Make all network interrupt service routines MPSAFE part 1/3.
[dragonfly.git] / sys / netinet / ip_output.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
e90fc8a2 13 * 3. Neither the name of the University nor the names of its contributors
984263bc
MD
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
30 * $FreeBSD: src/sys/netinet/ip_output.c,v 1.99.2.37 2003/04/15 06:44:45 silby Exp $
78195a76 31 * $DragonFly: src/sys/netinet/ip_output.c,v 1.33 2005/11/28 17:13:46 dillon Exp $
984263bc
MD
32 */
33
34#define _IP_VHL
35
36#include "opt_ipfw.h"
37#include "opt_ipdn.h"
38#include "opt_ipdivert.h"
39#include "opt_ipfilter.h"
40#include "opt_ipsec.h"
984263bc
MD
41#include "opt_mbuf_stress_test.h"
42
43#include <sys/param.h>
44#include <sys/systm.h>
45#include <sys/kernel.h>
46#include <sys/malloc.h>
47#include <sys/mbuf.h>
48#include <sys/protosw.h>
49#include <sys/socket.h>
50#include <sys/socketvar.h>
51#include <sys/proc.h>
52#include <sys/sysctl.h>
1cae611f 53#include <sys/thread2.h>
3f9db7f8 54#include <sys/in_cksum.h>
984263bc
MD
55
56#include <net/if.h>
9eeaa8a9 57#include <net/netisr.h>
e7e55f42 58#include <net/pfil.h>
984263bc
MD
59#include <net/route.h>
60
61#include <netinet/in.h>
62#include <netinet/in_systm.h>
63#include <netinet/ip.h>
64#include <netinet/in_pcb.h>
65#include <netinet/in_var.h>
66#include <netinet/ip_var.h>
67
984263bc
MD
68static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
69
70#ifdef IPSEC
71#include <netinet6/ipsec.h>
d2438d69 72#include <netproto/key/key.h>
984263bc 73#ifdef IPSEC_DEBUG
d2438d69 74#include <netproto/key/key_debug.h>
984263bc
MD
75#else
76#define KEYDEBUG(lev,arg)
77#endif
78#endif /*IPSEC*/
79
80#ifdef FAST_IPSEC
bf844ffa
JH
81#include <netproto/ipsec/ipsec.h>
82#include <netproto/ipsec/xform.h>
83#include <netproto/ipsec/key.h>
984263bc
MD
84#endif /*FAST_IPSEC*/
85
1f2de5d4
MD
86#include <net/ipfw/ip_fw.h>
87#include <net/dummynet/ip_dummynet.h>
984263bc
MD
88
89#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\
90 x, (ntohl(a.s_addr)>>24)&0xFF,\
91 (ntohl(a.s_addr)>>16)&0xFF,\
92 (ntohl(a.s_addr)>>8)&0xFF,\
93 (ntohl(a.s_addr))&0xFF, y);
94
95u_short ip_id;
96
97#ifdef MBUF_STRESS_TEST
98int mbuf_frag_size = 0;
99SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
100 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
101#endif
102
103static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
104static struct ifnet *ip_multicast_if(struct in_addr *, int *);
105static void ip_mloopback
106 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
107static int ip_getmoptions
108 (struct sockopt *, struct ip_moptions *);
109static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
110static int ip_setmoptions
111 (struct sockopt *, struct ip_moptions **);
112
113int ip_optcopy(struct ip *, struct ip *);
984263bc
MD
114
115
116extern struct protosw inetsw[];
117
118/*
119 * IP output. The packet in mbuf chain m contains a skeletal IP
120 * header (with len, off, ttl, proto, tos, src, dst).
121 * The mbuf chain containing the packet will be freed.
122 * The mbuf opt, if present, will not be freed.
123 */
124int
f1f552f6 125ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
5fe66e68 126 int flags, struct ip_moptions *imo, struct inpcb *inp)
984263bc 127{
f1f552f6 128 struct ip *ip;
984263bc
MD
129 struct ifnet *ifp = NULL; /* keep compiler happy */
130 struct mbuf *m;
5fe66e68 131 int hlen = sizeof(struct ip);
984263bc
MD
132 int len, off, error = 0;
133 struct sockaddr_in *dst = NULL; /* keep compiler happy */
134 struct in_ifaddr *ia = NULL;
135 int isbroadcast, sw_csum;
136 struct in_addr pkt_dst;
984263bc 137 struct route iproute;
fade9ce3 138#ifdef IPSEC
984263bc
MD
139 struct secpolicy *sp = NULL;
140 struct socket *so = inp ? inp->inp_socket : NULL;
141#endif
142#ifdef FAST_IPSEC
984263bc
MD
143 struct m_tag *mtag;
144 struct secpolicy *sp = NULL;
145 struct tdb_ident *tdbi;
984263bc
MD
146#endif /* FAST_IPSEC */
147 struct ip_fw_args args;
148 int src_was_INADDR_ANY = 0; /* as the name says... */
149
150 args.eh = NULL;
151 args.rule = NULL;
152 args.next_hop = NULL;
984263bc
MD
153
154 /* Grab info from MT_TAG mbufs prepended to the chain. */
5fe66e68 155 while (m0 != NULL && m0->m_type == MT_TAG) {
984263bc 156 switch(m0->_m_tag_id) {
984263bc
MD
157 case PACKET_TAG_DUMMYNET:
158 /*
159 * the packet was already tagged, so part of the
160 * processing was already done, and we need to go down.
161 * Get parameters from the header.
162 */
163 args.rule = ((struct dn_pkt *)m0)->rule;
164 opt = NULL ;
5fe66e68 165 ro = &((struct dn_pkt *)m0)->ro;
984263bc
MD
166 imo = NULL ;
167 dst = ((struct dn_pkt *)m0)->dn_dst ;
168 ifp = ((struct dn_pkt *)m0)->ifp ;
169 flags = ((struct dn_pkt *)m0)->flags ;
170 break;
984263bc
MD
171 case PACKET_TAG_IPFORWARD:
172 args.next_hop = (struct sockaddr_in *)m0->m_data;
173 break;
5fe66e68
JH
174 default:
175 printf("ip_output: unrecognised MT_TAG tag %d\n",
176 m0->_m_tag_id);
177 break;
984263bc 178 }
5fe66e68 179 m0 = m0->m_next;
984263bc
MD
180 }
181 m = m0;
5fe66e68 182 KASSERT(m != NULL && (m->m_flags & M_PKTHDR), ("ip_output: no HDR"));
fade9ce3
JS
183
184 if (ro == NULL) {
185 ro = &iproute;
5fe66e68 186 bzero(ro, sizeof *ro);
fade9ce3 187 }
984263bc
MD
188
189 if (args.rule != NULL) { /* dummynet already saw us */
190 ip = mtod(m, struct ip *);
191 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
192 if (ro->ro_rt)
193 ia = ifatoia(ro->ro_rt->rt_ifa);
194 goto sendit;
195 }
196
197 if (opt) {
198 len = 0;
199 m = ip_insertoptions(m, opt, &len);
200 if (len != 0)
201 hlen = len;
202 }
203 ip = mtod(m, struct ip *);
204 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
205
206 /*
207 * Fill in IP header.
208 */
f23061d4 209 if (!(flags & (IP_FORWARDING|IP_RAWOUTPUT))) {
984263bc
MD
210 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
211 ip->ip_off &= IP_DF;
6277137d 212 ip->ip_id = ip_newid();
984263bc
MD
213 ipstat.ips_localout++;
214 } else {
215 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
216 }
217
984263bc
MD
218 dst = (struct sockaddr_in *)&ro->ro_dst;
219 /*
220 * If there is a cached route,
221 * check that it is to the same destination
222 * and is still up. If not, free it and try again.
223 * The address family should also be checked in case of sharing the
224 * cache with IPv6.
225 */
f23061d4
JH
226 if (ro->ro_rt &&
227 (!(ro->ro_rt->rt_flags & RTF_UP) ||
228 dst->sin_family != AF_INET ||
229 dst->sin_addr.s_addr != pkt_dst.s_addr)) {
5fe66e68 230 rtfree(ro->ro_rt);
f23061d4 231 ro->ro_rt = (struct rtentry *)NULL;
984263bc 232 }
f23061d4 233 if (ro->ro_rt == NULL) {
5fe66e68 234 bzero(dst, sizeof *dst);
984263bc 235 dst->sin_family = AF_INET;
5fe66e68 236 dst->sin_len = sizeof *dst;
984263bc
MD
237 dst->sin_addr = pkt_dst;
238 }
239 /*
240 * If routing to interface only,
241 * short circuit routing lookup.
242 */
243 if (flags & IP_ROUTETOIF) {
f23061d4
JH
244 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
245 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) {
984263bc
MD
246 ipstat.ips_noroute++;
247 error = ENETUNREACH;
248 goto bad;
249 }
250 ifp = ia->ia_ifp;
251 ip->ip_ttl = 1;
252 isbroadcast = in_broadcast(dst->sin_addr, ifp);
253 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
f23061d4 254 imo != NULL && imo->imo_multicast_ifp != NULL) {
984263bc
MD
255 /*
256 * Bypass the normal routing lookup for multicast
257 * packets if the interface is specified.
258 */
259 ifp = imo->imo_multicast_ifp;
260 IFP_TO_IA(ifp, ia);
261 isbroadcast = 0; /* fool gcc */
262 } else {
263 /*
264 * If this is the case, we probably don't want to allocate
265 * a protocol-cloned route since we didn't get one from the
266 * ULP. This lets TCP do its thing, while not burdening
267 * forwarding or ICMP with the overhead of cloning a route.
