xref: /netbsd/sys/netinet/ip_output.c (revision 5e3074af)
1 /*	$NetBSD: ip_output.c,v 1.326 2023/04/19 22:00:18 mlelstv Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1998 The NetBSD Foundation, Inc.
34  * All rights reserved.
35  *
36  * This code is derived from software contributed to The NetBSD Foundation
37  * by Public Access Networks Corporation ("Panix").  It was developed under
38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
50  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
51  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
52  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
53  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
54  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
55  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
56  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
57  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
58  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
59  * POSSIBILITY OF SUCH DAMAGE.
60  */
61 
62 /*
63  * Copyright (c) 1982, 1986, 1988, 1990, 1993
64  *	The Regents of the University of California.  All rights reserved.
65  *
66  * Redistribution and use in source and binary forms, with or without
67  * modification, are permitted provided that the following conditions
68  * are met:
69  * 1. Redistributions of source code must retain the above copyright
70  *    notice, this list of conditions and the following disclaimer.
71  * 2. Redistributions in binary form must reproduce the above copyright
72  *    notice, this list of conditions and the following disclaimer in the
73  *    documentation and/or other materials provided with the distribution.
74  * 3. Neither the name of the University nor the names of its contributors
75  *    may be used to endorse or promote products derived from this software
76  *    without specific prior written permission.
77  *
78  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
79  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
80  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
81  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
82  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
83  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
84  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
85  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
86  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
87  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
88  * SUCH DAMAGE.
89  *
90  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
91  */
92 
93 #include <sys/cdefs.h>
94 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.326 2023/04/19 22:00:18 mlelstv Exp $");
95 
96 #ifdef _KERNEL_OPT
97 #include "opt_inet.h"
98 #include "opt_ipsec.h"
99 #include "opt_mrouting.h"
100 #include "opt_net_mpsafe.h"
101 #include "opt_mpls.h"
102 #endif
103 
104 #include "arp.h"
105 
106 #include <sys/param.h>
107 #include <sys/kmem.h>
108 #include <sys/mbuf.h>
109 #include <sys/socket.h>
110 #include <sys/socketvar.h>
111 #include <sys/kauth.h>
112 #include <sys/systm.h>
113 #include <sys/syslog.h>
114 
115 #include <net/if.h>
116 #include <net/if_types.h>
117 #include <net/route.h>
118 #include <net/pfil.h>
119 
120 #include <netinet/in.h>
121 #include <netinet/in_systm.h>
122 #include <netinet/ip.h>
123 #include <netinet/in_pcb.h>
124 #include <netinet/in_var.h>
125 #include <netinet/ip_var.h>
126 #include <netinet/ip_private.h>
127 #include <netinet/in_offload.h>
128 #include <netinet/portalgo.h>
129 #include <netinet/udp.h>
130 #include <netinet/udp_var.h>
131 
132 #ifdef INET6
133 #include <netinet6/ip6_var.h>
134 #endif
135 
136 #ifdef MROUTING
137 #include <netinet/ip_mroute.h>
138 #endif
139 
140 #ifdef IPSEC
141 #include <netipsec/ipsec.h>
142 #include <netipsec/key.h>
143 #endif
144 
145 #ifdef MPLS
146 #include <netmpls/mpls.h>
147 #include <netmpls/mpls_var.h>
148 #endif
149 
150 static int ip_pcbopts(struct inpcb *, const struct sockopt *);
151 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
152 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
153 static void ip_mloopback(struct ifnet *, struct mbuf *,
154     const struct sockaddr_in *);
155 static int ip_ifaddrvalid(const struct in_ifaddr *);
156 
157 extern pfil_head_t *inet_pfil_hook;			/* XXX */
158 
159 int ip_do_loopback_cksum = 0;
160 
161 static int
ip_mark_mpls(struct ifnet * const ifp,struct mbuf * const m,const struct rtentry * rt)162 ip_mark_mpls(struct ifnet * const ifp, struct mbuf * const m,
163     const struct rtentry *rt)
164 {
165 	int error = 0;
166 #ifdef MPLS
167 	union mpls_shim msh;
168 
169 	if (rt == NULL || rt_gettag(rt) == NULL ||
170 	    rt_gettag(rt)->sa_family != AF_MPLS ||
171 	    (m->m_flags & (M_MCAST | M_BCAST)) != 0 ||
172 	    ifp->if_type != IFT_ETHER)
173 		return 0;
174 
175 	msh.s_addr = MPLS_GETSADDR(rt);
176 	if (msh.shim.label != MPLS_LABEL_IMPLNULL) {
177 		struct m_tag *mtag;
178 		/*
179 		 * XXX tentative solution to tell ether_output
180 		 * it's MPLS. Need some more efficient solution.
