]> git.saurik.com Git - apple/xnu.git/blob - bsd/netinet/ip_output.c
projects / apple / xnu.git / blob
? search:


b6b0817cc860410069489a3e305991fdf44a6d62
[apple/xnu.git] / bsd / netinet / ip_output.c
1 /*
2 * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
3 *
4 * @APPLE_LICENSE_HEADER_START@
5 *
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. Please obtain a copy of the License at
10 * http://www.opensource.apple.com/apsl/ and read it before using this
11 * file.
12 *
13 * The Original Code and all software distributed under the License are
14 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
15 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
16 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
18 * Please see the License for the specific language governing rights and
19 * limitations under the License.
20 *
21 * @APPLE_LICENSE_HEADER_END@
22 */
23 /*
24 * Copyright (c) 1982, 1986, 1988, 1990, 1993
25 * The Regents of the University of California. All rights reserved.
26 *
27 * Redistribution and use in source and binary forms, with or without
28 * modification, are permitted provided that the following conditions
29 * are met:
30 * 1. Redistributions of source code must retain the above copyright
31 * notice, this list of conditions and the following disclaimer.
32 * 2. Redistributions in binary form must reproduce the above copyright
33 * notice, this list of conditions and the following disclaimer in the
34 * documentation and/or other materials provided with the distribution.
35 * 3. All advertising materials mentioning features or use of this software
36 * must display the following acknowledgement:
37 * This product includes software developed by the University of
38 * California, Berkeley and its contributors.
39 * 4. Neither the name of the University nor the names of its contributors
40 * may be used to endorse or promote products derived from this software
41 * without specific prior written permission.
42 *
43 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
44 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
45 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
46 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
47 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
48 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
49 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
50 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
51 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
52 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
53 * SUCH DAMAGE.
54 *
55 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
56 * $FreeBSD: src/sys/netinet/ip_output.c,v 1.99.2.16 2001/07/19 06:37:26 kris Exp $
57 */
58
59 #define _IP_VHL
60
61 #include <sys/param.h>
62 #include <sys/systm.h>
63 #include <sys/kernel.h>
64 #include <sys/malloc.h>
65 #include <sys/mbuf.h>
66 #include <sys/protosw.h>
67 #include <sys/socket.h>
68 #include <sys/socketvar.h>
69 #include <kern/locks.h>
70 #include <sys/sysctl.h>
71
72 #include <net/if.h>
73 #include <net/route.h>
74
75 #include <netinet/in.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/ip.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/in_var.h>
80 #include <netinet/ip_var.h>
81
82 #include <netinet/kpi_ipfilter_var.h>
83
84 #include "faith.h"
85
86 #include <net/dlil.h>
87 #include <sys/kdebug.h>
88
89 #define DBG_LAYER_BEG NETDBG_CODE(DBG_NETIP, 1)
90 #define DBG_LAYER_END NETDBG_CODE(DBG_NETIP, 3)
91 #define DBG_FNC_IP_OUTPUT NETDBG_CODE(DBG_NETIP, (1 << 8) | 1)
92 #define DBG_FNC_IPSEC4_OUTPUT NETDBG_CODE(DBG_NETIP, (2 << 8) | 1)
93
94
95 #if IPSEC
96 #include <netinet6/ipsec.h>
97 #include <netkey/key.h>
98 #if IPSEC_DEBUG
99 #include <netkey/key_debug.h>
100 #else
101 #define KEYDEBUG(lev,arg)
102 #endif
103 #endif /*IPSEC*/
104
105 #include <netinet/ip_fw.h>
106 #include <netinet/ip_divert.h>
107
108 #if DUMMYNET
109 #include <netinet/ip_dummynet.h>
110 #endif
111
112 #if IPFIREWALL_FORWARD_DEBUG
113 #define print_ip(a) printf("%ld.%ld.%ld.%ld",(ntohl(a.s_addr)>>24)&0xFF,\
114 (ntohl(a.s_addr)>>16)&0xFF,\
115 (ntohl(a.s_addr)>>8)&0xFF,\
116 (ntohl(a.s_addr))&0xFF);
117 #endif
118
119 #if IPSEC
120 extern lck_mtx_t *sadb_mutex;
121 #endif
122
123 u_short ip_id;
124
125 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
126 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
127 static void ip_mloopback(struct ifnet *, struct mbuf *,
128 struct sockaddr_in *, int);
129 static int ip_getmoptions(struct sockopt *, struct ip_moptions *);
130 static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
131 static int ip_setmoptions(struct sockopt *, struct ip_moptions **);
132
133 int ip_createmoptions(struct ip_moptions **imop);
134 int ip_addmembership(struct ip_moptions *imo, struct ip_mreq *mreq);
135 int ip_dropmembership(struct ip_moptions *imo, struct ip_mreq *mreq);
136 int ip_optcopy(struct ip *, struct ip *);
137 extern int (*fr_checkp)(struct ip *, int, struct ifnet *, int, struct mbuf **);
138 #ifdef __APPLE__
139 extern struct mbuf* m_dup(register struct mbuf *m, int how);
140 #endif
141
142 extern int apple_hwcksum_tx;
143 extern u_long route_generation;
144
145 extern struct protosw inetsw[];
146
147 extern struct ip_linklocal_stat ip_linklocal_stat;
148 extern lck_mtx_t *ip_mutex;
149
150 /* temporary: for testing */
151 #if IPSEC
152 extern int ipsec_bypass;
153 #endif
154
155 static int ip_maxchainsent = 0;
156 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxchainsent, CTLFLAG_RW,
157 &ip_maxchainsent, 0, "use dlil_output_list");
158 /*
159 * IP output. The packet in mbuf chain m contains a skeletal IP
160 * header (with len, off, ttl, proto, tos, src, dst).
161 * The mbuf chain containing the packet will be freed.
162 * The mbuf opt, if present, will not be freed.
163 */
164 int
165 ip_output(
166 struct mbuf *m0,
167 struct mbuf *opt,
168 struct route *ro,
169 int flags,
170 struct ip_moptions *imo)
171 {
172 int error;
173 error = ip_output_list(m0, 0, opt, ro, flags, imo);
174 return error;
175 }
176
177 int
178 ip_output_list(
179 struct mbuf *m0,
180 int packetchain,
181 struct mbuf *opt,
182 struct route *ro,
183 int flags,
184 struct ip_moptions *imo)
185 {
186 struct ip *ip, *mhip;
187 struct ifnet *ifp = NULL;
188 struct mbuf *m = m0;
189 int hlen = sizeof (struct ip);
190 int len, off, error = 0;
191 struct sockaddr_in *dst = NULL;
192 struct in_ifaddr *ia = NULL;
193 int isbroadcast, sw_csum;
194 struct in_addr pkt_dst;
195 #if IPSEC
196 struct route iproute;
197 struct socket *so = NULL;
198 struct secpolicy *sp = NULL;
199 #endif
200 #if IPFIREWALL_FORWARD
201 int fwd_rewrite_src = 0;
202 #endif
203 struct ip_fw_args args;
204 int didfilter = 0;
205 ipfilter_t inject_filter_ref = 0;
206 struct m_tag *tag;
207 struct route dn_route;
208 struct mbuf * packetlist;
209 int pktcnt = 0;
210
211 lck_mtx_lock(ip_mutex);
212
213 KERNEL_DEBUG(DBG_FNC_IP_OUTPUT | DBG_FUNC_START, 0,0,0,0,0);
214
215 packetlist = m0;
216 args.eh = NULL;
217 args.rule = NULL;
218 args.next_hop = NULL;
219 args.divert_rule = 0; /* divert cookie */
220
221 /* Grab info from mtags prepended to the chain */
222 #if DUMMYNET
223 if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_DUMMYNET, NULL)) != NULL) {
224 struct dn_pkt_tag *dn_tag;
225
226 dn_tag = (struct dn_pkt_tag *)(tag+1);
227 args.rule = dn_tag->rule;
228 opt = NULL;
229 dn_route = dn_tag->ro;
230 ro = &dn_route;
231
232 imo = NULL;
233 dst = dn_tag->dn_dst;
234 ifp = dn_tag->ifp;
235 flags = dn_tag->flags;
236
237 m_tag_delete(m0, tag);
238 }
239 #endif /* DUMMYNET */
240
241 if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_DIVERT, NULL)) != NULL) {
242 struct divert_tag *div_tag;
243
244 div_tag = (struct divert_tag *)(tag+1);
245 args.divert_rule = div_tag->cookie;
246
247 m_tag_delete(m0, tag);
248 }
249 if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_IPFORWARD, NULL)) != NULL) {
250 struct ip_fwd_tag *ipfwd_tag;
251
252 ipfwd_tag = (struct ip_fwd_tag *)(tag+1);
253 args.next_hop = ipfwd_tag->next_hop;
254
255 m_tag_delete(m0, tag);
256 }
257
258 m = m0;
259
260 #if DIAGNOSTIC
261 if ( !m || (m->m_flags & M_PKTHDR) != 0)
262 panic("ip_output no HDR");
263 if (!ro)
264 panic("ip_output no route, proto = %d",
265 mtod(m, struct ip *)->ip_p);
266 #endif
267
268 if (args.rule != NULL) { /* dummynet already saw us */
269 ip = mtod(m, struct ip *);
270 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
271 if (ro->ro_rt != NULL)
272 ia = (struct in_ifaddr *)ro->ro_rt->rt_ifa;
273 if (ia)
274 ifaref(&ia->ia_ifa);
275 #if IPSEC
276 if (ipsec_bypass == 0 && (flags & IP_NOIPSEC) == 0) {
277 so = ipsec_getsocket(m);
278 (void)ipsec_setsocket(m, NULL);
279 }
280 #endif
281 goto sendit;
282 }
283
284 #if IPSEC
285 if (ipsec_bypass == 0 && (flags & IP_NOIPSEC) == 0) {
286 so = ipsec_getsocket(m);
287 (void)ipsec_setsocket(m, NULL);
288 }
289 #endif
290 loopit:
291 /*
292 * No need to proccess packet twice if we've
293 * already seen it
294 */
295 inject_filter_ref = ipf_get_inject_filter(m);
296
297 if (opt) {
298 m = ip_insertoptions(m, opt, &len);
299 hlen = len;
300 }
301 ip = mtod(m, struct ip *);
302 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
303
304 /*
305 * Fill in IP header.