268 * Of course, we still want to do any cloning requested by
269 * the link layer, as this is probably required in all cases
270 * for correct operation (as it is for ARP).
271 */
f23061d4 272 if (ro->ro_rt == NULL)
984263bc 273 rtalloc_ign(ro, RTF_PRCLONING);
f23061d4 274 if (ro->ro_rt == NULL) {
984263bc
MD
275 ipstat.ips_noroute++;
276 error = EHOSTUNREACH;
277 goto bad;
278 }
279 ia = ifatoia(ro->ro_rt->rt_ifa);
280 ifp = ro->ro_rt->rt_ifp;
281 ro->ro_rt->rt_use++;
282 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
283 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
284 if (ro->ro_rt->rt_flags & RTF_HOST)
285 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
286 else
287 isbroadcast = in_broadcast(dst->sin_addr, ifp);
288 }
289 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
290 struct in_multi *inm;
291
292 m->m_flags |= M_MCAST;
293 /*
294 * IP destination address is multicast. Make sure "dst"
295 * still points to the address in "ro". (It may have been
296 * changed to point to a gateway address, above.)
297 */
298 dst = (struct sockaddr_in *)&ro->ro_dst;
299 /*
300 * See if the caller provided any multicast options
301 */
302 if (imo != NULL) {
303 ip->ip_ttl = imo->imo_multicast_ttl;
304 if (imo->imo_multicast_vif != -1)
305 ip->ip_src.s_addr =
306 ip_mcast_src ?
307 ip_mcast_src(imo->imo_multicast_vif) :
308 INADDR_ANY;
309 } else
310 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
311 /*
312 * Confirm that the outgoing interface supports multicast.
313 */
314 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
f23061d4 315 if (!(ifp->if_flags & IFF_MULTICAST)) {
984263bc
MD
316 ipstat.ips_noroute++;
317 error = ENETUNREACH;
318 goto bad;
319 }
320 }
321 /*
322 * If source address not specified yet, use address
323 * of outgoing interface.
324 */
325 if (ip->ip_src.s_addr == INADDR_ANY) {
326 /* Interface may have no addresses. */
327 if (ia != NULL)
328 ip->ip_src = IA_SIN(ia)->sin_addr;
329 }
330
331 IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
332 if (inm != NULL &&
333 (imo == NULL || imo->imo_multicast_loop)) {
334 /*
335 * If we belong to the destination multicast group
336 * on the outgoing interface, and the caller did not
337 * forbid loopback, loop back a copy.
338 */
339 ip_mloopback(ifp, m, dst, hlen);
340 }
341 else {
342 /*
343 * If we are acting as a multicast router, perform
344 * multicast forwarding as if the packet had just
345 * arrived on the interface to which we are about
346 * to send. The multicast forwarding function
347 * recursively calls this function, using the
348 * IP_FORWARDING flag to prevent infinite recursion.
349 *
350 * Multicasts that are looped back by ip_mloopback(),
351 * above, will be forwarded by the ip_input() routine,
352 * if necessary.
353 */
f23061d4 354 if (ip_mrouter && !(flags & IP_FORWARDING)) {
984263bc
MD
355 /*
356 * If rsvp daemon is not running, do not
357 * set ip_moptions. This ensures that the packet
358 * is multicast and not just sent down one link
359 * as prescribed by rsvpd.
360 */
361 if (!rsvp_on)
362 imo = NULL;
363 if (ip_mforward &&
364 ip_mforward(ip, ifp, m, imo) != 0) {
365 m_freem(m);
366 goto done;
367 }
368 }
369 }
370
371 /*
372 * Multicasts with a time-to-live of zero may be looped-
373 * back, above, but must not be transmitted on a network.
374 * Also, multicasts addressed to the loopback interface
375 * are not sent -- the above call to ip_mloopback() will
376 * loop back a copy if this host actually belongs to the
377 * destination group on the loopback interface.
378 */
379 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
380 m_freem(m);
381 goto done;
382 }
383
384 goto sendit;
385 }
386#ifndef notdef
387 /*
388 * If the source address is not specified yet, use the address
389 * of the outoing interface. In case, keep note we did that, so
390 * if the the firewall changes the next-hop causing the output
391 * interface to change, we can fix that.
392 */
393 if (ip->ip_src.s_addr == INADDR_ANY) {
394 /* Interface may have no addresses. */
395 if (ia != NULL) {
396 ip->ip_src = IA_SIN(ia)->sin_addr;
397 src_was_INADDR_ANY = 1;
398 }
399 }
400#endif /* notdef */
4d723e5a
JS
401#ifdef ALTQ
402 /*
403 * Disable packet drop hack.
404 * Packetdrop should be done by queueing.
405 */
406#else /* !ALTQ */
984263bc
MD
407 /*
408 * Verify that we have any chance at all of being able to queue
409 * the packet or packet fragments
410 */
411 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
412 ifp->if_snd.ifq_maxlen) {
413 error = ENOBUFS;
414 ipstat.ips_odropped++;
415 goto bad;
416 }
4d723e5a 417#endif /* !ALTQ */
984263bc
MD
418
419 /*
420 * Look for broadcast address and
421 * verify user is allowed to send
422 * such a packet.
423 */
424 if (isbroadcast) {
f23061d4 425 if (!(ifp->if_flags & IFF_BROADCAST)) {
984263bc
MD
426 error = EADDRNOTAVAIL;
427 goto bad;
428 }
f23061d4 429 if (!(flags & IP_ALLOWBROADCAST)) {
984263bc
MD
430 error = EACCES;
431 goto bad;
432 }
433 /* don't allow broadcast messages to be fragmented */
f1f552f6 434 if (ip->ip_len > ifp->if_mtu) {
984263bc
MD
435 error = EMSGSIZE;
436 goto bad;
437 }
438 m->m_flags |= M_BCAST;
439 } else {
440 m->m_flags &= ~M_BCAST;
441 }
442
443sendit:
444#ifdef IPSEC
445 /* get SP for this packet */
446 if (so == NULL)
447 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
448 else
449 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
450
451 if (sp == NULL) {
452 ipsecstat.out_inval++;
453 goto bad;
454 }
455
456 error = 0;
457
458 /* check policy */
459 switch (sp->policy) {
460 case IPSEC_POLICY_DISCARD:
461 /*
462 * This packet is just discarded.
463 */
464 ipsecstat.out_polvio++;
465 goto bad;
466
467 case IPSEC_POLICY_BYPASS:
468 case IPSEC_POLICY_NONE:
469 /* no need to do IPsec. */
470 goto skip_ipsec;
5fe66e68 471
984263bc
MD
472 case IPSEC_POLICY_IPSEC:
473 if (sp->req == NULL) {
474 /* acquire a policy */
475 error = key_spdacquire(sp);
476 goto bad;
477 }
478 break;
479
480 case IPSEC_POLICY_ENTRUST:
481 default:
482 printf("ip_output: Invalid policy found. %d\n", sp->policy);
483 }
484 {
485 struct ipsec_output_state state;
5fe66e68 486 bzero(&state, sizeof state);
984263bc
MD
487 state.m = m;
488 if (flags & IP_ROUTETOIF) {
489 state.ro = &iproute;
5fe66e68 490 bzero(&iproute, sizeof iproute);
984263bc
MD
491 } else
492 state.ro = ro;
493 state.dst = (struct sockaddr *)dst;
494
495 ip->ip_sum = 0;
496
497 /*
498 * XXX
499 * delayed checksums are not currently compatible with IPsec
500 */
501 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
502 in_delayed_cksum(m);
503 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
504 }
505
506 ip->ip_len = htons(ip->ip_len);
507 ip->ip_off = htons(ip->ip_off);
508
509 error = ipsec4_output(&state, sp, flags);
510
511 m = state.m;
512 if (flags & IP_ROUTETOIF) {
513 /*
514 * if we have tunnel mode SA, we may need to ignore
515 * IP_ROUTETOIF.
516 */
517 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
518 flags &= ~IP_ROUTETOIF;
519 ro = state.ro;
520 }
521 } else
522 ro = state.ro;
523 dst = (struct sockaddr_in *)state.dst;
524 if (error) {
525 /* mbuf is already reclaimed in ipsec4_output. */
526 m0 = NULL;
527 switch (error) {
528 case EHOSTUNREACH:
529 case ENETUNREACH:
530 case EMSGSIZE:
531 case ENOBUFS:
532 case ENOMEM:
533 break;
534 default:
535 printf("ip4_output (ipsec): error code %d\n", error);
536 /*fall through*/
537 case ENOENT:
538 /* don't show these error codes to the user */
539 error = 0;
540 break;
541 }
542 goto bad;
543 }
544 }
545
546 /* be sure to update variables that are affected by ipsec4_output() */
547 ip = mtod(m, struct ip *);
548#ifdef _IP_VHL
549 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
550#else
551 hlen = ip->ip_hl << 2;
552#endif
553 if (ro->ro_rt == NULL) {
f23061d4 554 if (!(flags & IP_ROUTETOIF)) {
984263bc
MD
555 printf("ip_output: "
556 "can't update route after IPsec processing\n");
557 error = EHOSTUNREACH; /*XXX*/
558 goto bad;
559 }
560 } else {
561 ia = ifatoia(ro->ro_rt->rt_ifa);
562 ifp = ro->ro_rt->rt_ifp;
563 }
564
565 /* make it flipped, again. */
566 ip->ip_len = ntohs(ip->ip_len);
567 ip->ip_off = ntohs(ip->ip_off);
568skip_ipsec:
569#endif /*IPSEC*/
570#ifdef FAST_IPSEC
571 /*
572 * Check the security policy (SP) for the packet and, if
573 * required, do IPsec-related processing. There are two
574 * cases here; the first time a packet is sent through
575 * it will be untagged and handled by ipsec4_checkpolicy.