181 		 */
182 		mtag = m_tag_get(PACKET_TAG_MPLS,
183 		    sizeof(int) /* dummy */,
184 		    M_NOWAIT);
185 		if (mtag == NULL)
186 			return ENOMEM;
187 		m_tag_prepend(m, mtag);
188 	}
189 #endif
190 	return error;
191 }
192 
193 /*
194  * Send an IP packet to a host.
195  */
196 int
ip_if_output(struct ifnet * const ifp,struct mbuf * const m,const struct sockaddr * const dst,const struct rtentry * rt)197 ip_if_output(struct ifnet * const ifp, struct mbuf * const m,
198     const struct sockaddr * const dst, const struct rtentry *rt)
199 {
200 	int error = 0;
201 
202 	if (rt != NULL) {
203 		error = rt_check_reject_route(rt, ifp);
204 		if (error != 0) {
205 			IP_STATINC(IP_STAT_RTREJECT);
206 			m_freem(m);
207 			return error;
208 		}
209 	}
210 
211 	error = ip_mark_mpls(ifp, m, rt);
212 	if (error != 0) {
213 		m_freem(m);
214 		return error;
215 	}
216 
217 	error = if_output_lock(ifp, ifp, m, dst, rt);
218 
219 	return error;
220 }
221 
222 /*
223  * IP output.  The packet in mbuf chain m contains a skeletal IP
224  * header (with len, off, ttl, proto, tos, src, dst).
225  * The mbuf chain containing the packet will be freed.
226  * The mbuf opt, if present, will not be freed.
227  */
228 int
ip_output(struct mbuf * m0,struct mbuf * opt,struct route * ro,int flags,struct ip_moptions * imo,struct inpcb * inp)229 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, int flags,
230     struct ip_moptions *imo, struct inpcb *inp)
231 {
232 	struct rtentry *rt;
233 	struct ip *ip;
234 	struct ifnet *ifp, *mifp = NULL;
235 	struct mbuf *m = m0;
236 	int len, hlen, error = 0;
237 	struct route iproute;
238 	const struct sockaddr_in *dst;
239 	struct in_ifaddr *ia = NULL;
240 	struct ifaddr *ifa;
241 	int isbroadcast;
242 	int sw_csum;
243 	u_long mtu;
244 	bool natt_frag = false;
245 	bool rtmtu_nolock;
246 	union {
247 		struct sockaddr		sa;
248 		struct sockaddr_in	sin;
249 	} udst, usrc;
250 	struct sockaddr *rdst = &udst.sa;	/* real IP destination, as
251 						 * opposed to the nexthop
252 						 */
253 	struct psref psref, psref_ia;
254 	int bound;
255 	bool bind_need_restore = false;
256 	const struct sockaddr *sa;
257 
258 	len = 0;
259 
260 	MCLAIM(m, &ip_tx_mowner);
261 
262 	KASSERT((m->m_flags & M_PKTHDR) != 0);
263 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 0);
264 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) !=
265 	    (M_CSUM_TCPv4|M_CSUM_UDPv4));
266 	KASSERT(m->m_len >= sizeof(struct ip));
267 
268 	hlen = sizeof(struct ip);
269 	if (opt) {
270 		m = ip_insertoptions(m, opt, &len);
271 		hlen = len;
272 	}
273 	ip = mtod(m, struct ip *);
274 
275 	/*
276 	 * Fill in IP header.
277 	 */
278 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
279 		ip->ip_v = IPVERSION;
280 		ip->ip_off = htons(0);
281 		/* ip->ip_id filled in after we find out source ia */
282 		ip->ip_hl = hlen >> 2;
283 		IP_STATINC(IP_STAT_LOCALOUT);
284 	} else {
285 		hlen = ip->ip_hl << 2;
286 	}
287 
288 	/*
289 	 * Route packet.
290 	 */
291 	if (ro == NULL) {
292 		memset(&iproute, 0, sizeof(iproute));
293 		ro = &iproute;
294 	}
295 	sockaddr_in_init(&udst.sin, &ip->ip_dst, 0);
296 	dst = satocsin(rtcache_getdst(ro));
297 
298 	/*
299 	 * If there is a cached route, check that it is to the same
300 	 * destination and is still up.  If not, free it and try again.