306 */
307 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
308 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
309 ip->ip_off &= IP_DF;
310 #if RANDOM_IP_ID
311 ip->ip_id = ip_randomid();
312 #else
313 ip->ip_id = htons(ip_id++);
314 #endif
315 ipstat.ips_localout++;
316 } else {
317 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
318 }
319
320 KERNEL_DEBUG(DBG_LAYER_BEG, ip->ip_dst.s_addr,
321 ip->ip_src.s_addr, ip->ip_p, ip->ip_off, ip->ip_len);
322
323 dst = (struct sockaddr_in *)&ro->ro_dst;
324
325 /*
326 * If there is a cached route,
327 * check that it is to the same destination
328 * and is still up. If not, free it and try again.
329 * The address family should also be checked in case of sharing the
330 * cache with IPv6.
331 */
332
333 {
334 if (ro->ro_rt && (ro->ro_rt->generation_id != route_generation) &&
335 ((flags & (IP_ROUTETOIF | IP_FORWARDING)) == 0) && (ip->ip_src.s_addr != INADDR_ANY) &&
336 (ifa_foraddr(ip->ip_src.s_addr) == 0)) {
337 error = EADDRNOTAVAIL;
338 goto bad;
339 }
340 }
341 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
342 dst->sin_family != AF_INET ||
343 dst->sin_addr.s_addr != pkt_dst.s_addr)) {
344 rtfree(ro->ro_rt);
345 ro->ro_rt = (struct rtentry *)0;
346 }
347 if (ro->ro_rt == 0) {
348 bzero(dst, sizeof(*dst));
349 dst->sin_family = AF_INET;
350 dst->sin_len = sizeof(*dst);
351 dst->sin_addr = pkt_dst;
352 }
353 /*
354 * If routing to interface only,
355 * short circuit routing lookup.
356 */
357 #define ifatoia(ifa) ((struct in_ifaddr *)(ifa))
358 #define sintosa(sin) ((struct sockaddr *)(sin))
359 if (flags & IP_ROUTETOIF) {
360 if (ia)
361 ifafree(&ia->ia_ifa);
362 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0) {
363 if ((ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
364 ipstat.ips_noroute++;
365 error = ENETUNREACH;
366 goto bad;
367 }
368 }
369 ifp = ia->ia_ifp;
370 ip->ip_ttl = 1;
371 isbroadcast = in_broadcast(dst->sin_addr, ifp);
372 } else {
373 /*
374 * If this is the case, we probably don't want to allocate
375 * a protocol-cloned route since we didn't get one from the
376 * ULP. This lets TCP do its thing, while not burdening
377 * forwarding or ICMP with the overhead of cloning a route.
378 * Of course, we still want to do any cloning requested by
379 * the link layer, as this is probably required in all cases
380 * for correct operation (as it is for ARP).
381 */
382 if (ro->ro_rt == 0)
383 rtalloc_ign(ro, RTF_PRCLONING);
384 if (ro->ro_rt == 0) {
385 ipstat.ips_noroute++;
386 error = EHOSTUNREACH;
387 goto bad;
388 }
389 if (ia)
390 ifafree(&ia->ia_ifa);
391 ia = ifatoia(ro->ro_rt->rt_ifa);
392 if (ia)
393 ifaref(&ia->ia_ifa);
394 ifp = ro->ro_rt->rt_ifp;
395 ro->ro_rt->rt_use++;
396 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
397 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
398 if (ro->ro_rt->rt_flags & RTF_HOST)
399 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
400 else
401 isbroadcast = in_broadcast(dst->sin_addr, ifp);
402 }
403 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
404 struct in_multi *inm;
405
406 m->m_flags |= M_MCAST;
407 /*
408 * IP destination address is multicast. Make sure "dst"
409 * still points to the address in "ro". (It may have been
410 * changed to point to a gateway address, above.)
411 */
412 dst = (struct sockaddr_in *)&ro->ro_dst;
413 /*
414 * See if the caller provided any multicast options
415 */
416 if (imo != NULL) {
417 if ((flags & IP_RAWOUTPUT) == 0) ip->ip_ttl = imo->imo_multicast_ttl;
418 if (imo->imo_multicast_ifp != NULL) {
419 ifp = imo->imo_multicast_ifp;
420 }
421 if (imo->imo_multicast_vif != -1 &&
422 ((flags & IP_RAWOUTPUT) == 0 || ip->ip_src.s_addr == INADDR_ANY))
423 ip->ip_src.s_addr =
424 ip_mcast_src(imo->imo_multicast_vif);
425 } else
426 if ((flags & IP_RAWOUTPUT) == 0) ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
427 /*
428 * Confirm that the outgoing interface supports multicast.
429 */
430 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
431 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
432 ipstat.ips_noroute++;
433 error = ENETUNREACH;
434 goto bad;
435 }
436 }
437 /*
438 * If source address not specified yet, use address
439 * of outgoing interface.
440 */
441 if (ip->ip_src.s_addr == INADDR_ANY) {
442 register struct in_ifaddr *ia1;
443
444 TAILQ_FOREACH(ia1, &in_ifaddrhead, ia_link)
445 if (ia1->ia_ifp == ifp) {
446 ip->ip_src = IA_SIN(ia1)->sin_addr;
447
448 break;
449 }
450 if (ip->ip_src.s_addr == INADDR_ANY) {
451 error = ENETUNREACH;
452 goto bad;
453 }
454 }
455
456 ifnet_lock_shared(ifp);
457 IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
458 ifnet_lock_done(ifp);
459 if (inm != NULL &&
460 (imo == NULL || imo->imo_multicast_loop)) {
461 /*
462 * If we belong to the destination multicast group
463 * on the outgoing interface, and the caller did not
464 * forbid loopback, loop back a copy.
465 */
466 if (!TAILQ_EMPTY(&ipv4_filters)) {
467 struct ipfilter *filter;
468 int seen = (inject_filter_ref == 0);
469 struct ipf_pktopts *ippo = 0, ipf_pktopts;
470
471 if (imo) {
472 ippo = &ipf_pktopts;
473 ipf_pktopts.ippo_mcast_ifnet = imo->imo_multicast_ifp;
474 ipf_pktopts.ippo_mcast_ttl = imo->imo_multicast_ttl;
475 ipf_pktopts.ippo_mcast_loop = imo->imo_multicast_loop;
476 }
477
478 lck_mtx_unlock(ip_mutex);
479 ipf_ref();
480 TAILQ_FOREACH(filter, &ipv4_filters, ipf_link) {
481 if (seen == 0) {
482 if ((struct ipfilter *)inject_filter_ref == filter)
483 seen = 1;
484 } else if (filter->ipf_filter.ipf_output) {
485 errno_t result;
486 result = filter->ipf_filter.ipf_output(filter->ipf_filter.cookie, (mbuf_t*)&m, ippo);
487 if (result == EJUSTRETURN) {
488 ipf_unref();
489 goto done;
490 }
491 if (result != 0) {
492 ipf_unref();
493 lck_mtx_lock(ip_mutex);
494 goto bad;
495 }
496 }
497 }
498 lck_mtx_lock(ip_mutex);
499 ipf_unref();
500 didfilter = 1;
501 }
502 ip_mloopback(ifp, m, dst, hlen);
503 }
504 else {
505 /*
506 * If we are acting as a multicast router, perform
507 * multicast forwarding as if the packet had just
508 * arrived on the interface to which we are about
509 * to send. The multicast forwarding function
510 * recursively calls this function, using the
511 * IP_FORWARDING flag to prevent infinite recursion.
512 *
513 * Multicasts that are looped back by ip_mloopback(),
514 * above, will be forwarded by the ip_input() routine,
515 * if necessary.
516 */
517 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
518 /*
519 * Check if rsvp daemon is running. If not, don't
520 * set ip_moptions. This ensures that the packet
521 * is multicast and not just sent down one link
522 * as prescribed by rsvpd.
523 */
524 if (!rsvp_on)
525 imo = NULL;
526 if (ip_mforward(ip, ifp, m, imo) != 0) {
527 m_freem(m);
528 lck_mtx_unlock(ip_mutex);
529 goto done;
530 }
531 }
532 }
533
534 /*
535 * Multicasts with a time-to-live of zero may be looped-
536 * back, above, but must not be transmitted on a network.
537 * Also, multicasts addressed to the loopback interface
538 * are not sent -- the above call to ip_mloopback() will
539 * loop back a copy if this host actually belongs to the
540 * destination group on the loopback interface.
541 */
542 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
543 m_freem(m);
544 lck_mtx_unlock(ip_mutex);
545 goto done;
546 }
547
548 goto sendit;
549 }
550 #ifndef notdef
551 /*
552 * If source address not specified yet, use address
553 * of outgoing interface.
554 */
555 if (ip->ip_src.s_addr == INADDR_ANY) {
556 ip->ip_src = IA_SIN(ia)->sin_addr;
557 #if IPFIREWALL_FORWARD
558 /* Keep note that we did this - if the firewall changes
559 * the next-hop, our interface may change, changing the
560 * default source IP. It's a shame so much effort happens
561 * twice. Oh well.
562 */
563 fwd_rewrite_src++;
564 #endif /* IPFIREWALL_FORWARD */
565 }
566 #endif /* notdef */
567
568 /*
569 * Look for broadcast address and
570 * and verify user is allowed to send
571 * such a packet.
572 */
573 if (isbroadcast) {
574 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
575 error = EADDRNOTAVAIL;
576 goto bad;
577 }
578 if ((flags & IP_ALLOWBROADCAST) == 0) {
579 error = EACCES;
580 goto bad;
581 }
582 /* don't allow broadcast messages to be fragmented */
583 if ((u_short)ip->ip_len > ifp->if_mtu) {
584 error = EMSGSIZE;
585 goto bad;
586 }
587 m->m_flags |= M_BCAST;
588 } else {
589 m->m_flags &= ~M_BCAST;
590 }
591
592 sendit:
593 /*
594 * Force IP TTL to 255 following draft-ietf-zeroconf-ipv4-linklocal.txt
595 */
596 if (IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) || IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
597 ip_linklocal_stat.iplls_out_total++;
598 if (ip->ip_ttl != MAXTTL) {
599 ip_linklocal_stat.iplls_out_badttl++;
600 ip->ip_ttl = MAXTTL;
601 }
602 }
603
604 injectit:
605 if (!didfilter && !TAILQ_EMPTY(&ipv4_filters)) {
606 struct ipfilter *filter;
607 int seen = (inject_filter_ref == 0);
608
609 lck_mtx_unlock(ip_mutex);
610 ipf_ref();
611 TAILQ_FOREACH(filter, &ipv4_filters, ipf_link) {
612 if (seen == 0) {
613 if ((struct ipfilter *)inject_filter_ref == filter)
614 seen = 1;
615 } else if (filter->ipf_filter.ipf_output) {
616 errno_t result;
617 result = filter->ipf_filter.ipf_output(filter->ipf_filter.cookie, (mbuf_t*)&m, 0);
618 if (result == EJUSTRETURN) {
619 ipf_unref();
620 goto done;
621 }
622 if (result != 0) {
623 ipf_unref();
624 lck_mtx_lock(ip_mutex);
625 goto bad;
626 }
627 }
628 }
629 ipf_unref();
630 lck_mtx_lock(ip_mutex);
631 }
632
633 #if IPSEC
634 /* temporary for testing only: bypass ipsec alltogether */
635
636 if (ipsec_bypass != 0 || (flags & IP_NOIPSEC) != 0)
637 goto skip_ipsec;
638
639 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_START, 0,0,0,0,0);
640
641 lck_mtx_lock(sadb_mutex);
642
643 /* get SP for this packet */
644 if (so == NULL)
645 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
646 else
647 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
648
649 if (sp == NULL) {
650 ipsecstat.out_inval++;
651 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 0,0,0,0,0);
652 lck_mtx_unlock(sadb_mutex);
653 goto bad;
654 }
655
656 error = 0;
657
658 /* check policy */
659 switch (sp->policy) {
660 case IPSEC_POLICY_DISCARD:
661 /*
662 * This packet is just discarded.