576 * If the packet is resubmitted to ip_output (e.g. after
577 * AH, ESP, etc. processing), there will be a tag to bypass
578 * the lookup and related policy checking.
579 */
580 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
1cae611f 581 crit_enter();
984263bc 582 if (mtag != NULL) {
d031aa80 583 tdbi = (struct tdb_ident *)m_tag_data(mtag);
984263bc
MD
584 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
585 if (sp == NULL)
586 error = -EINVAL; /* force silent drop */
587 m_tag_delete(m, mtag);
588 } else {
589 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
590 &error, inp);
591 }
592 /*
593 * There are four return cases:
f23061d4 594 * sp != NULL apply IPsec policy
984263bc
MD
595 * sp == NULL, error == 0 no IPsec handling needed
596 * sp == NULL, error == -EINVAL discard packet w/o error
597 * sp == NULL, error != 0 discard packet, report error
598 */
599 if (sp != NULL) {
600 /* Loop detection, check if ipsec processing already done */
601 KASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
602 for (mtag = m_tag_first(m); mtag != NULL;
603 mtag = m_tag_next(m, mtag)) {
604 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
605 continue;
606 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
607 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
608 continue;
609 /*
610 * Check if policy has an SA associated with it.
611 * This can happen when an SP has yet to acquire
612 * an SA; e.g. on first reference. If it occurs,
613 * then we let ipsec4_process_packet do its thing.
614 */
615 if (sp->req->sav == NULL)
616 break;
d031aa80 617 tdbi = (struct tdb_ident *)m_tag_data(mtag);
984263bc
MD
618 if (tdbi->spi == sp->req->sav->spi &&
619 tdbi->proto == sp->req->sav->sah->saidx.proto &&
620 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
5fe66e68 621 sizeof(union sockaddr_union)) == 0) {
984263bc
MD
622 /*
623 * No IPsec processing is needed, free
624 * reference to SP.
625 *
626 * NB: null pointer to avoid free at
627 * done: below.
628 */
629 KEY_FREESP(&sp), sp = NULL;
1cae611f 630 crit_exit();
984263bc
MD
631 goto spd_done;
632 }
633 }
634
635 /*
636 * Do delayed checksums now because we send before
637 * this is done in the normal processing path.
638 */
639 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
640 in_delayed_cksum(m);
641 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
642 }
643
644 ip->ip_len = htons(ip->ip_len);
645 ip->ip_off = htons(ip->ip_off);
646
647 /* NB: callee frees mbuf */
648 error = ipsec4_process_packet(m, sp->req, flags, 0);
649 /*
650 * Preserve KAME behaviour: ENOENT can be returned
651 * when an SA acquire is in progress. Don't propagate
652 * this to user-level; it confuses applications.
653 *
654 * XXX this will go away when the SADB is redone.
655 */
656 if (error == ENOENT)
657 error = 0;
1cae611f 658 crit_exit();
984263bc
MD
659 goto done;
660 } else {
1cae611f 661 crit_exit();
984263bc
MD
662
663 if (error != 0) {
664 /*
665 * Hack: -EINVAL is used to signal that a packet
666 * should be silently discarded. This is typically
667 * because we asked key management for an SA and
668 * it was delayed (e.g. kicked up to IKE).
669 */
670 if (error == -EINVAL)
671 error = 0;
672 goto bad;
673 } else {
674 /* No IPsec processing for this packet. */
675 }
676#ifdef notyet
677 /*
678 * If deferred crypto processing is needed, check that
679 * the interface supports it.
f23061d4 680 */
984263bc 681 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
f23061d4 682 if (mtag != NULL && !(ifp->if_capenable & IFCAP_IPSEC)) {
984263bc 683 /* notify IPsec to do its own crypto */
d031aa80 684 ipsp_skipcrypto_unmark((struct tdb_ident *)m_tag_data(mtag));
984263bc
MD
685 error = EHOSTUNREACH;
686 goto bad;
687 }
688#endif
689 }
690spd_done:
691#endif /* FAST_IPSEC */
692 /*
693 * IpHack's section.
694 * - Xlate: translate packet's addr/port (NAT).
695 * - Firewall: deny/allow/etc.
696 * - Wrap: fake packet's addr/port <unimpl.>
697 * - Encapsulate: put it in another IP and send out. <unimp.>
f23061d4 698 */
5e3f3b7a 699
e7e55f42
JR
700 /*
701 * Run through list of hooks for output packets.
702 */
afabe90c
MD
703 if (pfil_has_hooks(&inet_pfil_hook)) {
704 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT);
705 if (error != 0 || m == NULL)
706 goto done;
707 ip = mtod(m, struct ip *);
708 }
984263bc
MD
709
710 /*
711 * Check with the firewall...
712 * but not if we are already being fwd'd from a firewall.
713 */
714 if (fw_enable && IPFW_LOADED && !args.next_hop) {
715 struct sockaddr_in *old = dst;
716
717 args.m = m;
718 args.next_hop = dst;
719 args.oif = ifp;
720 off = ip_fw_chk_ptr(&args);
721 m = args.m;
722 dst = args.next_hop;
723
f23061d4 724 /*
984263bc
MD
725 * On return we must do the following:
726 * m == NULL -> drop the pkt (old interface, deprecated)
727 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface)
728 * 1<=off<= 0xffff -> DIVERT
729 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe
730 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet
731 * dst != old -> IPFIREWALL_FORWARD
732 * off==0, dst==old -> accept
733 * If some of the above modules are not compiled in, then
734 * we should't have to check the corresponding condition
735 * (because the ipfw control socket should not accept
736 * unsupported rules), but better play safe and drop
737 * packets in case of doubt.
738 */
739 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
740 if (m)
741 m_freem(m);
742 error = EACCES;
743 goto done;
744 }
745 ip = mtod(m, struct ip *);
746 if (off == 0 && dst == old) /* common case */
747 goto pass;
f23061d4 748 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG)) {
984263bc
MD
749 /*
750 * pass the pkt to dummynet. Need to include
751 * pipe number, m, ifp, ro, dst because these are
752 * not recomputed in the next pass.
753 * All other parameters have been already used and
f23061d4 754 * so they are not needed anymore.
984263bc
MD
755 * XXX note: if the ifp or ro entry are deleted
756 * while a pkt is in dummynet, we are in trouble!
f23061d4 757 */
984263bc
MD
758 args.ro = ro;
759 args.dst = dst;
760 args.flags = flags;
761
762 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT,
763 &args);
764 goto done;
765 }
766#ifdef IPDIVERT
f23061d4 767 if (off != 0 && !(off & IP_FW_PORT_DYNT_FLAG)) {
984263bc
MD
768 struct mbuf *clone = NULL;
769
770 /* Clone packet if we're doing a 'tee' */
f23061d4 771 if ((off & IP_FW_PORT_TEE_FLAG))
74f1caca 772 clone = m_dup(m, MB_DONTWAIT);
984263bc
MD
773
774 /*
775 * XXX
776 * delayed checksums are not currently compatible
777 * with divert sockets.
778 */
779 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
780 in_delayed_cksum(m);
781 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
782 }
783
784 /* Restore packet header fields to original values */
785 ip->ip_len = htons(ip->ip_len);
786 ip->ip_off = htons(ip->ip_off);
787
788 /* Deliver packet to divert input routine */
e5ecc832 789 divert_packet(m, 0, off & 0xffff);
984263bc
MD
790
791 /* If 'tee', continue with original packet */
792 if (clone != NULL) {
793 m = clone;
794 ip = mtod(m, struct ip *);
795 goto pass;
796 }
797 goto done;
798 }
799#endif
800
801 /* IPFIREWALL_FORWARD */
802 /*
803 * Check dst to make sure it is directly reachable on the
804 * interface we previously thought it was.
805 * If it isn't (which may be likely in some situations) we have
806 * to re-route it (ie, find a route for the next-hop and the
807 * associated interface) and set them here. This is nested
808 * forwarding which in most cases is undesirable, except where
809 * such control is nigh impossible. So we do it here.
810 * And I'm babbling.
811 */
812 if (off == 0 && old != dst) { /* FORWARD, dst has changed */
813#if 0
814 /*
815 * XXX To improve readability, this block should be
816 * changed into a function call as below:
817 */
818 error = ip_ipforward(&m, &dst, &ifp);
819 if (error)
820 goto bad;
821 if (m == NULL) /* ip_input consumed the mbuf */
822 goto done;
823#else
824 struct in_ifaddr *ia;
825
826 /*
827 * XXX sro_fwd below is static, and a pointer
828 * to it gets passed to routines downstream.
829 * This could have surprisingly bad results in
830 * practice, because its content is overwritten
831 * by subsequent packets.
