, 1, 1);
2444 imo
->imo_multicast_ttl
= ttl
;
2447 error
= sooptcopyin(sopt
, &ttl
, sizeof ttl
,
2454 imo
->imo_multicast_ttl
= ttl
;
2458 case IP_MULTICAST_LOOP
:
2460 * Set the loopback flag for outgoing multicast packets.
2461 * Must be zero or one. The original multicast API required a
2462 * char argument, which is inconsistent with the rest
2463 * of the socket API. We allow either a char or an int.
2465 if (sopt
->sopt_valsize
== 1) {
2467 error
= sooptcopyin(sopt
, &loop
, 1, 1);
2470 imo
->imo_multicast_loop
= !!loop
;
2473 error
= sooptcopyin(sopt
, &loop
, sizeof loop
,
2477 imo
->imo_multicast_loop
= !!loop
;
2481 case IP_ADD_MEMBERSHIP
:
2483 * Add a multicast group membership.
2484 * Group must be a valid IP multicast address.
2486 error
= sooptcopyin(sopt
, &mreq
, sizeof mreq
, sizeof mreq
);
2490 error
= ip_addmembership(imo
, &mreq
);
2493 case IP_DROP_MEMBERSHIP
:
2495 * Drop a multicast group membership.
2496 * Group must be a valid IP multicast address.
2498 error
= sooptcopyin(sopt
, &mreq
, sizeof mreq
, sizeof mreq
);
2502 error
= ip_dropmembership(imo
, &mreq
);
2511 * If all options have default values, no need to keep the mbuf.
2513 if (imo
->imo_multicast_ifp
== NULL
&&
2514 imo
->imo_multicast_vif
== (u_long
)-1 &&
2515 imo
->imo_multicast_ttl
== IP_DEFAULT_MULTICAST_TTL
&&
2516 imo
->imo_multicast_loop
== IP_DEFAULT_MULTICAST_LOOP
&&
2517 imo
->imo_num_memberships
== 0) {
2518 FREE(*imop
, M_IPMOPTS
);
2526 * Set the IP multicast options in response to user setsockopt().
2528 __private_extern__
int
2530 struct ip_moptions
**imop
)
2532 struct ip_moptions
*imo
;
2533 imo
= (struct ip_moptions
*) _MALLOC(sizeof(*imo
), M_IPMOPTS
,
2539 imo
->imo_multicast_ifp
= NULL
;
2540 imo
->imo_multicast_addr
.s_addr
= INADDR_ANY
;
2541 imo
->imo_multicast_vif
= -1;
2542 imo
->imo_multicast_ttl
= IP_DEFAULT_MULTICAST_TTL
;
2543 imo
->imo_multicast_loop
= IP_DEFAULT_MULTICAST_LOOP
;
2544 imo
->imo_num_memberships
= 0;
2550 * Add membership to an IPv4 multicast.
2552 __private_extern__
int
2554 struct ip_moptions
*imo
,
2555 struct ip_mreq
*mreq
)
2558 struct sockaddr_in
*dst
;
2559 struct ifnet
*ifp
= NULL
;
2563 if (!IN_MULTICAST(ntohl(mreq
->imr_multiaddr
.s_addr
))) {
2568 * If no interface address was provided, use the interface of
2569 * the route to the given multicast address.
2571 if (mreq
->imr_interface
.s_addr
== INADDR_ANY
) {
2572 bzero((caddr_t
)&ro
, sizeof(ro
));
2573 dst
= (struct sockaddr_in
*)&ro
.ro_dst
;
2574 dst
->sin_len
= sizeof(*dst
);
2575 dst
->sin_family
= AF_INET
;
2576 dst
->sin_addr
= mreq
->imr_multiaddr
;
2577 lck_mtx_lock(rt_mtx
);
2578 rtalloc_ign_locked(&ro
, 0UL);
2579 if (ro
.ro_rt
!= NULL
) {
2580 ifp
= ro
.ro_rt
->rt_ifp
;
2581 rtfree_locked(ro
.ro_rt
);
2584 /* If there's no default route, try using loopback */
2585 mreq
->imr_interface
.s_addr
= INADDR_LOOPBACK
;
2587 lck_mtx_unlock(rt_mtx
);
2591 ifp
= ip_multicast_if(&mreq
->imr_interface
, NULL
);
2595 * See if we found an interface, and confirm that it
2596 * supports multicast.