301 	 * The address family should also be checked in case of sharing
302 	 * the cache with IPv6.
303 	 */
304 	if (dst && (dst->sin_family != AF_INET ||
305 	    !in_hosteq(dst->sin_addr, ip->ip_dst)))
306 		rtcache_free(ro);
307 
308 	/* XXX must be before rtcache operations */
309 	bound = curlwp_bind();
310 	bind_need_restore = true;
311 
312 	if ((rt = rtcache_validate(ro)) == NULL &&
313 	    (rt = rtcache_update(ro, 1)) == NULL) {
314 		dst = &udst.sin;
315 		error = rtcache_setdst(ro, &udst.sa);
316 		if (error != 0) {
317 			IP_STATINC(IP_STAT_ODROPPED);
318 			goto bad;
319 		}
320 	}
321 
322 	/*
323 	 * If routing to interface only, short circuit routing lookup.
324 	 */
325 	if (flags & IP_ROUTETOIF) {
326 		ifa = ifa_ifwithladdr_psref(sintocsa(dst), &psref_ia);
327 		if (ifa == NULL) {
328 			IP_STATINC(IP_STAT_NOROUTE);
329 			error = ENETUNREACH;
330 			goto bad;
331 		}
332 		/* ia is already referenced by psref_ia */
333 		ia = ifatoia(ifa);
334 
335 		ifp = ia->ia_ifp;
336 		mtu = ifp->if_mtu;
337 		ip->ip_ttl = 1;
338 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
339 	} else if (((IN_MULTICAST(ip->ip_dst.s_addr) ||
340 	    ip->ip_dst.s_addr == INADDR_BROADCAST) ||
341 	    (flags & IP_ROUTETOIFINDEX)) &&
342 	    imo != NULL && imo->imo_multicast_if_index != 0) {
343 		ifp = mifp = if_get_byindex(imo->imo_multicast_if_index, &psref);
344 		if (ifp == NULL) {
345 			IP_STATINC(IP_STAT_NOROUTE);
346 			error = ENETUNREACH;
347 			goto bad;
348 		}
349 		mtu = ifp->if_mtu;
350 		ia = in_get_ia_from_ifp_psref(ifp, &psref_ia);
351 		if (IN_MULTICAST(ip->ip_dst.s_addr) ||
352 		    ip->ip_dst.s_addr == INADDR_BROADCAST) {
353 			isbroadcast = 0;
354 		} else {
355 			/* IP_ROUTETOIFINDEX */
356 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
357 			if ((isbroadcast == 0) && ((ifp->if_flags &
358 			    (IFF_LOOPBACK | IFF_POINTOPOINT)) == 0) &&
359 			    (in_direct(dst->sin_addr, ifp) == 0)) {
360 				/* gateway address required */
361 				if (rt == NULL)
362 					rt = rtcache_init(ro);
363 				if (rt == NULL || rt->rt_ifp != ifp) {
364 					IP_STATINC(IP_STAT_NOROUTE);
365 					error = EHOSTUNREACH;
366 					goto bad;
367 				}
368 				rt->rt_use++;
369 				if (rt->rt_flags & RTF_GATEWAY)
370 					dst = satosin(rt->rt_gateway);
371 				if (rt->rt_flags & RTF_HOST)
372 					isbroadcast =
373 					    rt->rt_flags & RTF_BROADCAST;
374 			}
375 		}
376 	} else {
377 		if (rt == NULL)
378 			rt = rtcache_init(ro);
379 		if (rt == NULL) {
380 			IP_STATINC(IP_STAT_NOROUTE);
381 			error = EHOSTUNREACH;
382 			goto bad;
383 		}
384 		if (ifa_is_destroying(rt->rt_ifa)) {
385 			rtcache_unref(rt, ro);
386 			rt = NULL;
387 			IP_STATINC(IP_STAT_NOROUTE);
388 			error = EHOSTUNREACH;
389 			goto bad;
390 		}
391 		ifa_acquire(rt->rt_ifa, &psref_ia);
392 		ia = ifatoia(rt->rt_ifa);
393 		ifp = rt->rt_ifp;
394 		if ((mtu = rt->rt_rmx.rmx_mtu) == 0)
395 			mtu = ifp->if_mtu;
396 		rt->rt_use++;
397 		if (rt->rt_flags & RTF_GATEWAY)
398 			dst = satosin(rt->rt_gateway);
399 		if (rt->rt_flags & RTF_HOST)
400 			isbroadcast = rt->rt_flags & RTF_BROADCAST;
401 		else
402 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
403 	}
404 	rtmtu_nolock = rt && (rt->rt_rmx.rmx_locks & RTV_MTU) == 0;
405 
406 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
407 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
408 		bool inmgroup;
409 
410 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
411 		    M_BCAST : M_MCAST;
412 		/*
413 		 * See if the caller provided any multicast options
414 		 */
415 		if (imo != NULL)
416 			ip->ip_ttl = imo->imo_multicast_ttl;
417 		else
418 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
419 
420 		/*
421 		 * if we don't know the outgoing ifp yet, we can't generate
422 		 * output
423 		 */
424 		if (!ifp) {
425 			IP_STATINC(IP_STAT_NOROUTE);
426 			error = ENETUNREACH;
427 			goto bad;
428 		}
429 
430 		/*
431 		 * If the packet is multicast or broadcast, confirm that
432 		 * the outgoing interface can transmit it.