832 */
833 /* There must be a better way to do this next line... */
834 static struct route sro_fwd;
835 struct route *ro_fwd = &sro_fwd;
836
837#if 0
838 print_ip("IPFIREWALL_FORWARD: New dst ip: ",
839 dst->sin_addr, "\n");
840#endif
841
842 /*
843 * We need to figure out if we have been forwarded
f23061d4 844 * to a local socket. If so, then we should somehow
984263bc
MD
845 * "loop back" to ip_input, and get directed to the
846 * PCB as if we had received this packet. This is
847 * because it may be dificult to identify the packets
848 * you want to forward until they are being output
849 * and have selected an interface. (e.g. locally
850 * initiated packets) If we used the loopback inteface,
f23061d4 851 * we would not be able to control what happens
984263bc
MD
852 * as the packet runs through ip_input() as
853 * it is done through a ISR.
854 */
f23061d4
JH
855 LIST_FOREACH(ia, INADDR_HASH(dst->sin_addr.s_addr),
856 ia_hash) {
984263bc
MD
857 /*
858 * If the addr to forward to is one
859 * of ours, we pretend to
860 * be the destination for this packet.
861 */
862 if (IA_SIN(ia)->sin_addr.s_addr ==
863 dst->sin_addr.s_addr)
864 break;
865 }
f23061d4 866 if (ia != NULL) { /* tell ip_input "dont filter" */
984263bc
MD
867 struct m_hdr tag;
868
869 tag.mh_type = MT_TAG;
870 tag.mh_flags = PACKET_TAG_IPFORWARD;
871 tag.mh_data = (caddr_t)args.next_hop;
872 tag.mh_next = m;
873
874 if (m->m_pkthdr.rcvif == NULL)
875 m->m_pkthdr.rcvif = ifunit("lo0");
876 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
877 m->m_pkthdr.csum_flags |=
878 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
879 m0->m_pkthdr.csum_data = 0xffff;
880 }
881 m->m_pkthdr.csum_flags |=
882 CSUM_IP_CHECKED | CSUM_IP_VALID;
883 ip->ip_len = htons(ip->ip_len);
884 ip->ip_off = htons(ip->ip_off);
4f277347 885 ip_input((struct mbuf *)&tag);
984263bc
MD
886 goto done;
887 }
f23061d4 888 /* Some of the logic for this was nicked from above.
984263bc
MD
889 *
890 * This rewrites the cached route in a local PCB.
891 * Is this what we want to do?
892 */
5fe66e68 893 bcopy(dst, &ro_fwd->ro_dst, sizeof *dst);
f23061d4 894 ro_fwd->ro_rt = NULL;
984263bc 895
5fe66e68 896 rtalloc_ign(ro_fwd, RTF_PRCLONING);
f23061d4 897 if (ro_fwd->ro_rt == NULL) {
984263bc
MD
898 ipstat.ips_noroute++;
899 error = EHOSTUNREACH;
900 goto bad;
901 }
902
903 ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
904 ifp = ro_fwd->ro_rt->rt_ifp;
905 ro_fwd->ro_rt->rt_use++;
906 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
907 dst = (struct sockaddr_in *)
5fe66e68 908 ro_fwd->ro_rt->rt_gateway;
984263bc
MD
909 if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
910 isbroadcast =
911 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
912 else
913 isbroadcast = in_broadcast(dst->sin_addr, ifp);
5fe66e68
JH
914 if (ro->ro_rt != NULL)
915 rtfree(ro->ro_rt);
984263bc
MD
916 ro->ro_rt = ro_fwd->ro_rt;
917 dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
918
919#endif /* ... block to be put into a function */
920 /*
921 * If we added a default src ip earlier,
922 * which would have been gotten from the-then
923 * interface, do it again, from the new one.
924 */
925 if (src_was_INADDR_ANY)
926 ip->ip_src = IA_SIN(ia)->sin_addr;
927 goto pass ;
928 }
929
f23061d4
JH
930 /*
931 * if we get here, none of the above matches, and
932 * we have to drop the pkt
933 */
984263bc 934 m_freem(m);
f23061d4
JH
935 error = EACCES; /* not sure this is the right error msg */
936 goto done;
984263bc
MD
937 }
938
939pass:
940 /* 127/8 must not appear on wire - RFC1122. */
941 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
942 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
f23061d4 943 if (!(ifp->if_flags & IFF_LOOPBACK)) {
984263bc
MD
944 ipstat.ips_badaddr++;
945 error = EADDRNOTAVAIL;
946 goto bad;
947 }
948 }
949
950 m->m_pkthdr.csum_flags |= CSUM_IP;
951 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
952 if (sw_csum & CSUM_DELAY_DATA) {
953 in_delayed_cksum(m);
954 sw_csum &= ~CSUM_DELAY_DATA;
955 }
956 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
957
958 /*
959 * If small enough for interface, or the interface will take
960 * care of the fragmentation for us, can just send directly.
961 */
7fb962be 962 if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
f23061d4 963 !(ip->ip_off & IP_DF))) {
984263bc
MD
964 ip->ip_len = htons(ip->ip_len);
965 ip->ip_off = htons(ip->ip_off);
966 ip->ip_sum = 0;
967 if (sw_csum & CSUM_DELAY_IP) {
968 if (ip->ip_vhl == IP_VHL_BORING) {
969 ip->ip_sum = in_cksum_hdr(ip);
970 } else {
971 ip->ip_sum = in_cksum(m, hlen);
972 }
973 }
974
975 /* Record statistics for this interface address. */
976 if (!(flags & IP_FORWARDING) && ia) {
977 ia->ia_ifa.if_opackets++;
978 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
979 }
980
981#ifdef IPSEC
982 /* clean ipsec history once it goes out of the node */
983 ipsec_delaux(m);
984#endif
985
986#ifdef MBUF_STRESS_TEST
987 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) {
988 struct mbuf *m1, *m2;
989 int length, tmp;
990
991 tmp = length = m->m_pkthdr.len;
992
993 while ((length -= mbuf_frag_size) >= 1) {
74f1caca 994 m1 = m_split(m, length, MB_DONTWAIT);
984263bc
MD
995 if (m1 == NULL)
996 break;
984263bc
MD
997 m2 = m;
998 while (m2->m_next != NULL)
999 m2 = m2->m_next;
1000 m2->m_next = m1;
1001 }
1002 m->m_pkthdr.len = tmp;
1003 }
1004#endif
78195a76 1005 lwkt_serialize_enter(ifp->if_serializer);
5fe66e68
JH
1006 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst,
1007 ro->ro_rt);
78195a76 1008 lwkt_serialize_exit(ifp->if_serializer);
984263bc
MD
1009 goto done;
1010 }
f1f552f6 1011
984263bc
MD
1012 if (ip->ip_off & IP_DF) {
1013 error = EMSGSIZE;
1014 /*
1015 * This case can happen if the user changed the MTU
1016 * of an interface after enabling IP on it. Because
1017 * most netifs don't keep track of routes pointing to
1018 * them, there is no way for one to update all its
1019 * routes when the MTU is changed.
1020 */
f1f552f6
JH
1021 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
1022 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
1023 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
984263bc
MD
1024 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
1025 }
1026 ipstat.ips_cantfrag++;
1027 goto bad;
1028 }
f1f552f6
JH
1029
1030 /*
1031 * Too large for interface; fragment if possible. If successful,
1032 * on return, m will point to a list of packets to be sent.
1033 */
1034 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
1035 if (error)
984263bc 1036 goto bad;
f1f552f6
JH
1037 for (; m; m = m0) {
1038 m0 = m->m_nextpkt;
f23061d4 1039 m->m_nextpkt = NULL;
f1f552f6
JH
1040#ifdef IPSEC
1041 /* clean ipsec history once it goes out of the node */
1042 ipsec_delaux(m);
1043#endif
1044 if (error == 0) {
1045 /* Record statistics for this interface address. */
1046 if (ia != NULL) {
1047 ia->ia_ifa.if_opackets++;
1048 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1049 }
78195a76 1050 lwkt_serialize_enter(ifp->if_serializer);
f1f552f6 1051 error = (*ifp->if_output)(ifp, m,
5fe66e68
JH
1052 (struct sockaddr *)dst,
1053 ro->ro_rt);
78195a76
MD
1054 lwkt_serialize_exit(ifp->if_serializer);
1055 } else {
f1f552f6 1056 m_freem(m);
78195a76 1057 }
984263bc
MD
1058 }
1059
f1f552f6
JH
1060 if (error == 0)
1061 ipstat.ips_fragmented++;
1062
1063done:
5fe66e68 1064 if (ro == &iproute && ro->ro_rt != NULL) {
f1f552f6
JH
1065 RTFREE(ro->ro_rt);
1066 ro->ro_rt = NULL;
1067 }
fade9ce3 1068#ifdef IPSEC
f1f552f6
JH
1069 if (sp != NULL) {
1070 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1071 printf("DP ip_output call free SP:%p\n", sp));
1072 key_freesp(sp);
1073 }
1074#endif
1075#ifdef FAST_IPSEC
f1f552f6
JH
1076 if (sp != NULL)
1077 KEY_FREESP(&sp);
1078#endif
1079 return (error);
1080bad:
1081 m_freem(m);
1082 goto done;
1083}
1084
1085/*
1086 * Create a chain of fragments which fit the given mtu. m_frag points to the
1087 * mbuf to be fragmented; on return it points to the chain with the fragments.
1088 * Return 0 if no error. If error, m_frag may contain a partially built
1089 * chain of fragments that should be freed by the caller.
1090 *
1091 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
1092 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
1093 */
1094int
1095ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
1096 u_long if_hwassist_flags, int sw_csum)
1097{
1098 int error = 0;
1099 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1100 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
1101 int off;
1102 struct mbuf *m0 = *m_frag; /* the original packet */
1103 int firstlen;
1104 struct mbuf **mnext;
1105 int nfrags;
1106
1107 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
1108 ipstat.ips_cantfrag++;
1109 return EMSGSIZE;
1110 }
1111
1112 /*
1113 * Must be able to put at least 8 bytes per fragment.