2598 if (ifp
== NULL
|| (ifp
->if_flags
& IFF_MULTICAST
) == 0) {
2599 error
= EADDRNOTAVAIL
;
2603 * See if the membership already exists or if all the
2604 * membership slots are full.
2606 for (i
= 0; i
< imo
->imo_num_memberships
; ++i
) {
2607 if (imo
->imo_membership
[i
]->inm_ifp
== ifp
&&
2608 imo
->imo_membership
[i
]->inm_addr
.s_addr
2609 == mreq
->imr_multiaddr
.s_addr
)
2612 if (i
< imo
->imo_num_memberships
) {
2616 if (i
== IP_MAX_MEMBERSHIPS
) {
2617 error
= ETOOMANYREFS
;
2621 * Everything looks good; add a new record to the multicast
2622 * address list for the given interface.
2624 if ((imo
->imo_membership
[i
] =
2625 in_addmulti(&mreq
->imr_multiaddr
, ifp
)) == NULL
) {
2629 ++imo
->imo_num_memberships
;
2635 * Drop membership of an IPv4 multicast.
2637 __private_extern__
int
2639 struct ip_moptions
*imo
,
2640 struct ip_mreq
*mreq
)
2643 struct ifnet
* ifp
= NULL
;
2646 if (!IN_MULTICAST(ntohl(mreq
->imr_multiaddr
.s_addr
))) {
2652 * If an interface address was specified, get a pointer
2653 * to its ifnet structure.
2655 if (mreq
->imr_interface
.s_addr
== INADDR_ANY
)
2658 ifp
= ip_multicast_if(&mreq
->imr_interface
, NULL
);
2660 error
= EADDRNOTAVAIL
;
2665 * Find the membership in the membership array.
2667 for (i
= 0; i
< imo
->imo_num_memberships
; ++i
) {
2669 imo
->imo_membership
[i
]->inm_ifp
== ifp
) &&
2670 imo
->imo_membership
[i
]->inm_addr
.s_addr
==
2671 mreq
->imr_multiaddr
.s_addr
)
2674 if (i
== imo
->imo_num_memberships
) {
2675 error
= EADDRNOTAVAIL
;
2679 * Give up the multicast address record to which the
2680 * membership points.
2682 in_delmulti(&imo
->imo_membership
[i
]);
2684 * Remove the gap in the membership array.
2686 for (++i
; i
< imo
->imo_num_memberships
; ++i
)
2687 imo
->imo_membership
[i
-1] = imo
->imo_membership
[i
];
2688 --imo
->imo_num_memberships
;
2694 * Return the IP multicast options in response to user getsockopt().
2697 ip_getmoptions(sopt
, imo
)
2698 struct sockopt
*sopt
;
2699 register struct ip_moptions
*imo
;
2701 struct in_addr addr
;
2702 struct in_ifaddr
*ia
;
2707 switch (sopt
->sopt_name
) {
2709 case IP_MULTICAST_VIF
:
2711 optval
= imo
->imo_multicast_vif
;
2714 error
= sooptcopyout(sopt
, &optval
, sizeof optval
);
2716 #endif /* MROUTING */
2718 case IP_MULTICAST_IF
:
2719 if (imo
== NULL
|| imo
->imo_multicast_ifp
== NULL
)
2720 addr
.s_addr
= INADDR_ANY
;
2721 else if (imo
->imo_multicast_addr
.s_addr
) {
2722 /* return the value user has set */
2723 addr
= imo
->imo_multicast_addr
;
2725 IFP_TO_IA(imo
->imo_multicast_ifp
, ia
);
2726 addr
.s_addr
= (ia
== NULL
) ? INADDR_ANY
2727 : IA_SIN(ia
)->sin_addr
.s_addr
;
2729 error
= sooptcopyout(sopt
, &addr
, sizeof addr
);
2732 case IP_MULTICAST_TTL
:
2734 optval
= coptval
= IP_DEFAULT_MULTICAST_TTL
;
2736 optval
= coptval
= imo
->imo_multicast_ttl
;
2737 if (sopt
->sopt_valsize
== 1)
2738 error
= sooptcopyout(sopt
, &coptval
, 1);
2740 error
= sooptcopyout(sopt
, &optval
, sizeof optval
);
2743 case IP_MULTICAST_LOOP
:
2745 optval
= coptval
= IP_DEFAULT_MULTICAST_LOOP
;
2747 optval
= coptval
= imo
->imo_multicast_loop
;
2748 if (sopt
->sopt_valsize
== 1)
2749 error
= sooptcopyout(sopt
, &coptval
, 1);
2751 error
= sooptcopyout(sopt
, &optval
, sizeof optval
);
2755 error
= ENOPROTOOPT
;
2762 * Discard the IP multicast options.