433 		 */
434 		if (((m->m_flags & M_MCAST) &&
435 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
436 		    ((m->m_flags & M_BCAST) &&
437 		     (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0))  {
438 			IP_STATINC(IP_STAT_NOROUTE);
439 			error = ENETUNREACH;
440 			goto bad;
441 		}
442 		/*
443 		 * If source address not specified yet, use an address
444 		 * of outgoing interface.
445 		 */
446 		if (in_nullhost(ip->ip_src)) {
447 			struct in_ifaddr *xia;
448 			struct ifaddr *xifa;
449 			struct psref _psref;
450 
451 			xia = in_get_ia_from_ifp_psref(ifp, &_psref);
452 			if (!xia) {
453 				IP_STATINC(IP_STAT_IFNOADDR);
454 				error = EADDRNOTAVAIL;
455 				goto bad;
456 			}
457 			xifa = &xia->ia_ifa;
458 			if (xifa->ifa_getifa != NULL) {
459 				ia4_release(xia, &_psref);
460 				/* FIXME ifa_getifa is NOMPSAFE */
461 				xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst));
462 				if (xia == NULL) {
463 					IP_STATINC(IP_STAT_IFNOADDR);
464 					error = EADDRNOTAVAIL;
465 					goto bad;
466 				}
467 				ia4_acquire(xia, &_psref);
468 			}
469 			ip->ip_src = xia->ia_addr.sin_addr;
470 			ia4_release(xia, &_psref);
471 		}
472 
473 		inmgroup = in_multi_group(ip->ip_dst, ifp, flags);
474 		if (inmgroup && (imo == NULL || imo->imo_multicast_loop)) {
475 			/*
476 			 * If we belong to the destination multicast group
477 			 * on the outgoing interface, and the caller did not
478 			 * forbid loopback, loop back a copy.
479 			 */
480 			ip_mloopback(ifp, m, &udst.sin);
481 		}
482 #ifdef MROUTING
483 		else {
484 			/*
485 			 * If we are acting as a multicast router, perform
486 			 * multicast forwarding as if the packet had just
487 			 * arrived on the interface to which we are about
488 			 * to send.  The multicast forwarding function
489 			 * recursively calls this function, using the
490 			 * IP_FORWARDING flag to prevent infinite recursion.
491 			 *
492 			 * Multicasts that are looped back by ip_mloopback(),
493 			 * above, will be forwarded by the ip_input() routine,
494 			 * if necessary.
495 			 */
496 			extern struct socket *ip_mrouter;
497 
498 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
499 				if (ip_mforward(m, ifp) != 0) {
500 					m_freem(m);
501 					goto done;
502 				}
503 			}
504 		}
505 #endif
506 		/*
507 		 * Multicasts with a time-to-live of zero may be looped-
508 		 * back, above, but must not be transmitted on a network.
509 		 * Also, multicasts addressed to the loopback interface
510 		 * are not sent -- the above call to ip_mloopback() will
511 		 * loop back a copy if this host actually belongs to the
512 		 * destination group on the loopback interface.
513 		 */
514 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
515 			IP_STATINC(IP_STAT_ODROPPED);
516 			m_freem(m);
517 			goto done;
518 		}
519 		goto sendit;
520 	}
521 
522 	/*
523 	 * If source address not specified yet, use address
524 	 * of outgoing interface.