1114 */
1115 if (len < 8)
1116 return EMSGSIZE;
1117
984263bc 1118 /*
f1f552f6 1119 * If the interface will not calculate checksums on
984263bc
MD
1120 * fragmented packets, then do it here.
1121 */
f23061d4
JH
1122 if ((m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) &&
1123 !(if_hwassist_flags & CSUM_IP_FRAGS)) {
f1f552f6
JH
1124 in_delayed_cksum(m0);
1125 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
984263bc
MD
1126 }
1127
f1f552f6 1128 if (len > PAGE_SIZE) {
f23061d4
JH
1129 /*
1130 * Fragment large datagrams such that each segment
1131 * contains a multiple of PAGE_SIZE amount of data,
1132 * plus headers. This enables a receiver to perform
f1f552f6
JH
1133 * page-flipping zero-copy optimizations.
1134 *
1135 * XXX When does this help given that sender and receiver
1136 * could have different page sizes, and also mtu could
1137 * be less than the receiver's page size ?
1138 */
1139 int newlen;
1140 struct mbuf *m;
1141
1142 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
1143 off += m->m_len;
1144
1145 /*
f23061d4 1146 * firstlen (off - hlen) must be aligned on an
f1f552f6
JH
1147 * 8-byte boundary
1148 */
1149 if (off < hlen)
1150 goto smart_frag_failure;
1151 off = ((off - hlen) & ~7) + hlen;
1152 newlen = (~PAGE_MASK) & mtu;
5fe66e68 1153 if ((newlen + sizeof(struct ip)) > mtu) {
f1f552f6
JH
1154 /* we failed, go back the default */
1155smart_frag_failure:
1156 newlen = len;
1157 off = hlen + len;
1158 }
1159 len = newlen;
1160
1161 } else {
1162 off = hlen + len;
1163 }
1164
1165 firstlen = off - hlen;
1166 mnext = &m0->m_nextpkt; /* pointer to next packet */
984263bc
MD
1167
1168 /*
1169 * Loop through length of segment after first fragment,
1170 * make new header and copy data of each part and link onto chain.
f1f552f6
JH
1171 * Here, m0 is the original packet, m is the fragment being created.
1172 * The fragments are linked off the m_nextpkt of the original
1173 * packet, which after processing serves as the first fragment.
984263bc 1174 */
f1f552f6
JH
1175 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
1176 struct ip *mhip; /* ip header on the fragment */
1177 struct mbuf *m;
5fe66e68 1178 int mhlen = sizeof(struct ip);
f1f552f6 1179
74f1caca 1180 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
f23061d4 1181 if (m == NULL) {
984263bc
MD
1182 error = ENOBUFS;
1183 ipstat.ips_odropped++;
f1f552f6 1184 goto done;
984263bc
MD
1185 }
1186 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
f1f552f6
JH
1187 /*
1188 * In the first mbuf, leave room for the link header, then
1189 * copy the original IP header including options. The payload
1190 * goes into an additional mbuf chain returned by m_copy().
1191 */
984263bc
MD
1192 m->m_data += max_linkhdr;
1193 mhip = mtod(m, struct ip *);
1194 *mhip = *ip;
5fe66e68
JH
1195 if (hlen > sizeof(struct ip)) {
1196 mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip);
984263bc
MD
1197 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
1198 }
1199 m->m_len = mhlen;
f1f552f6 1200 /* XXX do we need to add ip->ip_off below ? */
984263bc 1201 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
f1f552f6
JH
1202 if (off + len >= ip->ip_len) { /* last fragment */
1203 len = ip->ip_len - off;
1204 m->m_flags |= M_LASTFRAG;
1205 } else
984263bc
MD
1206 mhip->ip_off |= IP_MF;
1207 mhip->ip_len = htons((u_short)(len + mhlen));
1208 m->m_next = m_copy(m0, off, len);
f23061d4 1209 if (m->m_next == NULL) { /* copy failed */
f1f552f6 1210 m_free(m);
984263bc
MD
1211 error = ENOBUFS; /* ??? */
1212 ipstat.ips_odropped++;
f1f552f6 1213 goto done;
984263bc
MD
1214 }
1215 m->m_pkthdr.len = mhlen + len;
f23061d4 1216 m->m_pkthdr.rcvif = (struct ifnet *)NULL;
984263bc
MD
1217 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1218 mhip->ip_off = htons(mhip->ip_off);
1219 mhip->ip_sum = 0;
f1f552f6
JH
1220 if (sw_csum & CSUM_DELAY_IP)
1221 mhip->ip_sum = in_cksum(m, mhlen);
984263bc
MD
1222 *mnext = m;
1223 mnext = &m->m_nextpkt;
984263bc
MD
1224 }
1225 ipstat.ips_ofragments += nfrags;
1226
f1f552f6 1227 /* set first marker for fragment chain */
984263bc
MD
1228 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1229 m0->m_pkthdr.csum_data = nfrags;
1230
1231 /*
1232 * Update first fragment by trimming what's been copied out
f1f552f6 1233 * and updating header.
984263bc 1234 */
f1f552f6
JH
1235 m_adj(m0, hlen + firstlen - ip->ip_len);
1236 m0->m_pkthdr.len = hlen + firstlen;
1237 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
984263bc
MD
1238 ip->ip_off |= IP_MF;
1239 ip->ip_off = htons(ip->ip_off);
1240 ip->ip_sum = 0;
f1f552f6
JH
1241 if (sw_csum & CSUM_DELAY_IP)
1242 ip->ip_sum = in_cksum(m0, hlen);
984263bc 1243
984263bc 1244done:
f1f552f6
JH
1245 *m_frag = m0;
1246 return error;
984263bc
MD
1247}
1248
1249void
1250in_delayed_cksum(struct mbuf *m)
1251{
1252 struct ip *ip;
1253 u_short csum, offset;
1254
1255 ip = mtod(m, struct ip *);
1256 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1257 csum = in_cksum_skip(m, ip->ip_len, offset);
1258 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1259 csum = 0xffff;
1260 offset += m->m_pkthdr.csum_data; /* checksum offset */
1261
1262 if (offset + sizeof(u_short) > m->m_len) {
1263 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1264 m->m_len, offset, ip->ip_p);
1265 /*
1266 * XXX
1267 * this shouldn't happen, but if it does, the
1268 * correct behavior may be to insert the checksum
1269 * in the existing chain instead of rearranging it.
1270 */
1271 m = m_pullup(m, offset + sizeof(u_short));
1272 }
1273 *(u_short *)(m->m_data + offset) = csum;
1274}
1275
1276/*
1277 * Insert IP options into preformed packet.
1278 * Adjust IP destination as required for IP source routing,
1279 * as indicated by a non-zero in_addr at the start of the options.
1280 *
1281 * XXX This routine assumes that the packet has no options in place.
1282 */
1283static struct mbuf *
5fe66e68 1284ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
984263bc 1285{
2256ba69 1286 struct ipoption *p = mtod(opt, struct ipoption *);
984263bc 1287 struct mbuf *n;
2256ba69 1288 struct ip *ip = mtod(m, struct ip *);
984263bc
MD
1289 unsigned optlen;
1290
5fe66e68 1291 optlen = opt->m_len - sizeof p->ipopt_dst;
984263bc
MD
1292 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
1293 *phlen = 0;
1294 return (m); /* XXX should fail */
1295 }
1296 if (p->ipopt_dst.s_addr)
1297 ip->ip_dst = p->ipopt_dst;
1298 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
74f1caca 1299 MGETHDR(n, MB_DONTWAIT, MT_HEADER);
f23061d4 1300 if (n == NULL) {
984263bc
MD
1301 *phlen = 0;
1302 return (m);
1303 }
f23061d4 1304 n->m_pkthdr.rcvif = (struct ifnet *)NULL;
984263bc
MD
1305 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1306 m->m_len -= sizeof(struct ip);
1307 m->m_data += sizeof(struct ip);
1308 n->m_next = m;
1309 m = n;
1310 m->m_len = optlen + sizeof(struct ip);
1311 m->m_data += max_linkhdr;
f23061d4 1312 memcpy(mtod(m, void *), ip, sizeof(struct ip));
984263bc
MD
1313 } else {
1314 m->m_data -= optlen;
1315 m->m_len += optlen;
1316 m->m_pkthdr.len += optlen;
f23061d4 1317 ovbcopy(ip, mtod(m, caddr_t), sizeof(struct ip));
984263bc
MD
1318 }
1319 ip = mtod(m, struct ip *);
1320 bcopy(p->ipopt_list, ip + 1, optlen);
1321 *phlen = sizeof(struct ip) + optlen;
1322 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1323 ip->ip_len += optlen;
1324 return (m);
1325}
1326
1327/*
1328 * Copy options from ip to jp,
1329 * omitting those not copied during fragmentation.