2765 ip_freemoptions(imo
)
2766 register struct ip_moptions
*imo
;
2771 for (i
= 0; i
< imo
->imo_num_memberships
; ++i
)
2772 in_delmulti(&imo
->imo_membership
[i
]);
2773 FREE(imo
, M_IPMOPTS
);
2778 * Routine called from ip_output() to loop back a copy of an IP multicast
2779 * packet to the input queue of a specified interface. Note that this
2780 * calls the output routine of the loopback "driver", but with an interface
2781 * pointer that might NOT be a loopback interface -- evil, but easier than
2782 * replicating that code here.
2785 ip_mloopback(ifp
, m
, dst
, hlen
)
2787 register struct mbuf
*m
;
2788 register struct sockaddr_in
*dst
;
2791 register struct ip
*ip
;
2793 int sw_csum
= (apple_hwcksum_tx
== 0);
2795 copym
= m_copy(m
, 0, M_COPYALL
);
2796 if (copym
!= NULL
&& (copym
->m_flags
& M_EXT
|| copym
->m_len
< hlen
))
2797 copym
= m_pullup(copym
, hlen
);
2803 * We don't bother to fragment if the IP length is greater
2804 * than the interface's MTU. Can this possibly matter?
2806 ip
= mtod(copym
, struct ip
*);
2810 ip
->ip_sum
= in_cksum(copym
, hlen
);
2813 * It's not clear whether there are any lingering
2814 * reentrancy problems in other areas which might
2815 * be exposed by using ip_input directly (in
2816 * particular, everything which modifies the packet
2817 * in-place). Yet another option is using the
2818 * protosw directly to deliver the looped back
2819 * packet. For the moment, we'll err on the side
2820 * of safety by using if_simloop().
2823 if (dst
->sin_family
!= AF_INET
) {
2824 printf("ip_mloopback: bad address family %d\n",
2826 dst
->sin_family
= AF_INET
;
2831 * Mark checksum as valid or calculate checksum for loopback.
2833 * This is done this way because we have to embed the ifp of
2834 * the interface we will send the original copy of the packet
2835 * out on in the mbuf. ip_input will check if_hwassist of the
2836 * embedded ifp and ignore all csum_flags if if_hwassist is 0.
2837 * The UDP checksum has not been calculated yet.
2839 if (sw_csum
|| (copym
->m_pkthdr
.csum_flags
& CSUM_DELAY_DATA
)) {
2840 if (!sw_csum
&& IF_HWASSIST_CSUM_FLAGS(ifp
->if_hwassist
)) {
2841 copym
->m_pkthdr
.csum_flags
|=
2842 CSUM_DATA_VALID
| CSUM_PSEUDO_HDR
|
2843 CSUM_IP_CHECKED
| CSUM_IP_VALID
;
2844 copym
->m_pkthdr
.csum_data
= 0xffff;
2847 in_delayed_cksum(copym
);
2854 * We need to send all loopback traffic down to dlil in case
2855 * a filter has tapped-in.
2859 * Stuff the 'real' ifp into the pkthdr, to be used in matching
2860 * in ip_input(); we need the loopback ifp/dl_tag passed as args
2861 * to make the loopback driver compliant with the data link
2865 copym
->m_pkthdr
.rcvif
= ifp
;
2866 dlil_output(lo_ifp
, PF_INET
, copym
, 0,
2867 (struct sockaddr
*) dst
, 0);
2869 printf("Warning: ip_output call to dlil_find_dltag failed!\n");