663 */
664 ipsecstat.out_polvio++;
665 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 1,0,0,0,0);
666 lck_mtx_unlock(sadb_mutex);
667 goto bad;
668
669 case IPSEC_POLICY_BYPASS:
670 case IPSEC_POLICY_NONE:
671 /* no need to do IPsec. */
672 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 2,0,0,0,0);
673 lck_mtx_unlock(sadb_mutex);
674 goto skip_ipsec;
675
676 case IPSEC_POLICY_IPSEC:
677 if (sp->req == NULL) {
678 /* acquire a policy */
679 error = key_spdacquire(sp);
680 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 3,0,0,0,0);
681 lck_mtx_unlock(sadb_mutex);
682 goto bad;
683 }
684 break;
685
686 case IPSEC_POLICY_ENTRUST:
687 default:
688 printf("ip_output: Invalid policy found. %d\n", sp->policy);
689 }
690 {
691 struct ipsec_output_state state;
692 bzero(&state, sizeof(state));
693 state.m = m;
694 if (flags & IP_ROUTETOIF) {
695 state.ro = &iproute;
696 bzero(&iproute, sizeof(iproute));
697 } else
698 state.ro = ro;
699 state.dst = (struct sockaddr *)dst;
700
701 ip->ip_sum = 0;
702
703 /*
704 * XXX
705 * delayed checksums are not currently compatible with IPsec
706 */
707 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
708 in_delayed_cksum(m);
709 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
710 }
711
712 HTONS(ip->ip_len);
713 HTONS(ip->ip_off);
714
715 lck_mtx_unlock(ip_mutex);
716 error = ipsec4_output(&state, sp, flags);
717 lck_mtx_unlock(sadb_mutex);
718 lck_mtx_lock(ip_mutex);
719
720 m0 = m = state.m;
721
722 if (flags & IP_ROUTETOIF) {
723 /*
724 * if we have tunnel mode SA, we may need to ignore
725 * IP_ROUTETOIF.
726 */
727 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
728 flags &= ~IP_ROUTETOIF;
729 ro = state.ro;
730 }
731 } else
732 ro = state.ro;
733
734 dst = (struct sockaddr_in *)state.dst;
735 if (error) {
736 /* mbuf is already reclaimed in ipsec4_output. */
737 m0 = NULL;
738 switch (error) {
739 case EHOSTUNREACH:
740 case ENETUNREACH:
741 case EMSGSIZE:
742 case ENOBUFS:
743 case ENOMEM:
744 break;
745 default:
746 printf("ip4_output (ipsec): error code %d\n", error);
747 /*fall through*/
748 case ENOENT:
749 /* don't show these error codes to the user */
750 error = 0;
751 break;
752 }
753 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 4,0,0,0,0);
754 goto bad;
755 }
756 }
757
758 /* be sure to update variables that are affected by ipsec4_output() */
759 ip = mtod(m, struct ip *);
760
761 #ifdef _IP_VHL
762 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
763 #else
764 hlen = ip->ip_hl << 2;
765 #endif
766 /* Check that there wasn't a route change and src is still valid */
767
768 if (ro->ro_rt->generation_id != route_generation) {
769 if (ifa_foraddr(ip->ip_src.s_addr) == 0 && ((flags & (IP_ROUTETOIF | IP_FORWARDING)) == 0)) {
770 error = EADDRNOTAVAIL;
771 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 5,0,0,0,0);
772 goto bad;
773 }
774 rtfree(ro->ro_rt);
775 ro->ro_rt = NULL;
776 }
777
778 if (ro->ro_rt == NULL) {
779 if ((flags & IP_ROUTETOIF) == 0) {
780 printf("ip_output: "
781 "can't update route after IPsec processing\n");
782 error = EHOSTUNREACH; /*XXX*/
783 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 6,0,0,0,0);
784 goto bad;
785 }
786 } else {
787 if (ia)
788 ifafree(&ia->ia_ifa);
789 ia = ifatoia(ro->ro_rt->rt_ifa);
790 if (ia)
791 ifaref(&ia->ia_ifa);
792 ifp = ro->ro_rt->rt_ifp;
793 }
794
795 /* make it flipped, again. */
796 NTOHS(ip->ip_len);
797 NTOHS(ip->ip_off);
798 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 7,0xff,0xff,0xff,0xff);
799
800 /* Pass to filters again */
801 if (!TAILQ_EMPTY(&ipv4_filters)) {
802 struct ipfilter *filter;
803
804 lck_mtx_unlock(ip_mutex);
805 ipf_ref();
806 TAILQ_FOREACH(filter, &ipv4_filters, ipf_link) {
807 if (filter->ipf_filter.ipf_output) {
808 errno_t result;
809 result = filter->ipf_filter.ipf_output(filter->ipf_filter.cookie, (mbuf_t*)&m, 0);
810 if (result == EJUSTRETURN) {
811 ipf_unref();
812 goto done;
813 }
814 if (result != 0) {
815 ipf_unref();
816 lck_mtx_lock(ip_mutex);
817 goto bad;
818 }
819 }
820 }
821 ipf_unref();
822 lck_mtx_lock(ip_mutex);
823 }
824 skip_ipsec:
825 #endif /*IPSEC*/
826
827 /*
828 * IpHack's section.
829 * - Xlate: translate packet's addr/port (NAT).
830 * - Firewall: deny/allow/etc.
831 * - Wrap: fake packet's addr/port <unimpl.>
832 * - Encapsulate: put it in another IP and send out. <unimp.>
833 */
834 if (fr_checkp) {
835 struct mbuf *m1 = m;
836
837 if ((error = (*fr_checkp)(ip, hlen, ifp, 1, &m1)) || !m1) {
838 lck_mtx_unlock(ip_mutex);
839 goto done;
840 }
841 ip = mtod(m0 = m = m1, struct ip *);
842 }
843
844 /*
845 * Check with the firewall...
846 * but not if we are already being fwd'd from a firewall.
847 */
848 if (fw_enable && IPFW_LOADED && !args.next_hop) {
849 struct sockaddr_in *old = dst;
850
851 args.m = m;
852 args.next_hop = dst;
853 args.oif = ifp;
854 lck_mtx_unlock(ip_mutex);
855 off = ip_fw_chk_ptr(&args);
856 m = args.m;
857 dst = args.next_hop;
858
859 /*
860 * On return we must do the following:
861 * IP_FW_PORT_DENY_FLAG -> drop the pkt (XXX new)
862 * 1<=off<= 0xffff -> DIVERT
863 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe
864 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet
865 * dst != old -> IPFIREWALL_FORWARD
866 * off==0, dst==old -> accept
867 * If some of the above modules is not compiled in, then
868 * we should't have to check the corresponding condition
869 * (because the ipfw control socket should not accept
870 * unsupported rules), but better play safe and drop
871 * packets in case of doubt.
872 */
873 m0 = m;
874 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
875 if (m)
876 m_freem(m);
877 error = EACCES ;
878 goto done ;
879 }
880 ip = mtod(m, struct ip *);
881 if (off == 0 && dst == old) {/* common case */
882 lck_mtx_lock(ip_mutex);
883 goto pass ;
884 }
885 #if DUMMYNET
886 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) {
887 /*
888 * pass the pkt to dummynet. Need to include
889 * pipe number, m, ifp, ro, dst because these are
890 * not recomputed in the next pass.
891 * All other parameters have been already used and
892 * so they are not needed anymore.
893 * XXX note: if the ifp or ro entry are deleted
894 * while a pkt is in dummynet, we are in trouble!
895 */
896 args.ro = ro;
897 args.dst = dst;
898 args.flags = flags;
899
900 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT,
901 &args);
902 goto done;
903 }
904 #endif /* DUMMYNET */
905 lck_mtx_lock(ip_mutex);
906 #if IPDIVERT
907 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) {
908 struct mbuf *clone = NULL;
909
910 /* Clone packet if we're doing a 'tee' */
911 if ((off & IP_FW_PORT_TEE_FLAG) != 0)
912 clone = m_dup(m, M_DONTWAIT);
913 /*
914 * XXX
915 * delayed checksums are not currently compatible
916 * with divert sockets.
917 */
918 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
919 in_delayed_cksum(m);
920 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
921 }
922
923 /* Restore packet header fields to original values */
924 HTONS(ip->ip_len);
925 HTONS(ip->ip_off);
926
927 /* Deliver packet to divert input routine */
928 divert_packet(m, 0, off & 0xffff, args.divert_rule);
929
930 /* If 'tee', continue with original packet */
931 if (clone != NULL) {
932 m0 = m = clone;
933 ip = mtod(m, struct ip *);
934 goto pass;
935 }
936 lck_mtx_unlock(ip_mutex);
937 goto done;
938 }
939 #endif
940
941 #if IPFIREWALL_FORWARD
942 /* Here we check dst to make sure it's directly reachable on the
943 * interface we previously thought it was.
944 * If it isn't (which may be likely in some situations) we have
945 * to re-route it (ie, find a route for the next-hop and the
946 * associated interface) and set them here. This is nested
947 * forwarding which in most cases is undesirable, except where
948 * such control is nigh impossible. So we do it here.
949 * And I'm babbling.