1330 */
1331int
5fe66e68 1332ip_optcopy(struct ip *ip, struct ip *jp)
984263bc 1333{
2256ba69 1334 u_char *cp, *dp;
984263bc
MD
1335 int opt, optlen, cnt;
1336
1337 cp = (u_char *)(ip + 1);
1338 dp = (u_char *)(jp + 1);
5fe66e68 1339 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
984263bc
MD
1340 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1341 opt = cp[0];
1342 if (opt == IPOPT_EOL)
1343 break;
1344 if (opt == IPOPT_NOP) {
1345 /* Preserve for IP mcast tunnel's LSRR alignment. */
1346 *dp++ = IPOPT_NOP;
1347 optlen = 1;
1348 continue;
1349 }
1350
5fe66e68 1351 KASSERT(cnt >= IPOPT_OLEN + sizeof *cp,
984263bc
MD
1352 ("ip_optcopy: malformed ipv4 option"));
1353 optlen = cp[IPOPT_OLEN];
5fe66e68 1354 KASSERT(optlen >= IPOPT_OLEN + sizeof *cp && optlen <= cnt,
984263bc
MD
1355 ("ip_optcopy: malformed ipv4 option"));
1356
1357 /* bogus lengths should have been caught by ip_dooptions */
1358 if (optlen > cnt)
1359 optlen = cnt;
1360 if (IPOPT_COPIED(opt)) {
1361 bcopy(cp, dp, optlen);
1362 dp += optlen;
1363 }
1364 }
1365 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1366 *dp++ = IPOPT_EOL;
1367 return (optlen);
1368}
1369
1370/*
1371 * IP socket option processing.
1372 */
1373int
5fe66e68 1374ip_ctloutput(struct socket *so, struct sockopt *sopt)
984263bc 1375{
ed894f8c 1376 struct inpcb *inp = so->so_pcb;
984263bc
MD
1377 int error, optval;
1378
1379 error = optval = 0;
1380 if (sopt->sopt_level != IPPROTO_IP) {
1381 return (EINVAL);
1382 }
1383
1384 switch (sopt->sopt_dir) {
1385 case SOPT_SET:
1386 switch (sopt->sopt_name) {
1387 case IP_OPTIONS:
1388#ifdef notyet
1389 case IP_RETOPTS:
1390#endif
1391 {
1392 struct mbuf *m;
1393 if (sopt->sopt_valsize > MLEN) {
1394 error = EMSGSIZE;
1395 break;
1396 }
74f1caca 1397 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
f23061d4 1398 if (m == NULL) {
984263bc
MD
1399 error = ENOBUFS;
1400 break;
1401 }
1402 m->m_len = sopt->sopt_valsize;
1403 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1404 m->m_len);
5fe66e68 1405
984263bc
MD
1406 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1407 m));
1408 }
1409
1410 case IP_TOS:
1411 case IP_TTL:
1412 case IP_RECVOPTS:
1413 case IP_RECVRETOPTS:
1414 case IP_RECVDSTADDR:
1415 case IP_RECVIF:
1416 case IP_FAITH:
1417 error = sooptcopyin(sopt, &optval, sizeof optval,
1418 sizeof optval);
1419 if (error)
1420 break;
1421
1422 switch (sopt->sopt_name) {
1423 case IP_TOS:
1424 inp->inp_ip_tos = optval;
1425 break;
1426
1427 case IP_TTL:
1428 inp->inp_ip_ttl = optval;
1429 break;
1430#define OPTSET(bit) \
1431 if (optval) \
1432 inp->inp_flags |= bit; \
1433 else \
1434 inp->inp_flags &= ~bit;
1435
1436 case IP_RECVOPTS:
1437 OPTSET(INP_RECVOPTS);
1438 break;
1439
1440 case IP_RECVRETOPTS:
1441 OPTSET(INP_RECVRETOPTS);
1442 break;
1443
1444 case IP_RECVDSTADDR:
1445 OPTSET(INP_RECVDSTADDR);
1446 break;
1447
1448 case IP_RECVIF:
1449 OPTSET(INP_RECVIF);
1450 break;
1451
1452 case IP_FAITH:
1453 OPTSET(INP_FAITH);
1454 break;
1455 }
1456 break;
1457#undef OPTSET
1458
1459 case IP_MULTICAST_IF:
1460 case IP_MULTICAST_VIF:
1461 case IP_MULTICAST_TTL:
1462 case IP_MULTICAST_LOOP:
1463 case IP_ADD_MEMBERSHIP:
1464 case IP_DROP_MEMBERSHIP:
1465 error = ip_setmoptions(sopt, &inp->inp_moptions);
1466 break;
1467
1468 case IP_PORTRANGE:
1469 error = sooptcopyin(sopt, &optval, sizeof optval,
1470 sizeof optval);
1471 if (error)
1472 break;
1473
1474 switch (optval) {
1475 case IP_PORTRANGE_DEFAULT:
1476 inp->inp_flags &= ~(INP_LOWPORT);
1477 inp->inp_flags &= ~(INP_HIGHPORT);
1478 break;
1479
1480 case IP_PORTRANGE_HIGH:
1481 inp->inp_flags &= ~(INP_LOWPORT);
1482 inp->inp_flags |= INP_HIGHPORT;
1483 break;
1484
1485 case IP_PORTRANGE_LOW:
1486 inp->inp_flags &= ~(INP_HIGHPORT);
1487 inp->inp_flags |= INP_LOWPORT;
1488 break;
1489
1490 default:
1491 error = EINVAL;
1492 break;
1493 }
1494 break;
1495
1496#if defined(IPSEC) || defined(FAST_IPSEC)
1497 case IP_IPSEC_POLICY:
1498 {
1499 caddr_t req;
1500 size_t len = 0;
1501 int priv;
1502 struct mbuf *m;
1503 int optname;
1504
1505 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1506 break;
1507 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1508 break;
7b95be2a
MD
1509 priv = (sopt->sopt_td != NULL &&
1510 suser(sopt->sopt_td) != 0) ? 0 : 1;
984263bc
MD
1511 req = mtod(m, caddr_t);
1512 len = m->m_len;
1513 optname = sopt->sopt_name;
1514 error = ipsec4_set_policy(inp, optname, req, len, priv);
1515 m_freem(m);
1516 break;
1517 }
1518#endif /*IPSEC*/
1519
1520 default:
1521 error = ENOPROTOOPT;
1522 break;
1523 }
1524 break;
1525
1526 case SOPT_GET:
1527 switch (sopt->sopt_name) {
1528 case IP_OPTIONS:
1529 case IP_RETOPTS:
1530 if (inp->inp_options)
f23061d4 1531 error = sooptcopyout(sopt,
984263bc
MD
1532 mtod(inp->inp_options,
1533 char *),
1534 inp->inp_options->m_len);
1535 else
1536 sopt->sopt_valsize = 0;
1537 break;
1538
1539 case IP_TOS:
1540 case IP_TTL:
1541 case IP_RECVOPTS:
1542 case IP_RECVRETOPTS:
1543 case IP_RECVDSTADDR:
1544 case IP_RECVIF:
1545 case IP_PORTRANGE:
1546 case IP_FAITH:
1547 switch (sopt->sopt_name) {
1548
1549 case IP_TOS:
1550 optval = inp->inp_ip_tos;
1551 break;
1552
1553 case IP_TTL:
1554 optval = inp->inp_ip_ttl;
1555 break;
1556
1557#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1558
1559 case IP_RECVOPTS:
1560 optval = OPTBIT(INP_RECVOPTS);
1561 break;
1562
1563 case IP_RECVRETOPTS:
1564 optval = OPTBIT(INP_RECVRETOPTS);
1565 break;
1566
1567 case IP_RECVDSTADDR:
1568 optval = OPTBIT(INP_RECVDSTADDR);
1569 break;
1570
1571 case IP_RECVIF:
1572 optval = OPTBIT(INP_RECVIF);
1573 break;
1574
1575 case IP_PORTRANGE:
1576 if (inp->inp_flags & INP_HIGHPORT)
1577 optval = IP_PORTRANGE_HIGH;
1578 else if (inp->inp_flags & INP_LOWPORT)
1579 optval = IP_PORTRANGE_LOW;
1580 else
1581 optval = 0;
1582 break;
1583
1584 case IP_FAITH:
1585 optval = OPTBIT(INP_FAITH);
1586 break;
1587 }
1588 error = sooptcopyout(sopt, &optval, sizeof optval);
1589 break;
1590
1591 case IP_MULTICAST_IF:
1592 case IP_MULTICAST_VIF:
1593 case IP_MULTICAST_TTL:
1594 case IP_MULTICAST_LOOP:
1595 case IP_ADD_MEMBERSHIP:
1596 case IP_DROP_MEMBERSHIP:
1597 error = ip_getmoptions(sopt, inp->inp_moptions);
1598 break;
1599
1600#if defined(IPSEC) || defined(FAST_IPSEC)
1601 case IP_IPSEC_POLICY:
1602 {
1603 struct mbuf *m = NULL;
1604 caddr_t req = NULL;
1605 size_t len = 0;
1606
f23061d4 1607 if (m != NULL) {
984263bc
MD
1608 req = mtod(m, caddr_t);
1609 len = m->m_len;
1610 }
ed894f8c 1611 error = ipsec4_get_policy(so->so_pcb, req, len, &m);
984263bc
MD
1612 if (error == 0)
1613 error = soopt_mcopyout(sopt, m); /* XXX */
1614 if (error == 0)
1615 m_freem(m);
1616 break;
1617 }
1618#endif /*IPSEC*/
1619
1620 default:
1621 error = ENOPROTOOPT;
1622 break;
1623 }
1624 break;
1625 }
1626 return (error);
1627}
1628
1629/*
1630 * Set up IP options in pcb for insertion in output packets.
1631 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1632 * with destination address if source routed.