525 	 */
526 	if (in_nullhost(ip->ip_src)) {
527 		struct ifaddr *xifa;
528 
529 		xifa = &ia->ia_ifa;
530 		if (xifa->ifa_getifa != NULL) {
531 			ia4_release(ia, &psref_ia);
532 			/* FIXME ifa_getifa is NOMPSAFE */
533 			ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst));
534 			if (ia == NULL) {
535 				error = EADDRNOTAVAIL;
536 				goto bad;
537 			}
538 			ia4_acquire(ia, &psref_ia);
539 		}
540 		ip->ip_src = ia->ia_addr.sin_addr;
541 	}
542 
543 	/*
544 	 * Packets with Class-D address as source are not valid per
545 	 * RFC1112.
546 	 */
547 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
548 		IP_STATINC(IP_STAT_ODROPPED);
549 		error = EADDRNOTAVAIL;
550 		goto bad;
551 	}
552 
553 	/*
554 	 * Look for broadcast address and verify user is allowed to
555 	 * send such a packet.
556 	 */
557 	if (isbroadcast) {
558 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
559 			IP_STATINC(IP_STAT_BCASTDENIED);
560 			error = EADDRNOTAVAIL;
561 			goto bad;
562 		}
563 		if ((flags & IP_ALLOWBROADCAST) == 0) {
564 			IP_STATINC(IP_STAT_BCASTDENIED);
565 			error = EACCES;
566 			goto bad;
567 		}
568 		/* don't allow broadcast messages to be fragmented */
569 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
570 			IP_STATINC(IP_STAT_BCASTDENIED);
571 			error = EMSGSIZE;
572 			goto bad;
573 		}
574 		m->m_flags |= M_BCAST;
575 	} else
576 		m->m_flags &= ~M_BCAST;
577 
578 sendit:
579 	if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) {
580 		if (m->m_pkthdr.len < IP_MINFRAGSIZE) {
581 			ip->ip_id = 0;
582 		} else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
583 			ip->ip_id = ip_newid(ia);
584 		} else {
585 			/*
586 			 * TSO capable interfaces (typically?) increment
587 			 * ip_id for each segment.
588 			 * "allocate" enough ids here to increase the chance
589 			 * for them to be unique.
590 			 *
591 			 * note that the following calculation is not
592 			 * needed to be precise.  wasting some ip_id is fine.
593 			 */
594 
595 			unsigned int segsz = m->m_pkthdr.segsz;
596 			unsigned int datasz = ntohs(ip->ip_len) - hlen;
597 			unsigned int num = howmany(datasz, segsz);
598 
599 			ip->ip_id = ip_newid_range(ia, num);
600 		}
601 	}
602 	if (ia != NULL) {
603 		ia4_release(ia, &psref_ia);
604 		ia = NULL;
605 	}
606 
607 	/*
608 	 * If we're doing Path MTU Discovery, we need to set DF unless
609 	 * the route's MTU is locked.
610 	 */
611 	if ((flags & IP_MTUDISC) != 0 && rtmtu_nolock) {
612 		ip->ip_off |= htons(IP_DF);
613 	}
614 
615 #ifdef IPSEC
616 	if (ipsec_used) {
617 		bool ipsec_done = false;
618 		bool count_drop = false;
619 
620 		/* Perform IPsec processing, if any. */
621 		error = ipsec4_output(m, inp, flags, &mtu, &natt_frag,
622 		    &ipsec_done, &count_drop);
623 		if (count_drop)
624 			IP_STATINC(IP_STAT_IPSECDROP_OUT);
625 		if (error || ipsec_done)
626 			goto done;
627 	}
628 
629 	if (!ipsec_used || !natt_frag)
630 #endif
631 	{
632 		/*
633 		 * Run through list of hooks for output packets.
634 		 */
635 		error = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_OUT);
636 		if (error || m == NULL) {
637 			IP_STATINC(IP_STAT_PFILDROP_OUT);
638 			goto done;
639 		}
640 	}
641 
642 	ip = mtod(m, struct ip *);
643 	hlen = ip->ip_hl << 2;
644 
645 	m->m_pkthdr.csum_data |= hlen << 16;
646 
647 	/*
648 	 * search for the source address structure to
649 	 * maintain output statistics, and verify address
650 	 * validity
651 	 */
652 	KASSERT(ia == NULL);
653 	sockaddr_in_init(&usrc.sin, &ip->ip_src, 0);
654 	ifa = ifaof_ifpforaddr_psref(&usrc.sa, ifp, &psref_ia);
655 	if (ifa != NULL)
656 		ia = ifatoia(ifa);
657 
658 	/*
659 	 * Ensure we only send from a valid address.