950 */
951 if (off == 0 && old != dst) {
952 struct in_ifaddr *ia_fw;
953
954 /* It's changed... */
955 /* There must be a better way to do this next line... */
956 static struct route sro_fwd, *ro_fwd = &sro_fwd;
957 #if IPFIREWALL_FORWARD_DEBUG
958 printf("IPFIREWALL_FORWARD: New dst ip: ");
959 print_ip(dst->sin_addr);
960 printf("\n");
961 #endif
962 /*
963 * We need to figure out if we have been forwarded
964 * to a local socket. If so then we should somehow
965 * "loop back" to ip_input, and get directed to the
966 * PCB as if we had received this packet. This is
967 * because it may be dificult to identify the packets
968 * you want to forward until they are being output
969 * and have selected an interface. (e.g. locally
970 * initiated packets) If we used the loopback inteface,
971 * we would not be able to control what happens
972 * as the packet runs through ip_input() as
973 * it is done through a ISR.
974 */
975 TAILQ_FOREACH(ia_fw, &in_ifaddrhead, ia_link) {
976 /*
977 * If the addr to forward to is one
978 * of ours, we pretend to
979 * be the destination for this packet.
980 */
981 if (IA_SIN(ia_fw)->sin_addr.s_addr ==
982 dst->sin_addr.s_addr)
983 break;
984 }
985 if (ia) {
986 /* tell ip_input "dont filter" */
987 struct m_tag *fwd_tag;
988 struct ip_fwd_tag *ipfwd_tag;
989
990 fwd_tag = m_tag_alloc(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_IPFORWARD,
991 sizeof(struct sockaddr_in), M_NOWAIT);
992 if (fwd_tag == NULL) {
993 error = ENOBUFS;
994 goto bad;
995 }
996
997 ipfwd_tag = (struct ip_fwd_tag *)(fwd_tag+1);
998 ipfwd_tag->next_hop = args.next_hop;
999
1000 m_tag_prepend(m, fwd_tag);
1001
1002 if (m->m_pkthdr.rcvif == NULL)
1003 m->m_pkthdr.rcvif = ifunit("lo0");
1004 if ((~IF_HWASSIST_CSUM_FLAGS(m->m_pkthdr.rcvif->if_hwassist) &
1005 m->m_pkthdr.csum_flags) == 0) {
1006 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1007 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1008 m->m_pkthdr.csum_flags |=
1009 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1010 m->m_pkthdr.csum_data = 0xffff;
1011 }
1012 m->m_pkthdr.csum_flags |=
1013 CSUM_IP_CHECKED | CSUM_IP_VALID;
1014 }
1015 else if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1016 in_delayed_cksum(m);
1017 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1018 ip->ip_sum = in_cksum(m, hlen);
1019 }
1020 HTONS(ip->ip_len);
1021 HTONS(ip->ip_off);
1022
1023 lck_mtx_unlock(ip_mutex);
1024
1025 /* we need to call dlil_output to run filters
1026 * and resync to avoid recursion loops.
1027 */
1028 if (lo_ifp) {
1029 dlil_output(lo_ifp, PF_INET, m, 0, (struct sockaddr *)dst, 0);
1030 }
1031 else {
1032 printf("ip_output: no loopback ifp for forwarding!!!\n");
1033 }
1034 goto done;
1035 }
1036 /* Some of the logic for this was
1037 * nicked from above.
1038 *
1039 * This rewrites the cached route in a local PCB.
1040 * Is this what we want to do?
1041 */
1042 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst));
1043
1044 ro_fwd->ro_rt = 0;
1045 rtalloc_ign(ro_fwd, RTF_PRCLONING);
1046
1047 if (ro_fwd->ro_rt == 0) {
1048 ipstat.ips_noroute++;
1049 error = EHOSTUNREACH;
1050 goto bad;
1051 }
1052
1053 ia_fw = ifatoia(ro_fwd->ro_rt->rt_ifa);
1054 ifp = ro_fwd->ro_rt->rt_ifp;
1055 ro_fwd->ro_rt->rt_use++;
1056 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
1057 dst = (struct sockaddr_in *)ro_fwd->ro_rt->rt_gateway;
1058 if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
1059 isbroadcast =
1060 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
1061 else
1062 isbroadcast = in_broadcast(dst->sin_addr, ifp);
1063 rtfree(ro->ro_rt);
1064 ro->ro_rt = ro_fwd->ro_rt;
1065 dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
1066
1067 /*
1068 * If we added a default src ip earlier,
1069 * which would have been gotten from the-then
1070 * interface, do it again, from the new one.
1071 */
1072 if (fwd_rewrite_src)
1073 ip->ip_src = IA_SIN(ia_fw)->sin_addr;
1074 goto pass ;
1075 }
1076 #endif /* IPFIREWALL_FORWARD */
1077 /*
1078 * if we get here, none of the above matches, and
1079 * we have to drop the pkt
1080 */
1081 m_freem(m);
1082 error = EACCES; /* not sure this is the right error msg */
1083 lck_mtx_unlock(ip_mutex);
1084 goto done;
1085 }
1086
1087 pass:
1088 #if __APPLE__
1089 /* Do not allow loopback address to wind up on a wire */
1090 if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
1091 ((ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
1092 (ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)) {
1093 ipstat.ips_badaddr++;
1094 m_freem(m);
1095 /*
1096 * Do not simply drop the packet just like a firewall -- we want the
1097 * the application to feel the pain.
1098 * Return ENETUNREACH like ip6_output does in some similar cases.
1099 * This can startle the otherwise clueless process that specifies
1100 * loopback as the source address.
1101 */
1102 error = ENETUNREACH;
1103 lck_mtx_unlock(ip_mutex);
1104 goto done;
1105 }
1106 #endif
1107 m->m_pkthdr.csum_flags |= CSUM_IP;
1108 sw_csum = m->m_pkthdr.csum_flags
1109 & ~IF_HWASSIST_CSUM_FLAGS(ifp->if_hwassist);
1110
1111 if ((ifp->if_hwassist & CSUM_TCP_SUM16) != 0) {
1112 /*
1113 * Special case code for GMACE
1114 * frames that can be checksumed by GMACE SUM16 HW:
1115 * frame >64, no fragments, no UDP
1116 */
1117 if (apple_hwcksum_tx && (m->m_pkthdr.csum_flags & CSUM_TCP)
1118 && (ip->ip_len > 50) && (ip->ip_len <= ifp->if_mtu)) {
1119 /* Apple GMAC HW, expects STUFF_OFFSET << 16 | START_OFFSET */
1120 u_short offset = (IP_VHL_HL(ip->ip_vhl) << 2) +14 ; /* IP+Enet header length */
1121 u_short csumprev= m->m_pkthdr.csum_data & 0xFFFF;
1122 m->m_pkthdr.csum_flags = CSUM_DATA_VALID | CSUM_TCP_SUM16; /* for GMAC */
1123 m->m_pkthdr.csum_data = (csumprev + offset) << 16 ;
1124 m->m_pkthdr.csum_data += offset;
1125 sw_csum = CSUM_DELAY_IP; /* do IP hdr chksum in software */
1126 }
1127 else {
1128 /* let the software handle any UDP or TCP checksums */
1129 sw_csum |= (CSUM_DELAY_DATA & m->m_pkthdr.csum_flags);
1130 }
1131 }
1132
1133 if (sw_csum & CSUM_DELAY_DATA) {
1134 in_delayed_cksum(m);
1135 sw_csum &= ~CSUM_DELAY_DATA;
1136 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1137 }
1138
1139 m->m_pkthdr.csum_flags &= IF_HWASSIST_CSUM_FLAGS(ifp->if_hwassist);
1140
1141 /*
1142 * If small enough for interface, or the interface will take
1143 * care of the fragmentation for us, can just send directly.
1144 */
1145 if ((u_short)ip->ip_len <= ifp->if_mtu ||
1146 ifp->if_hwassist & CSUM_FRAGMENT) {
1147 HTONS(ip->ip_len);
1148 HTONS(ip->ip_off);
1149 ip->ip_sum = 0;
1150 if (sw_csum & CSUM_DELAY_IP) {
1151 ip->ip_sum = in_cksum(m, hlen);
1152 }
1153
1154 #ifndef __APPLE__
1155 /* Record statistics for this interface address. */
1156 if (!(flags & IP_FORWARDING) && ia != NULL) {
1157 ia->ia_ifa.if_opackets++;
1158 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1159 }
1160 #endif
1161
1162 #if IPSEC
1163 /* clean ipsec history once it goes out of the node */
1164 if (ipsec_bypass == 0 && (flags & IP_NOIPSEC) == 0)
1165 ipsec_delaux(m);
1166 #endif
1167 if (packetchain == 0) {
1168 lck_mtx_unlock(ip_mutex);
1169 error = dlil_output(ifp, PF_INET, m, (void *) ro->ro_rt,
1170 (struct sockaddr *)dst, 0);
1171 goto done;
1172 }
1173 else { /* packet chaining allows us to reuse the route for all packets */
1174 m = m->m_nextpkt;
1175 if (m == NULL) {
1176 if (pktcnt > ip_maxchainsent)
1177 ip_maxchainsent = pktcnt;
1178 //send
1179 lck_mtx_unlock(ip_mutex);
1180 error = dlil_output_list(ifp, PF_INET, packetlist, (void *) ro->ro_rt,
1181 (struct sockaddr *)dst, 0);
1182 pktcnt = 0;
1183 goto done;
1184
1185 }
1186 m0 = m;
1187 pktcnt++;
1188 goto loopit;
1189 }
1190 }
1191 /*
1192 * Too large for interface; fragment if possible.
1193 * Must be able to put at least 8 bytes per fragment.
1194 */
1195 if (ip->ip_off & IP_DF) {
1196 error = EMSGSIZE;
1197 /*
1198 * This case can happen if the user changed the MTU
1199 * of an interface after enabling IP on it. Because
1200 * most netifs don't keep track of routes pointing to
1201 * them, there is no way for one to update all its
1202 * routes when the MTU is changed.
1203 */
1204 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST))
1205 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU)
1206 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
1207 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
1208 }
1209 ipstat.ips_cantfrag++;
1210 goto bad;
1211 }
1212 len = (ifp->if_mtu - hlen) &~ 7;
1213 if (len < 8) {
1214 error = EMSGSIZE;
1215 goto bad;
1216 }
1217
1218 /*
1219 * if the interface will not calculate checksums on
1220 * fragmented packets, then do it here.