1633 */
1634static int
f23061d4 1635ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
984263bc 1636{
2256ba69
RG
1637 int cnt, optlen;
1638 u_char *cp;
984263bc
MD
1639 u_char opt;
1640
1641 /* turn off any old options */
1642 if (*pcbopt)
f23061d4 1643 m_free(*pcbopt);
984263bc 1644 *pcbopt = 0;
f23061d4 1645 if (m == NULL || m->m_len == 0) {
984263bc
MD
1646 /*
1647 * Only turning off any previous options.
1648 */
f23061d4
JH
1649 if (m != NULL)
1650 m_free(m);
984263bc
MD
1651 return (0);
1652 }
1653
1654 if (m->m_len % sizeof(int32_t))
1655 goto bad;
1656 /*
1657 * IP first-hop destination address will be stored before
1658 * actual options; move other options back
1659 * and clear it when none present.
1660 */
1661 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1662 goto bad;
1663 cnt = m->m_len;
1664 m->m_len += sizeof(struct in_addr);
1665 cp = mtod(m, u_char *) + sizeof(struct in_addr);
f23061d4 1666 ovbcopy(mtod(m, caddr_t), cp, cnt);
984263bc
MD
1667 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1668
1669 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1670 opt = cp[IPOPT_OPTVAL];
1671 if (opt == IPOPT_EOL)
1672 break;
1673 if (opt == IPOPT_NOP)
1674 optlen = 1;
1675 else {
5fe66e68 1676 if (cnt < IPOPT_OLEN + sizeof *cp)
984263bc
MD
1677 goto bad;
1678 optlen = cp[IPOPT_OLEN];
5fe66e68 1679 if (optlen < IPOPT_OLEN + sizeof *cp || optlen > cnt)
984263bc
MD
1680 goto bad;
1681 }
1682 switch (opt) {
1683
1684 default:
1685 break;
1686
1687 case IPOPT_LSRR:
1688 case IPOPT_SSRR:
1689 /*
1690 * user process specifies route as:
1691 * ->A->B->C->D
1692 * D must be our final destination (but we can't
1693 * check that since we may not have connected yet).
1694 * A is first hop destination, which doesn't appear in
1695 * actual IP option, but is stored before the options.
1696 */
1697 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1698 goto bad;
1699 m->m_len -= sizeof(struct in_addr);
1700 cnt -= sizeof(struct in_addr);
1701 optlen -= sizeof(struct in_addr);
1702 cp[IPOPT_OLEN] = optlen;
1703 /*
1704 * Move first hop before start of options.
1705 */
f23061d4 1706 bcopy(&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
5fe66e68 1707 sizeof(struct in_addr));
984263bc
MD
1708 /*
1709 * Then copy rest of options back
1710 * to close up the deleted entry.
1711 */
f23061d4
JH
1712 ovbcopy(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr),
1713 &cp[IPOPT_OFFSET+1],
1714 cnt - (IPOPT_MINOFF - 1));
984263bc
MD
1715 break;
1716 }
1717 }
1718 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1719 goto bad;
1720 *pcbopt = m;
1721 return (0);
1722
1723bad:
f23061d4 1724 m_free(m);
984263bc
MD
1725 return (EINVAL);
1726}
1727
1728/*
1729 * XXX
1730 * The whole multicast option thing needs to be re-thought.
1731 * Several of these options are equally applicable to non-multicast
1732 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1733 * standard option (IP_TTL).
1734 */
1735
1736/*
1737 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1738 */
1739static struct ifnet *
f23061d4 1740ip_multicast_if(struct in_addr *a, int *ifindexp)
984263bc
MD
1741{
1742 int ifindex;
1743 struct ifnet *ifp;
1744
1745 if (ifindexp)
1746 *ifindexp = 0;
1747 if (ntohl(a->s_addr) >> 24 == 0) {
1748 ifindex = ntohl(a->s_addr) & 0xffffff;
1749 if (ifindex < 0 || if_index < ifindex)
1750 return NULL;
1751 ifp = ifindex2ifnet[ifindex];
1752 if (ifindexp)
1753 *ifindexp = ifindex;
1754 } else {
1755 INADDR_TO_IFP(*a, ifp);
1756 }
1757 return ifp;
1758}
1759
1760/*
1761 * Set the IP multicast options in response to user setsockopt().
1762 */
1763static int
f23061d4 1764ip_setmoptions(struct sockopt *sopt, struct ip_moptions **imop)
984263bc
MD
1765{
1766 int error = 0;
1767 int i;
1768 struct in_addr addr;
1769 struct ip_mreq mreq;
1770 struct ifnet *ifp;
1771 struct ip_moptions *imo = *imop;
984263bc 1772 int ifindex;
984263bc
MD
1773
1774 if (imo == NULL) {
1775 /*
1776 * No multicast option buffer attached to the pcb;
1777 * allocate one and initialize to default values.
1778 */
5fe66e68 1779 imo = malloc(sizeof *imo, M_IPMOPTS, M_WAITOK);
984263bc
MD
1780
1781 if (imo == NULL)
1782 return (ENOBUFS);
1783 *imop = imo;
1784 imo->imo_multicast_ifp = NULL;
1785 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1786 imo->imo_multicast_vif = -1;
1787 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1788 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1789 imo->imo_num_memberships = 0;
1790 }
1791
1792 switch (sopt->sopt_name) {
1793 /* store an index number for the vif you wanna use in the send */
1794 case IP_MULTICAST_VIF:
1795 if (legal_vif_num == 0) {
1796 error = EOPNOTSUPP;
1797 break;
1798 }
1799 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
1800 if (error)
1801 break;
1802 if (!legal_vif_num(i) && (i != -1)) {
1803 error = EINVAL;
1804 break;
1805 }
1806 imo->imo_multicast_vif = i;
1807 break;
1808
1809 case IP_MULTICAST_IF:
1810 /*
1811 * Select the interface for outgoing multicast packets.
1812 */
1813 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr);
1814 if (error)
1815 break;
1816 /*
1817 * INADDR_ANY is used to remove a previous selection.
1818 * When no interface is selected, a default one is
1819 * chosen every time a multicast packet is sent.
1820 */
1821 if (addr.s_addr == INADDR_ANY) {
1822 imo->imo_multicast_ifp = NULL;
1823 break;
1824 }
1825 /*
1826 * The selected interface is identified by its local
1827 * IP address. Find the interface and confirm that
1828 * it supports multicasting.
1829 */
1cae611f 1830 crit_enter();
984263bc 1831 ifp = ip_multicast_if(&addr, &ifindex);
f23061d4 1832 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
1cae611f 1833 crit_exit();
984263bc
MD
1834 error = EADDRNOTAVAIL;
1835 break;
1836 }
1837 imo->imo_multicast_ifp = ifp;
1838 if (ifindex)
1839 imo->imo_multicast_addr = addr;
1840 else
1841 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1cae611f 1842 crit_exit();
984263bc
MD
1843 break;
1844
1845 case IP_MULTICAST_TTL:
1846 /*
1847 * Set the IP time-to-live for outgoing multicast packets.
1848 * The original multicast API required a char argument,
1849 * which is inconsistent with the rest of the socket API.
1850 * We allow either a char or an int.
1851 */
1852 if (sopt->sopt_valsize == 1) {
1853 u_char ttl;
1854 error = sooptcopyin(sopt, &ttl, 1, 1);
1855 if (error)
1856 break;
1857 imo->imo_multicast_ttl = ttl;
1858 } else {
1859 u_int ttl;
5fe66e68 1860 error = sooptcopyin(sopt, &ttl, sizeof ttl, sizeof ttl);
984263bc
MD
1861 if (error)
1862 break;
1863 if (ttl > 255)
1864 error = EINVAL;
1865 else
1866 imo->imo_multicast_ttl = ttl;
1867 }
1868 break;
1869
1870 case IP_MULTICAST_LOOP:
1871 /*
1872 * Set the loopback flag for outgoing multicast packets.
1873 * Must be zero or one. The original multicast API required a
1874 * char argument, which is inconsistent with the rest
1875 * of the socket API. We allow either a char or an int.
1876 */
1877 if (sopt->sopt_valsize == 1) {
1878 u_char loop;
5fe66e68 1879
984263bc
MD
1880 error = sooptcopyin(sopt, &loop, 1, 1);
1881 if (error)
1882 break;
1883 imo->imo_multicast_loop = !!loop;
1884 } else {
1885 u_int loop;
5fe66e68 1886
984263bc
MD
1887 error = sooptcopyin(sopt, &loop, sizeof loop,
1888 sizeof loop);
1889 if (error)
1890 break;
1891 imo->imo_multicast_loop = !!loop;
1892 }
1893 break;
1894
1895 case IP_ADD_MEMBERSHIP:
1896 /*
1897 * Add a multicast group membership.
1898 * Group must be a valid IP multicast address.
1899 */
1900 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1901 if (error)
1902 break;
1903
1904 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1905 error = EINVAL;
1906 break;
1907 }
1cae611f 1908 crit_enter();
984263bc
MD
1909 /*
1910 * If no interface address was provided, use the interface of
1911 * the route to the given multicast address.
1912 */
1913 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1ce00f4f
JH
1914 struct sockaddr_in dst;
1915 struct rtentry *rt;
1916
1917 bzero(&dst, sizeof(struct sockaddr_in));
1918 dst.sin_len = sizeof(struct sockaddr_in);
1919 dst.sin_family = AF_INET;
1920 dst.sin_addr = mreq.imr_multiaddr;
1921 rt = rtlookup((struct sockaddr *)&dst);
1922 if (rt == NULL) {
984263bc 1923 error = EADDRNOTAVAIL;
1cae611f 1924 crit_exit();
984263bc
MD
1925 break;
1926 }
1ce00f4f
JH
1927 --rt->rt_refcnt;
1928 ifp = rt->rt_ifp;
1929 } else {
984263bc
MD
1930 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1931 }
1932
1933 /*
1934 * See if we found an interface, and confirm that it
1935 * supports multicast.