660 	 * A NULL address is valid because the packet could be
661 	 * generated from a packet filter.
662 	 */
663 	if (ia != NULL && (flags & IP_FORWARDING) == 0 &&
664 	    (error = ip_ifaddrvalid(ia)) != 0)
665 	{
666 		ARPLOG(LOG_ERR,
667 		    "refusing to send from invalid address %s (pid %d)\n",
668 		    ARPLOGADDR(&ip->ip_src), curproc->p_pid);
669 		IP_STATINC(IP_STAT_ODROPPED);
670 		if (error == 1)
671 			/*
672 			 * Address exists, but is tentative or detached.
673 			 * We can't send from it because it's invalid,
674 			 * so we drop the packet.
675 			 */
676 			error = 0;
677 		else
678 			error = EADDRNOTAVAIL;
679 		goto bad;
680 	}
681 
682 	/* Maybe skip checksums on loopback interfaces. */
683 	if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) {
684 		m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
685 	}
686 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
687 
688 	/* Need to fragment the packet */
689 	if (ntohs(ip->ip_len) > mtu &&
690 	    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
691 		goto fragment;
692 	}
693 
694 #if IFA_STATS
695 	if (ia)
696 		ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
697 #endif
698 	/*
699 	 * Always initialize the sum to 0!  Some HW assisted
700 	 * checksumming requires this.
701 	 */
702 	ip->ip_sum = 0;
703 
704 	if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
705 		/*
706 		 * Perform any checksums that the hardware can't do
707 		 * for us.
708 		 *
709 		 * XXX Does any hardware require the {th,uh}_sum
710 		 * XXX fields to be 0?
711 		 */
712 		if (sw_csum & M_CSUM_IPv4) {
713 			KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4));
714 			ip->ip_sum = in_cksum(m, hlen);
715 			m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
716 		}
717 		if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
718 			if (IN_NEED_CHECKSUM(ifp,
719 			    sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
720 				in_undefer_cksum_tcpudp(m);
721 			}
722 			m->m_pkthdr.csum_flags &=
723 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
724 		}
725 	}
726 
727 	sa = (m->m_flags & M_MCAST) ? sintocsa(rdst) : sintocsa(dst);
728 
729 	/* Send it */
730 	if (__predict_false(sw_csum & M_CSUM_TSOv4)) {
731 		/*
732 		 * TSO4 is required by a packet, but disabled for
733 		 * the interface.
734 		 */
735 		error = ip_tso_output(ifp, m, sa, rt);
736 	} else
737 		error = ip_if_output(ifp, m, sa, rt);
738 	goto done;
739 
740 fragment:
741 	/*
742 	 * We can't use HW checksumming if we're about to fragment the packet.
743 	 *
744 	 * XXX Some hardware can do this.
745 	 */
746 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
747 		if (IN_NEED_CHECKSUM(ifp,
748 		    m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
749 			in_undefer_cksum_tcpudp(m);
750 		}
751 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
752 	}
753 
754 	/*
755 	 * Too large for interface; fragment if possible.
756 	 * Must be able to put at least 8 bytes per fragment.
757 	 */
758 	if (ntohs(ip->ip_off) & IP_DF) {
759 		if (flags & IP_RETURNMTU) {
760 			KASSERT(inp != NULL);
761 			in4p_errormtu(inp) = mtu;
762 		}
763 		error = EMSGSIZE;
764 		IP_STATINC(IP_STAT_CANTFRAG);
765 		goto bad;
766 	}
767 
768 	error = ip_fragment(m, ifp, mtu);
769 	if (error) {
770 		m = NULL;
771 		goto bad;
772 	}
773 
774 	for (; m; m = m0) {
775 		m0 = m->m_nextpkt;
776 		m->m_nextpkt = NULL;
777 		if (error) {
778 			m_freem(m);
779 			continue;
780 		}
781 #if IFA_STATS
782 		if (ia)
783 			ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
784 #endif
785 		/*
786 		 * If we get there, the packet has not been handled by
787 		 * IPsec whereas it should have. Now that it has been
788 		 * fragmented, re-inject it in ip_output so that IPsec
789 		 * processing can occur.