1221 */
1222 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
1223 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) {
1224 in_delayed_cksum(m);
1225 if (m == NULL) {
1226 lck_mtx_unlock(ip_mutex);
1227 return(ENOMEM);
1228 }
1229 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1230 }
1231
1232
1233 {
1234 int mhlen, firstlen = len;
1235 struct mbuf **mnext = &m->m_nextpkt;
1236 int nfrags = 1;
1237
1238 /*
1239 * Loop through length of segment after first fragment,
1240 * make new header and copy data of each part and link onto chain.
1241 */
1242 m0 = m;
1243 mhlen = sizeof (struct ip);
1244 for (off = hlen + len; off < (u_short)ip->ip_len; off += len) {
1245 MGETHDR(m, M_DONTWAIT, MT_HEADER);
1246 if (m == 0) {
1247 error = ENOBUFS;
1248 ipstat.ips_odropped++;
1249 goto sendorfree;
1250 }
1251 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
1252 m->m_data += max_linkhdr;
1253 mhip = mtod(m, struct ip *);
1254 *mhip = *ip;
1255 if (hlen > sizeof (struct ip)) {
1256 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1257 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
1258 }
1259 m->m_len = mhlen;
1260 mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
1261 if (ip->ip_off & IP_MF)
1262 mhip->ip_off |= IP_MF;
1263 if (off + len >= (u_short)ip->ip_len)
1264 len = (u_short)ip->ip_len - off;
1265 else
1266 mhip->ip_off |= IP_MF;
1267 mhip->ip_len = htons((u_short)(len + mhlen));
1268 m->m_next = m_copy(m0, off, len);
1269 if (m->m_next == 0) {
1270 (void) m_free(m);
1271 error = ENOBUFS; /* ??? */
1272 ipstat.ips_odropped++;
1273 goto sendorfree;
1274 }
1275 m->m_pkthdr.len = mhlen + len;
1276 m->m_pkthdr.rcvif = 0;
1277 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1278 m->m_pkthdr.socket_id = m0->m_pkthdr.socket_id;
1279 HTONS(mhip->ip_off);
1280 mhip->ip_sum = 0;
1281 if (sw_csum & CSUM_DELAY_IP) {
1282 mhip->ip_sum = in_cksum(m, mhlen);
1283 }
1284 *mnext = m;
1285 mnext = &m->m_nextpkt;
1286 nfrags++;
1287 }
1288 ipstat.ips_ofragments += nfrags;
1289
1290 /* set first/last markers for fragment chain */
1291 m->m_flags |= M_LASTFRAG;
1292 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1293 m0->m_pkthdr.csum_data = nfrags;
1294
1295 /*
1296 * Update first fragment by trimming what's been copied out
1297 * and updating header, then send each fragment (in order).
1298 */
1299 m = m0;
1300 m_adj(m, hlen + firstlen - (u_short)ip->ip_len);
1301 m->m_pkthdr.len = hlen + firstlen;
1302 ip->ip_len = htons((u_short)m->m_pkthdr.len);
1303 ip->ip_off |= IP_MF;
1304 HTONS(ip->ip_off);
1305 ip->ip_sum = 0;
1306 if (sw_csum & CSUM_DELAY_IP) {
1307 ip->ip_sum = in_cksum(m, hlen);
1308 }
1309 sendorfree:
1310
1311 KERNEL_DEBUG(DBG_LAYER_END, ip->ip_dst.s_addr,
1312 ip->ip_src.s_addr, ip->ip_p, ip->ip_off, ip->ip_len);
1313
1314 lck_mtx_unlock(ip_mutex);
1315 for (m = m0; m; m = m0) {
1316 m0 = m->m_nextpkt;
1317 m->m_nextpkt = 0;
1318 #if IPSEC
1319 /* clean ipsec history once it goes out of the node */
1320 if (ipsec_bypass == 0 && (flags & IP_NOIPSEC) == 0)
1321 ipsec_delaux(m);
1322 #endif
1323 if (error == 0) {
1324 #ifndef __APPLE__
1325 /* Record statistics for this interface address. */
1326 if (ia != NULL) {
1327 ia->ia_ifa.if_opackets++;
1328 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1329 }
1330 #endif
1331 if ((packetchain != 0) && (pktcnt > 0))
1332 panic("ip_output: mix of packet in packetlist is wrong=%x", packetlist);
1333 error = dlil_output(ifp, PF_INET, m, (void *) ro->ro_rt,
1334 (struct sockaddr *)dst, 0);
1335 } else
1336 m_freem(m);
1337 }
1338
1339 if (error == 0)
1340 ipstat.ips_fragmented++;
1341 }
1342 done:
1343 if (ia) {
1344 ifafree(&ia->ia_ifa);
1345 ia = NULL;
1346 }
1347 #if IPSEC
1348 if (ipsec_bypass == 0 && (flags & IP_NOIPSEC) == 0) {
1349 if (ro == &iproute && ro->ro_rt) {
1350 rtfree(ro->ro_rt);
1351 ro->ro_rt = NULL;
1352 }
1353 if (sp != NULL) {
1354 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1355 printf("DP ip_output call free SP:%x\n", sp));
1356 lck_mtx_lock(sadb_mutex);
1357 key_freesp(sp);
1358 lck_mtx_unlock(sadb_mutex);
1359 }
1360 }
1361 #endif /* IPSEC */
1362
1363 KERNEL_DEBUG(DBG_FNC_IP_OUTPUT | DBG_FUNC_END, error,0,0,0,0);
1364 return (error);
1365 bad:
1366 m_freem(m0);
1367 lck_mtx_unlock(ip_mutex);
1368 goto done;
1369 }
1370
1371 void
1372 in_delayed_cksum_offset(struct mbuf *m, int ip_offset)
1373 {
1374 struct ip *ip;
1375 u_short csum, offset;
1376
1377 while (ip_offset > m->m_len) {
1378 ip_offset -= m->m_len;
1379 m = m->m_next;
1380 if (m) {
1381 printf("in_delayed_cksum_withoffset failed - ip_offset wasn't in the packet\n");
1382 return;
1383 }
1384 }
1385
1386 if (ip_offset + sizeof(struct ip) > m->m_len) {
1387 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1388 m->m_len, ip_offset, ip->ip_p);
1389 /*
1390 * XXX
1391 * this shouldn't happen
1392 */
1393 m = m_pullup(m, ip_offset + sizeof(struct ip));
1394 }
1395
1396 /* Gross */
1397 if (ip_offset) {
1398 m->m_len -= ip_offset;
1399 m->m_data += ip_offset;
1400 }
1401
1402 ip = mtod(m, struct ip*);
1403 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1404 csum = in_cksum_skip(m, ip->ip_len, offset);
1405 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1406 csum = 0xffff;
1407 offset += m->m_pkthdr.csum_data & 0xFFFF; /* checksum offset */
1408
1409 /* Gross */
1410 if (ip_offset) {
1411 if (M_LEADINGSPACE(m) < ip_offset)
1412 panic("in_delayed_cksum_withoffset - chain modified!\n");
1413 m->m_len += ip_offset;
1414 m->m_data -= ip_offset;
1415 }
1416
1417 if (offset > ip->ip_len) /* bogus offset */
1418 return;
1419
1420 if (offset + ip_offset + sizeof(u_short) > m->m_len) {
1421 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1422 m->m_len, offset + ip_offset, ip->ip_p);
1423 /*
1424 * XXX
1425 * this shouldn't happen, but if it does, the
1426 * correct behavior may be to insert the checksum
1427 * in the existing chain instead of rearranging it.
1428 */
1429 m = m_pullup(m, offset + ip_offset + sizeof(u_short));
1430 }
1431 *(u_short *)(m->m_data + offset + ip_offset) = csum;
1432 }
1433
1434 void
1435 in_delayed_cksum(struct mbuf *m)
1436 {
1437 in_delayed_cksum_offset(m, 0);
1438 }
1439
1440 void
1441 in_cksum_offset(struct mbuf* m, size_t ip_offset)
1442 {
1443 struct ip* ip = NULL;
1444 int hlen = 0;
1445
1446 while (ip_offset > m->m_len) {
1447 ip_offset -= m->m_len;
1448 m = m->m_next;
1449 if (m) {
1450 printf("in_cksum_offset failed - ip_offset wasn't in the packet\n");
1451 return;
1452 }
1453 }
1454
1455 if (ip_offset + sizeof(struct ip) > m->m_len) {
1456 printf("in_cksum_offset - delayed m_pullup, m->len: %d off: %d\n",
1457 m->m_len, ip_offset);
1458 /*
1459 * XXX
1460 * this shouldn't happen
1461 */
1462 m = m_pullup(m, ip_offset + sizeof(struct ip));
1463 }
1464
1465 /* Gross */
1466 if (ip_offset) {
1467 m->m_len -= ip_offset;
1468 m->m_data += ip_offset;
1469 }
1470
1471 ip = mtod(m, struct ip*);
1472
1473 #ifdef _IP_VHL
1474 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1475 #else
1476 hlen = ip->ip_hl << 2;
1477 #endif
1478
1479 ip->ip_sum = 0;
1480 ip->ip_sum = in_cksum(m, hlen);
1481
1482 /* Gross */
1483 if (ip_offset) {
1484 if (M_LEADINGSPACE(m) < ip_offset)
1485 panic("in_cksum_offset - chain modified!\n");
1486 m->m_len += ip_offset;
1487 m->m_data -= ip_offset;
1488 }
1489 }
1490
1491 /*
1492 * Insert IP options into preformed packet.
1493 * Adjust IP destination as required for IP source routing,
1494 * as indicated by a non-zero in_addr at the start of the options.
1495 *
1496 * XXX This routine assumes that the packet has no options in place.
1497 */
1498 static struct mbuf *
1499 ip_insertoptions(m, opt, phlen)
1500 register struct mbuf *m;
1501 struct mbuf *opt;
1502 int *phlen;
1503 {
1504 register struct ipoption *p = mtod(opt, struct ipoption *);
1505 struct mbuf *n;
1506 register struct ip *ip = mtod(m, struct ip *);
1507 unsigned optlen;
1508
1509 optlen = opt->m_len - sizeof(p->ipopt_dst);
1510 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET)
1511 return (m); /* XXX should fail */
1512 if (p->ipopt_dst.s_addr)
1513 ip->ip_dst = p->ipopt_dst;
1514 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1515 MGETHDR(n, M_DONTWAIT, MT_HEADER);
1516 if (n == 0)
1517 return (m);
1518 n->m_pkthdr.rcvif = 0;
1519 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1520 m->m_len -= sizeof(struct ip);
1521 m->m_data += sizeof(struct ip);
1522 n->m_next = m;
1523 m = n;
1524 m->m_len = optlen + sizeof(struct ip);
1525 m->m_data += max_linkhdr;
1526 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip));
1527 } else {
1528 m->m_data -= optlen;
1529 m->m_len += optlen;
1530 m->m_pkthdr.len += optlen;
1531 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1532 }
1533 ip = mtod(m, struct ip *);
1534 bcopy(p->ipopt_list, ip + 1, optlen);
1535 *phlen = sizeof(struct ip) + optlen;
1536 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1537 ip->ip_len += optlen;
1538 return (m);
1539 }
1540
1541 /*
1542 * Copy options from ip to jp,
1543 * omitting those not copied during fragmentation.