1936 */
f23061d4 1937 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
984263bc 1938 error = EADDRNOTAVAIL;
1cae611f 1939 crit_exit();
984263bc
MD
1940 break;
1941 }
1942 /*
1943 * See if the membership already exists or if all the
1944 * membership slots are full.
1945 */
1946 for (i = 0; i < imo->imo_num_memberships; ++i) {
1947 if (imo->imo_membership[i]->inm_ifp == ifp &&
1948 imo->imo_membership[i]->inm_addr.s_addr
1949 == mreq.imr_multiaddr.s_addr)
1950 break;
1951 }
1952 if (i < imo->imo_num_memberships) {
1953 error = EADDRINUSE;
1cae611f 1954 crit_exit();
984263bc
MD
1955 break;
1956 }
1957 if (i == IP_MAX_MEMBERSHIPS) {
1958 error = ETOOMANYREFS;
1cae611f 1959 crit_exit();
984263bc
MD
1960 break;
1961 }
1962 /*
1963 * Everything looks good; add a new record to the multicast
1964 * address list for the given interface.
1965 */
1966 if ((imo->imo_membership[i] =
5fe66e68 1967 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
984263bc 1968 error = ENOBUFS;
1cae611f 1969 crit_exit();
984263bc
MD
1970 break;
1971 }
1972 ++imo->imo_num_memberships;
1cae611f 1973 crit_exit();
984263bc
MD
1974 break;
1975
1976 case IP_DROP_MEMBERSHIP:
1977 /*
1978 * Drop a multicast group membership.
1979 * Group must be a valid IP multicast address.
1980 */
1981 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1982 if (error)
1983 break;
1984
1985 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1986 error = EINVAL;
1987 break;
1988 }
1989
1cae611f 1990 crit_enter();
984263bc
MD
1991 /*
1992 * If an interface address was specified, get a pointer
1993 * to its ifnet structure.
1994 */
1995 if (mreq.imr_interface.s_addr == INADDR_ANY)
1996 ifp = NULL;
1997 else {
1998 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1999 if (ifp == NULL) {
2000 error = EADDRNOTAVAIL;
1cae611f 2001 crit_exit();
984263bc
MD
2002 break;
2003 }
2004 }
2005 /*
2006 * Find the membership in the membership array.
2007 */
2008 for (i = 0; i < imo->imo_num_memberships; ++i) {
2009 if ((ifp == NULL ||
2010 imo->imo_membership[i]->inm_ifp == ifp) &&
5fe66e68
JH
2011 imo->imo_membership[i]->inm_addr.s_addr ==
2012 mreq.imr_multiaddr.s_addr)
984263bc
MD
2013 break;
2014 }
2015 if (i == imo->imo_num_memberships) {
2016 error = EADDRNOTAVAIL;
1cae611f 2017 crit_exit();
984263bc
MD
2018 break;
2019 }
2020 /*
2021 * Give up the multicast address record to which the
2022 * membership points.
2023 */
2024 in_delmulti(imo->imo_membership[i]);
2025 /*
2026 * Remove the gap in the membership array.
2027 */
2028 for (++i; i < imo->imo_num_memberships; ++i)
2029 imo->imo_membership[i-1] = imo->imo_membership[i];
2030 --imo->imo_num_memberships;
1cae611f 2031 crit_exit();
984263bc
MD
2032 break;
2033
2034 default:
2035 error = EOPNOTSUPP;
2036 break;
2037 }
2038
2039 /*
2040 * If all options have default values, no need to keep the mbuf.
2041 */
2042 if (imo->imo_multicast_ifp == NULL &&
2043 imo->imo_multicast_vif == -1 &&
2044 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2045 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2046 imo->imo_num_memberships == 0) {
2047 free(*imop, M_IPMOPTS);
2048 *imop = NULL;
2049 }
2050
2051 return (error);
2052}
2053
2054/*
2055 * Return the IP multicast options in response to user getsockopt().
2056 */
2057static int
f23061d4 2058ip_getmoptions(struct sockopt *sopt, struct ip_moptions *imo)
984263bc
MD
2059{
2060 struct in_addr addr;
2061 struct in_ifaddr *ia;
2062 int error, optval;
2063 u_char coptval;
2064
2065 error = 0;
2066 switch (sopt->sopt_name) {
f23061d4 2067 case IP_MULTICAST_VIF:
984263bc
MD
2068 if (imo != NULL)
2069 optval = imo->imo_multicast_vif;
2070 else
2071 optval = -1;
2072 error = sooptcopyout(sopt, &optval, sizeof optval);
2073 break;
2074
2075 case IP_MULTICAST_IF:
2076 if (imo == NULL || imo->imo_multicast_ifp == NULL)
2077 addr.s_addr = INADDR_ANY;
2078 else if (imo->imo_multicast_addr.s_addr) {
2079 /* return the value user has set */
2080 addr = imo->imo_multicast_addr;
2081 } else {
2082 IFP_TO_IA(imo->imo_multicast_ifp, ia);
2083 addr.s_addr = (ia == NULL) ? INADDR_ANY
2084 : IA_SIN(ia)->sin_addr.s_addr;
2085 }
2086 error = sooptcopyout(sopt, &addr, sizeof addr);
2087 break;
2088
2089 case IP_MULTICAST_TTL:
f23061d4 2090 if (imo == NULL)
984263bc
MD
2091 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
2092 else
2093 optval = coptval = imo->imo_multicast_ttl;
2094 if (sopt->sopt_valsize == 1)
2095 error = sooptcopyout(sopt, &coptval, 1);
2096 else
2097 error = sooptcopyout(sopt, &optval, sizeof optval);
2098 break;
2099
2100 case IP_MULTICAST_LOOP:
f23061d4 2101 if (imo == NULL)
984263bc
MD
2102 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
2103 else
2104 optval = coptval = imo->imo_multicast_loop;
2105 if (sopt->sopt_valsize == 1)
2106 error = sooptcopyout(sopt, &coptval, 1);
2107 else
2108 error = sooptcopyout(sopt, &optval, sizeof optval);
2109 break;
2110
2111 default:
2112 error = ENOPROTOOPT;
2113 break;
2114 }
2115 return (error);
2116}
2117
2118/*
2119 * Discard the IP multicast options.
2120 */
2121void
f23061d4 2122ip_freemoptions(struct ip_moptions *imo)
984263bc 2123{
2256ba69 2124 int i;
984263bc
MD
2125
2126 if (imo != NULL) {
2127 for (i = 0; i < imo->imo_num_memberships; ++i)
2128 in_delmulti(imo->imo_membership[i]);
2129 free(imo, M_IPMOPTS);
2130 }
2131}
2132
2133/*
2134 * Routine called from ip_output() to loop back a copy of an IP multicast
2135 * packet to the input queue of a specified interface. Note that this
2136 * calls the output routine of the loopback "driver", but with an interface
2137 * pointer that might NOT be a loopback interface -- evil, but easier than
2138 * replicating that code here.
2139 */
2140static void
5fe66e68
JH
2141ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
2142 int hlen)
984263bc 2143{
2256ba69 2144 struct ip *ip;
984263bc
MD
2145 struct mbuf *copym;
2146
f23061d4 2147 copym = m_copypacket(m, MB_DONTWAIT);
984263bc
MD
2148 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2149 copym = m_pullup(copym, hlen);
2150 if (copym != NULL) {
f23061d4 2151 /*
7db7d2da
MD
2152 * if the checksum hasn't been computed, mark it as valid
2153 */
2154 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2155 in_delayed_cksum(copym);
2156 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2157 copym->m_pkthdr.csum_flags |=
2158 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2159 copym->m_pkthdr.csum_data = 0xffff;
2160 }
984263bc
MD
2161 /*
2162 * We don't bother to fragment if the IP length is greater
2163 * than the interface's MTU. Can this possibly matter?
2164 */
2165 ip = mtod(copym, struct ip *);
2166 ip->ip_len = htons(ip->ip_len);
2167 ip->ip_off = htons(ip->ip_off);
2168 ip->ip_sum = 0;
2169 if (ip->ip_vhl == IP_VHL_BORING) {
2170 ip->ip_sum = in_cksum_hdr(ip);
2171 } else {
2172 ip->ip_sum = in_cksum(copym, hlen);
2173 }
2174 /*
2175 * NB:
2176 * It's not clear whether there are any lingering
2177 * reentrancy problems in other areas which might
2178 * be exposed by using ip_input directly (in
2179 * particular, everything which modifies the packet
2180 * in-place). Yet another option is using the
2181 * protosw directly to deliver the looped back
2182 * packet. For the moment, we'll err on the side
2183 * of safety by using if_simloop().
2184 */
2185#if 1 /* XXX */
2186 if (dst->sin_family != AF_INET) {
2187 printf("ip_mloopback: bad address family %d\n",
2188 dst->sin_family);
2189 dst->sin_family = AF_INET;
2190 }
2191#endif
2192
2193#ifdef notdef
2194 copym->m_pkthdr.rcvif = ifp;
2195 ip_input(copym);
2196#else
984263bc
MD
2197 if_simloop(ifp, copym, dst->sin_family, 0);
2198#endif
2199 }
2200}