1544 */
1545 int
1546 ip_optcopy(ip, jp)
1547 struct ip *ip, *jp;
1548 {
1549 register u_char *cp, *dp;
1550 int opt, optlen, cnt;
1551
1552 cp = (u_char *)(ip + 1);
1553 dp = (u_char *)(jp + 1);
1554 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1555 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1556 opt = cp[0];
1557 if (opt == IPOPT_EOL)
1558 break;
1559 if (opt == IPOPT_NOP) {
1560 /* Preserve for IP mcast tunnel's LSRR alignment. */
1561 *dp++ = IPOPT_NOP;
1562 optlen = 1;
1563 continue;
1564 }
1565 #if DIAGNOSTIC
1566 if (cnt < IPOPT_OLEN + sizeof(*cp))
1567 panic("malformed IPv4 option passed to ip_optcopy");
1568 #endif
1569 optlen = cp[IPOPT_OLEN];
1570 #if DIAGNOSTIC
1571 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1572 panic("malformed IPv4 option passed to ip_optcopy");
1573 #endif
1574 /* bogus lengths should have been caught by ip_dooptions */
1575 if (optlen > cnt)
1576 optlen = cnt;
1577 if (IPOPT_COPIED(opt)) {
1578 bcopy(cp, dp, optlen);
1579 dp += optlen;
1580 }
1581 }
1582 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1583 *dp++ = IPOPT_EOL;
1584 return (optlen);
1585 }
1586
1587 /*
1588 * IP socket option processing.
1589 */
1590 int
1591 ip_ctloutput(so, sopt)
1592 struct socket *so;
1593 struct sockopt *sopt;
1594 {
1595 struct inpcb *inp = sotoinpcb(so);
1596 int error, optval;
1597
1598 error = optval = 0;
1599 if (sopt->sopt_level != IPPROTO_IP) {
1600 return (EINVAL);
1601 }
1602
1603 switch (sopt->sopt_dir) {
1604 case SOPT_SET:
1605 switch (sopt->sopt_name) {
1606 case IP_OPTIONS:
1607 #ifdef notyet
1608 case IP_RETOPTS:
1609 #endif
1610 {
1611 struct mbuf *m;
1612 if (sopt->sopt_valsize > MLEN) {
1613 error = EMSGSIZE;
1614 break;
1615 }
1616 MGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT, MT_HEADER);
1617 if (m == 0) {
1618 error = ENOBUFS;
1619 break;
1620 }
1621 m->m_len = sopt->sopt_valsize;
1622 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1623 m->m_len);
1624 if (error)
1625 break;
1626
1627 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1628 m));
1629 }
1630
1631 case IP_TOS:
1632 case IP_TTL:
1633 case IP_RECVOPTS:
1634 case IP_RECVRETOPTS:
1635 case IP_RECVDSTADDR:
1636 case IP_RECVIF:
1637 case IP_RECVTTL:
1638 #if defined(NFAITH) && NFAITH > 0
1639 case IP_FAITH:
1640 #endif
1641 error = sooptcopyin(sopt, &optval, sizeof optval,
1642 sizeof optval);
1643 if (error)
1644 break;
1645
1646 switch (sopt->sopt_name) {
1647 case IP_TOS:
1648 inp->inp_ip_tos = optval;
1649 break;
1650
1651 case IP_TTL:
1652 inp->inp_ip_ttl = optval;
1653 break;
1654 #define OPTSET(bit) \
1655 if (optval) \
1656 inp->inp_flags |= bit; \
1657 else \
1658 inp->inp_flags &= ~bit;
1659
1660 case IP_RECVOPTS:
1661 OPTSET(INP_RECVOPTS);
1662 break;
1663
1664 case IP_RECVRETOPTS:
1665 OPTSET(INP_RECVRETOPTS);
1666 break;
1667
1668 case IP_RECVDSTADDR:
1669 OPTSET(INP_RECVDSTADDR);
1670 break;
1671
1672 case IP_RECVIF:
1673 OPTSET(INP_RECVIF);
1674 break;
1675
1676 case IP_RECVTTL:
1677 OPTSET(INP_RECVTTL);
1678 break;
1679
1680 #if defined(NFAITH) && NFAITH > 0
1681 case IP_FAITH:
1682 OPTSET(INP_FAITH);
1683 break;
1684 #endif
1685 }
1686 break;
1687 #undef OPTSET
1688
1689 case IP_MULTICAST_IF:
1690 case IP_MULTICAST_VIF:
1691 case IP_MULTICAST_TTL:
1692 case IP_MULTICAST_LOOP:
1693 case IP_ADD_MEMBERSHIP:
1694 case IP_DROP_MEMBERSHIP:
1695 error = ip_setmoptions(sopt, &inp->inp_moptions);
1696 break;
1697
1698 case IP_PORTRANGE:
1699 error = sooptcopyin(sopt, &optval, sizeof optval,
1700 sizeof optval);
1701 if (error)
1702 break;
1703
1704 switch (optval) {
1705 case IP_PORTRANGE_DEFAULT:
1706 inp->inp_flags &= ~(INP_LOWPORT);
1707 inp->inp_flags &= ~(INP_HIGHPORT);
1708 break;
1709
1710 case IP_PORTRANGE_HIGH:
1711 inp->inp_flags &= ~(INP_LOWPORT);
1712 inp->inp_flags |= INP_HIGHPORT;
1713 break;
1714
1715 case IP_PORTRANGE_LOW:
1716 inp->inp_flags &= ~(INP_HIGHPORT);
1717 inp->inp_flags |= INP_LOWPORT;
1718 break;
1719
1720 default:
1721 error = EINVAL;
1722 break;
1723 }
1724 break;
1725
1726 #if IPSEC
1727 case IP_IPSEC_POLICY:
1728 {
1729 caddr_t req = NULL;
1730 size_t len = 0;
1731 int priv;
1732 struct mbuf *m;
1733 int optname;
1734
1735 if (sopt->sopt_valsize > MCLBYTES) {
1736 error = EMSGSIZE;
1737 break;
1738 }
1739 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1740 break;
1741 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1742 break;
1743 priv = (sopt->sopt_p != NULL &&
1744 proc_suser(sopt->sopt_p) != 0) ? 0 : 1;
1745 if (m) {
1746 req = mtod(m, caddr_t);
1747 len = m->m_len;
1748 }
1749 optname = sopt->sopt_name;
1750 lck_mtx_lock(sadb_mutex);
1751 error = ipsec4_set_policy(inp, optname, req, len, priv);
1752 lck_mtx_unlock(sadb_mutex);
1753 m_freem(m);
1754 break;
1755 }
1756 #endif /*IPSEC*/
1757
1758 default:
1759 error = ENOPROTOOPT;
1760 break;
1761 }
1762 break;
1763
1764 case SOPT_GET:
1765 switch (sopt->sopt_name) {
1766 case IP_OPTIONS:
1767 case IP_RETOPTS:
1768 if (inp->inp_options)
1769 error = sooptcopyout(sopt,
1770 mtod(inp->inp_options,
1771 char *),
1772 inp->inp_options->m_len);
1773 else
1774 sopt->sopt_valsize = 0;
1775 break;
1776
1777 case IP_TOS:
1778 case IP_TTL:
1779 case IP_RECVOPTS:
1780 case IP_RECVRETOPTS:
1781 case IP_RECVDSTADDR:
1782 case IP_RECVIF:
1783 case IP_RECVTTL:
1784 case IP_PORTRANGE:
1785 #if defined(NFAITH) && NFAITH > 0
1786 case IP_FAITH:
1787 #endif
1788 switch (sopt->sopt_name) {
1789
1790 case IP_TOS:
1791 optval = inp->inp_ip_tos;
1792 break;
1793
1794 case IP_TTL:
1795 optval = inp->inp_ip_ttl;
1796 break;
1797
1798 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1799
1800 case IP_RECVOPTS:
1801 optval = OPTBIT(INP_RECVOPTS);
1802 break;
1803
1804 case IP_RECVRETOPTS:
1805 optval = OPTBIT(INP_RECVRETOPTS);
1806 break;
1807
1808 case IP_RECVDSTADDR:
1809 optval = OPTBIT(INP_RECVDSTADDR);
1810 break;
1811
1812 case IP_RECVIF:
1813 optval = OPTBIT(INP_RECVIF);
1814 break;
1815
1816 case IP_RECVTTL:
1817 optval = OPTBIT(INP_RECVTTL);
1818 break;
1819
1820 case IP_PORTRANGE:
1821 if (inp->inp_flags & INP_HIGHPORT)
1822 optval = IP_PORTRANGE_HIGH;
1823 else if (inp->inp_flags & INP_LOWPORT)
1824 optval = IP_PORTRANGE_LOW;
1825 else
1826 optval = 0;
1827 break;
1828
1829 #if defined(NFAITH) && NFAITH > 0
1830 case IP_FAITH:
1831 optval = OPTBIT(INP_FAITH);
1832 break;
1833 #endif
1834 }
1835 error = sooptcopyout(sopt, &optval, sizeof optval);
1836 break;
1837
1838 case IP_MULTICAST_IF:
1839 case IP_MULTICAST_VIF:
1840 case IP_MULTICAST_TTL:
1841 case IP_MULTICAST_LOOP:
1842 case IP_ADD_MEMBERSHIP:
1843 case IP_DROP_MEMBERSHIP:
1844 error = ip_getmoptions(sopt, inp->inp_moptions);
1845 break;
1846
1847 #if IPSEC
1848 case IP_IPSEC_POLICY:
1849 {
1850 struct mbuf *m = NULL;
1851 caddr_t req = NULL;
1852 size_t len = 0;
1853
1854 if (m != 0) {
1855 req = mtod(m, caddr_t);
1856 len = m->m_len;
1857 }
1858 lck_mtx_lock(sadb_mutex);
1859 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m);
1860 lck_mtx_unlock(sadb_mutex);
1861 if (error == 0)
1862 error = soopt_mcopyout(sopt, m); /* XXX */
1863 if (error == 0)
1864 m_freem(m);
1865 break;
1866 }
1867 #endif /*IPSEC*/
1868
1869 default:
1870 error = ENOPROTOOPT;
1871 break;
1872 }
1873 break;
1874 }
1875 return (error);
1876 }
1877
1878 /*
1879 * Set up IP options in pcb for insertion in output packets.
1880 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1881 * with destination address if source routed.
1882 */
1883 static int
1884 ip_pcbopts(optname, pcbopt, m)
1885 int optname;
1886 struct mbuf **pcbopt;
1887 register struct mbuf *m;
1888 {
1889 register int cnt, optlen;
1890 register u_char *cp;
1891 u_char opt;
1892
1893 /* turn off any old options */
1894 if (*pcbopt)
1895 (void)m_free(*pcbopt);
1896 *pcbopt = 0;
1897 if (m == (struct mbuf *)0 || m->m_len == 0) {
1898 /*
1899 * Only turning off any previous options.
1900 */
1901 if (m)
1902 (void)m_free(m);
1903 return (0);
1904 }
1905
1906 #ifndef vax
1907 if (m->m_len % sizeof(int32_t))
1908 goto bad;
1909 #endif
1910 /*
1911 * IP first-hop destination address will be stored before
1912 * actual options; move other options back
1913 * and clear it when none present.
1914 */
1915 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1916 goto bad;
1917 cnt = m->m_len;
1918 m->m_len += sizeof(struct in_addr);
1919 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1920 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
1921 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1922
1923 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1924 opt = cp[IPOPT_OPTVAL];
1925 if (opt == IPOPT_EOL)
1926 break;
1927 if (opt == IPOPT_NOP)
1928 optlen = 1;
1929 else {
1930 if (cnt < IPOPT_OLEN + sizeof(*cp))
1931 goto bad;
1932 optlen = cp[IPOPT_OLEN];
1933 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1934 goto bad;
1935 }
1936 switch (opt) {
1937
1938 default:
1939 break;
1940
1941 case IPOPT_LSRR:
1942 case IPOPT_SSRR:
1943 /*
1944 * user process specifies route as:
1945 * ->A->B->C->D
1946 * D must be our final destination (but we can't
1947 * check that since we may not have connected yet).
1948 * A is first hop destination, which doesn't appear in
1949 * actual IP option, but is stored before the options.
1950 */
1951 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1952 goto bad;
1953 m->m_len -= sizeof(struct in_addr);
1954 cnt -= sizeof(struct in_addr);
1955 optlen -= sizeof(struct in_addr);
1956 cp[IPOPT_OLEN] = optlen;
1957 /*
1958 * Move first hop before start of options.
1959 */
1960 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1961 sizeof(struct in_addr));
1962 /*
1963 * Then copy rest of options back
1964 * to close up the deleted entry.
1965 */
1966 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1967 sizeof(struct in_addr)),
1968 (caddr_t)&cp[IPOPT_OFFSET+1],
1969 (unsigned)cnt + sizeof(struct in_addr));
1970 break;
1971 }
1972 }
1973 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1974 goto bad;
1975 *pcbopt = m;
1976 return (0);
1977
1978 bad:
1979 (void)m_free(m);
1980 return (EINVAL);
1981 }
1982
1983 /*
1984 * XXX
1985 * The whole multicast option thing needs to be re-thought.
1986 * Several of these options are equally applicable to non-multicast
1987 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1988 * standard option (IP_TTL).
1989 */
1990
1991 /*
1992 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1993 */
1994 static struct ifnet *
1995 ip_multicast_if(a, ifindexp)
1996 struct in_addr *a;
1997 int *ifindexp;
1998 {
1999 int ifindex;
2000 struct ifnet *ifp;
2001
2002 if (ifindexp)
2003 *ifindexp = 0;
2004 if (ntohl(a->s_addr) >> 24 == 0) {
2005 ifindex = ntohl(a->s_addr) & 0xffffff;
2006 ifnet_head_lock_shared();
2007 if (ifindex < 0 || if_index < ifindex) {
2008 ifnet_head_done();
2009 return NULL;
2010 }
2011 ifp = ifindex2ifnet[ifindex];
2012 ifnet_head_done();
2013 if (ifindexp)
2014 *ifindexp = ifindex;
2015 } else {
2016 INADDR_TO_IFP(*a, ifp);
2017 }
2018 return ifp;
2019 }
2020
2021 /*
2022 * Set the IP multicast options in response to user setsockopt().
2023 */
2024 static int
2025 ip_setmoptions(sopt, imop)
2026 struct sockopt *sopt;
2027 struct ip_moptions **imop;
2028 {
2029 int error = 0;
2030 int i;
2031 struct in_addr addr;
2032 struct ip_mreq mreq;
2033 struct ifnet *ifp = NULL;
2034 struct ip_moptions *imo = *imop;
2035 int ifindex;
2036
2037 if (imo == NULL) {
2038 /*
2039 * No multicast option buffer attached to the pcb;
2040 * allocate one and initialize to default values.
2041 */
2042 error = ip_createmoptions(imop);
2043 if (error != 0)
2044 return error;
2045 imo = *imop;
2046 }
2047
2048 switch (sopt->sopt_name) {
2049 /* store an index number for the vif you wanna use in the send */
2050 case IP_MULTICAST_VIF:
2051 if (legal_vif_num == 0) {
2052 error = EOPNOTSUPP;
2053 break;
2054 }
2055 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
2056 if (error)
2057 break;
2058 if (!legal_vif_num(i) && (i != -1)) {
2059 error = EINVAL;
2060 break;
2061 }
2062 imo->imo_multicast_vif = i;
2063 break;
2064
2065 case IP_MULTICAST_IF:
2066 /*
2067 * Select the interface for outgoing multicast packets.
2068 */
2069 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr);
2070 if (error)
2071 break;
2072 /*
2073 * INADDR_ANY is used to remove a previous selection.
2074 * When no interface is selected, a default one is
2075 * chosen every time a multicast packet is sent.
2076 */
2077 if (addr.s_addr == INADDR_ANY) {
2078 imo->imo_multicast_ifp = NULL;
2079 break;
2080 }
2081 /*
2082 * The selected interface is identified by its local
2083 * IP address. Find the interface and confirm that
2084 * it supports multicasting.
2085 */
2086 ifp = ip_multicast_if(&addr, &ifindex);
2087 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2088 error = EADDRNOTAVAIL;
2089 break;
2090 }
2091 imo->imo_multicast_ifp = ifp;
2092 if (ifindex)
2093 imo->imo_multicast_addr = addr;
2094 else
2095 imo->imo_multicast_addr.s_addr = INADDR_ANY;
2096 break;
2097
2098 case IP_MULTICAST_TTL:
2099 /*
2100 * Set the IP time-to-live for outgoing multicast packets.
2101 * The original multicast API required a char argument,
2102 * which is inconsistent with the rest of the socket API.
2103 * We allow either a char or an int.
2104 */
2105 if (sopt->sopt_valsize == 1) {
2106 u_char ttl;
2107 error = sooptcopyin(sopt, &ttl, 1, 1);
2108 if (error)
2109 break;
2110 imo->imo_multicast_ttl = ttl;
2111 } else {
2112 u_int ttl;
2113 error = sooptcopyin(sopt, &ttl, sizeof ttl,
2114 sizeof ttl);
2115 if (error)
2116 break;
2117 if (ttl > 255)
2118 error = EINVAL;
2119 else
2120 imo->imo_multicast_ttl = ttl;
2121 }
2122 break;
2123
2124 case IP_MULTICAST_LOOP:
2125 /*
2126 * Set the loopback flag for outgoing multicast packets.
2127 * Must be zero or one. The original multicast API required a
2128 * char argument, which is inconsistent with the rest
2129 * of the socket API. We allow either a char or an int.
2130 */
2131 if (sopt->sopt_valsize == 1) {
2132 u_char loop;
2133 error = sooptcopyin(sopt, &loop, 1, 1);
2134 if (error)
2135 break;
2136 imo->imo_multicast_loop = !!loop;
2137 } else {
2138 u_int loop;
2139 error = sooptcopyin(sopt, &loop, sizeof loop,
2140 sizeof loop);
2141 if (error)
2142 break;
2143 imo->imo_multicast_loop = !!loop;
2144 }
2145 break;
2146
2147 case IP_ADD_MEMBERSHIP:
2148 /*
2149 * Add a multicast group membership.
2150 * Group must be a valid IP multicast address.
2151 */
2152 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
2153 if (error)
2154 break;
2155
2156 error = ip_addmembership(imo, &mreq);
2157 break;
2158
2159 case IP_DROP_MEMBERSHIP:
2160 /*
2161 * Drop a multicast group membership.
2162 * Group must be a valid IP multicast address.
2163 */
2164 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
2165 if (error)
2166 break;
2167
2168 error = ip_dropmembership(imo, &mreq);
2169 break;
2170
2171 default:
2172 error = EOPNOTSUPP;
2173 break;
2174 }
2175
2176 /*
2177 * If all options have default values, no need to keep the mbuf.
2178 */
2179 if (imo->imo_multicast_ifp == NULL &&
2180 imo->imo_multicast_vif == -1 &&
2181 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2182 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2183 imo->imo_num_memberships == 0) {
2184 FREE(*imop, M_IPMOPTS);
2185 *imop = NULL;
2186 }
2187
2188 return (error);
2189 }
2190
2191 /*
2192 * Set the IP multicast options in response to user setsockopt().
2193 */
2194 __private_extern__ int
2195 ip_createmoptions(
2196 struct ip_moptions **imop)
2197 {
2198 struct ip_moptions *imo;
2199 imo = (struct ip_moptions*) _MALLOC(sizeof(*imo), M_IPMOPTS,
2200 M_WAITOK);
2201
2202 if (imo == NULL)
2203 return (ENOBUFS);
2204 *imop = imo;
2205 imo->imo_multicast_ifp = NULL;
2206 imo->imo_multicast_addr.s_addr = INADDR_ANY;
2207 imo->imo_multicast_vif = -1;
2208 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
2209 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
2210 imo->imo_num_memberships = 0;
2211
2212 return 0;
2213 }
2214
2215 /*
2216 * Add membership to an IPv4 multicast.
2217 */
2218 __private_extern__ int
2219 ip_addmembership(
2220 struct ip_moptions *imo,
2221 struct ip_mreq *mreq)
2222 {
2223 struct route ro;
2224 struct sockaddr_in *dst;
2225 struct ifnet *ifp = NULL;
2226 int error = 0;
2227 int i;
2228
2229 if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
2230 error = EINVAL;
2231 return error;
2232 }
2233 /*
2234 * If no interface address was provided, use the interface of
2235 * the route to the given multicast address.
2236 */
2237 if (mreq->imr_interface.s_addr == INADDR_ANY) {
2238 bzero((caddr_t)&ro, sizeof(ro));
2239 dst = (struct sockaddr_in *)&ro.ro_dst;
2240 dst->sin_len = sizeof(*dst);
2241 dst->sin_family = AF_INET;
2242 dst->sin_addr = mreq->imr_multiaddr;
2243 rtalloc(&ro);
2244 if (ro.ro_rt != NULL) {
2245 ifp = ro.ro_rt->rt_ifp;
2246 rtfree(ro.ro_rt);
2247 }
2248 else {
2249 /* If there's no default route, try using loopback */
2250 mreq->imr_interface.s_addr = INADDR_LOOPBACK;
2251 }
2252 }
2253
2254 if (ifp == NULL) {
2255 ifp = ip_multicast_if(&mreq->imr_interface, NULL);
2256 }
2257
2258 /*
2259 * See if we found an interface, and confirm that it
2260 * supports multicast.
2261 */
2262 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2263 error = EADDRNOTAVAIL;
2264 return error;
2265 }
2266 /*
2267 * See if the membership already exists or if all the
2268 * membership slots are full.
2269 */
2270 for (i = 0; i < imo->imo_num_memberships; ++i) {
2271 if (imo->imo_membership[i]->inm_ifp == ifp &&
2272 imo->imo_membership[i]->inm_addr.s_addr
2273 == mreq->imr_multiaddr.s_addr)
2274 break;
2275 }
2276 if (i < imo->imo_num_memberships) {
2277 error = EADDRINUSE;
2278 return error;
2279 }
2280 if (i == IP_MAX_MEMBERSHIPS) {
2281 error = ETOOMANYREFS;
2282 return error;
2283 }
2284 /*
2285 * Everything looks good; add a new record to the multicast
2286 * address list for the given interface.
2287 */
2288 if ((imo->imo_membership[i] =
2289 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
2290 error = ENOBUFS;
2291 return error;
2292 }
2293 ++imo->imo_num_memberships;
2294
2295 return error;
2296 }
2297
2298 /*
2299 * Drop membership of an IPv4 multicast.
2300 */
2301 __private_extern__ int
2302 ip_dropmembership(
2303 struct ip_moptions *imo,
2304 struct ip_mreq *mreq)
2305 {
2306 int error = 0;
2307 struct ifnet* ifp = NULL;
2308 int i;
2309
2310 if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
2311 error = EINVAL;
2312 return error;
2313 }
2314
2315 /*
2316 * If an interface address was specified, get a pointer
2317 * to its ifnet structure.
2318 */
2319 if (mreq->imr_interface.s_addr == INADDR_ANY)
2320 ifp = NULL;
2321 else {
2322 ifp = ip_multicast_if(&mreq->imr_interface, NULL);
2323 if (ifp == NULL) {
2324 error = EADDRNOTAVAIL;
2325 return error;
2326 }
2327 }
2328 /*
2329 * Find the membership in the membership array.
2330 */
2331 for (i = 0; i < imo->imo_num_memberships; ++i) {
2332 if ((ifp == NULL ||
2333 imo->imo_membership[i]->inm_ifp == ifp) &&
2334 imo->imo_membership[i]->inm_addr.s_addr ==
2335 mreq->imr_multiaddr.s_addr)
2336 break;
2337 }
2338 if (i == imo->imo_num_memberships) {
2339 error = EADDRNOTAVAIL;
2340 return error;
2341 }
2342 /*
2343 * Give up the multicast address record to which the
2344 * membership points.
2345 */
2346 in_delmulti(&imo->imo_membership[i]);
2347 /*
2348 * Remove the gap in the membership array.
2349 */
2350 for (++i; i < imo->imo_num_memberships; ++i)
2351 imo->imo_membership[i-1] = imo->imo_membership[i];
2352 --imo->imo_num_memberships;
2353
2354 return error;
2355 }
2356
2357 /*
2358 * Return the IP multicast options in response to user getsockopt().
2359 */
2360 static int
2361 ip_getmoptions(sopt, imo)
2362 struct sockopt *sopt;
2363 register struct ip_moptions *imo;
2364 {
2365 struct in_addr addr;
2366 struct in_ifaddr *ia;
2367 int error, optval;
2368 u_char coptval;
2369
2370 error = 0;
2371 switch (sopt->sopt_name) {
2372 case IP_MULTICAST_VIF:
2373 if (imo != NULL)
2374 optval = imo->imo_multicast_vif;
2375 else
2376 optval = -1;
2377 error = sooptcopyout(sopt, &optval, sizeof optval);
2378 break;
2379
2380 case IP_MULTICAST_IF:
2381 if (imo == NULL || imo->imo_multicast_ifp == NULL)
2382 addr.s_addr = INADDR_ANY;
2383 else if (imo->imo_multicast_addr.s_addr) {
2384 /* return the value user has set */
2385 addr = imo->imo_multicast_addr;
2386 } else {
2387 IFP_TO_IA(imo->imo_multicast_ifp, ia);
2388 addr.s_addr = (ia == NULL) ? INADDR_ANY
2389 : IA_SIN(ia)->sin_addr.s_addr;
2390 }
2391 error = sooptcopyout(sopt, &addr, sizeof addr);
2392 break;
2393
2394 case IP_MULTICAST_TTL:
2395 if (imo == 0)
2396 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
2397 else
2398 optval = coptval = imo->imo_multicast_ttl;
2399 if (sopt->sopt_valsize == 1)
2400 error = sooptcopyout(sopt, &coptval, 1);
2401 else
2402 error = sooptcopyout(sopt, &optval, sizeof optval);
2403 break;
2404
2405 case IP_MULTICAST_LOOP:
2406 if (imo == 0)
2407 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
2408 else
2409 optval = coptval = imo->imo_multicast_loop;
2410 if (sopt->sopt_valsize == 1)
2411 error = sooptcopyout(sopt, &coptval, 1);
2412 else
2413 error = sooptcopyout(sopt, &optval, sizeof optval);
2414 break;
2415
2416 default:
2417 error = ENOPROTOOPT;
2418 break;
2419 }
2420 return (error);
2421 }
2422
2423 /*
2424 * Discard the IP multicast options.
2425 */
2426 void
2427 ip_freemoptions(imo)
2428 register struct ip_moptions *imo;
2429 {
2430 register int i;
2431
2432 if (imo != NULL) {
2433 for (i = 0; i < imo->imo_num_memberships; ++i)
2434 in_delmulti(&imo->imo_membership[i]);
2435 FREE(imo, M_IPMOPTS);
2436 }
2437 }
2438
2439 /*
2440 * Routine called from ip_output() to loop back a copy of an IP multicast
2441 * packet to the input queue of a specified interface. Note that this
2442 * calls the output routine of the loopback "driver", but with an interface
2443 * pointer that might NOT be a loopback interface -- evil, but easier than
2444 * replicating that code here.
2445 */
2446 static void
2447 ip_mloopback(ifp, m, dst, hlen)
2448 struct ifnet *ifp;
2449 register struct mbuf *m;
2450 register struct sockaddr_in *dst;
2451 int hlen;
2452 {
2453 register struct ip *ip;
2454 struct mbuf *copym;
2455
2456 copym = m_copy(m, 0, M_COPYALL);
2457 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2458 copym = m_pullup(copym, hlen);
2459 if (copym != NULL) {
2460 /*
2461 * We don't bother to fragment if the IP length is greater
2462 * than the interface's MTU. Can this possibly matter?
2463 */
2464 ip = mtod(copym, struct ip *);
2465 HTONS(ip->ip_len);
2466 HTONS(ip->ip_off);
2467 ip->ip_sum = 0;
2468 ip->ip_sum = in_cksum(copym, hlen);
2469 /*
2470 * NB:
2471 * It's not clear whether there are any lingering
2472 * reentrancy problems in other areas which might
2473 * be exposed by using ip_input directly (in
2474 * particular, everything which modifies the packet
2475 * in-place). Yet another option is using the
2476 * protosw directly to deliver the looped back
2477 * packet. For the moment, we'll err on the side
2478 * of safety by using if_simloop().
2479 */
2480 #if 1 /* XXX */
2481 if (dst->sin_family != AF_INET) {
2482 printf("ip_mloopback: bad address family %d\n",
2483 dst->sin_family);
2484 dst->sin_family = AF_INET;
2485 }
2486 #endif
2487
2488
2489 /*
2490 * Mark checksum as valid or calculate checksum for loopback.
2491 *
2492 * This is done this way because we have to embed the ifp of
2493 * the interface we will send the original copy of the packet
2494 * out on in the mbuf. ip_input will check if_hwassist of the
2495 * embedded ifp and ignore all csum_flags if if_hwassist is 0.
2496 * The UDP checksum has not been calculated yet.
2497 */
2498 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2499 if (IF_HWASSIST_CSUM_FLAGS(ifp->if_hwassist)) {
2500 copym->m_pkthdr.csum_flags |=
2501 CSUM_DATA_VALID | CSUM_PSEUDO_HDR |
2502 CSUM_IP_CHECKED | CSUM_IP_VALID;
2503 copym->m_pkthdr.csum_data = 0xffff;
2504 } else {
2505 NTOHS(ip->ip_len);
2506 in_delayed_cksum(copym);
2507 HTONS(ip->ip_len);
2508 }
2509 }
2510
2511
2512 /*
2513 * TedW:
2514 * We need to send all loopback traffic down to dlil in case
2515 * a filter has tapped-in.
2516 */
2517
2518 /*
2519 * Stuff the 'real' ifp into the pkthdr, to be used in matching
2520 * in ip_input(); we need the loopback ifp/dl_tag passed as args
2521 * to make the loopback driver compliant with the data link
2522 * requirements.
2523 */
2524 if (lo_ifp) {
2525 copym->m_pkthdr.rcvif = ifp;
2526 dlil_output(lo_ifp, PF_INET, copym, 0, (struct sockaddr *) dst, 0);
2527 } else {
2528 printf("Warning: ip_output call to dlil_find_dltag failed!\n");
2529 m_freem(copym);
2530 }
2531
2532 /* if_simloop(ifp, copym, (struct sockaddr *)dst, 0);*/
2533 }
2534 }
mirror of XNU