790 		 */
791 		if (natt_frag) {
792 			error = ip_output(m, opt, NULL,
793 			    flags | IP_RAWOUTPUT | IP_NOIPNEWID,
794 			    imo, inp);
795 		} else {
796 			KASSERT((m->m_pkthdr.csum_flags &
797 			    (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0);
798 			error = ip_if_output(ifp, m, (m->m_flags & M_MCAST) ?
799 			    sintocsa(rdst) : sintocsa(dst), rt);
800 		}
801 	}
802 	if (error == 0) {
803 		IP_STATINC(IP_STAT_FRAGMENTED);
804 	}
805 
806 done:
807 	ia4_release(ia, &psref_ia);
808 	rtcache_unref(rt, ro);
809 	if (ro == &iproute) {
810 		rtcache_free(&iproute);
811 	}
812 	if (mifp != NULL) {
813 		if_put(mifp, &psref);
814 	}
815 	if (bind_need_restore)
816 		curlwp_bindx(bound);
817 	return error;
818 
819 bad:
820 	m_freem(m);
821 	goto done;
822 }
823 
824 int
ip_fragment(struct mbuf * m,struct ifnet * ifp,u_long mtu)825 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
826 {
827 	struct ip *ip, *mhip;
828 	struct mbuf *m0;
829 	int len, hlen, off;
830 	int mhlen, firstlen;
831 	struct mbuf **mnext;
832 	int sw_csum = m->m_pkthdr.csum_flags;
833 	int fragments = 0;
834 	int error = 0;
835 	int ipoff, ipflg;
836 
837 	ip = mtod(m, struct ip *);
838 	hlen = ip->ip_hl << 2;
839 
840 	/* Preserve the offset and flags. */
841 	ipoff = ntohs(ip->ip_off) & IP_OFFMASK;
842 	ipflg = ntohs(ip->ip_off) & (IP_RF|IP_DF|IP_MF);
843 
844 	if (ifp != NULL)
845 		sw_csum &= ~ifp->if_csum_flags_tx;
846 
847 	len = (mtu - hlen) &~ 7;
848 	if (len < 8) {
849 		IP_STATINC(IP_STAT_CANTFRAG);
850 		m_freem(m);
851 		return EMSGSIZE;
852 	}
853 
854 	firstlen = len;
855 	mnext = &m->m_nextpkt;
856 
857 	/*
858 	 * Loop through length of segment after first fragment,
859 	 * make new header and copy data of each part and link onto chain.
860 	 */
861 	m0 = m;
862 	mhlen = sizeof(struct ip);
863 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
864 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
865 		if (m == NULL) {
866 			error = ENOBUFS;
867 			IP_STATINC(IP_STAT_ODROPPED);
868 			goto sendorfree;
869 		}
870 		MCLAIM(m, m0->m_owner);
871 
872 		*mnext = m;
873 		mnext = &m->m_nextpkt;
874 
875 		m->m_data += max_linkhdr;
876 		mhip = mtod(m, struct ip *);
877 		*mhip = *ip;
878 
879 		/* we must inherit the flags */
880 		m->m_flags |= m0->m_flags & M_COPYFLAGS;
881 
882 		if (hlen > sizeof(struct ip)) {
883 			mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip);
884 			mhip->ip_hl = mhlen >> 2;
885 		}
886 		m->m_len = mhlen;
887 
888 		mhip->ip_off = ((off - hlen) >> 3) + ipoff;
889 		mhip->ip_off |= ipflg;
890 		if (off + len >= ntohs(ip->ip_len))
891 			len = ntohs(ip->ip_len) - off;
892 		else
893 			mhip->ip_off |= IP_MF;
894 		HTONS(mhip->ip_off);
895 
896 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
897 		m->m_next = m_copym(m0, off, len, M_DONTWAIT);
898 		if (m->m_next == NULL) {
899 			error = ENOBUFS;
900 			IP_STATINC(IP_STAT_ODROPPED);
901 			goto sendorfree;
902 		}
903 
904 		m->m_pkthdr.len = mhlen + len;
905 		m_reset_rcvif(m);
906 
907 		mhip->ip_sum = 0;
908 		KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0);
909 		if (sw_csum & M_CSUM_IPv4) {
910 			mhip->ip_sum = in_cksum(m, mhlen);
911 		} else {
912 			/*
913 			 * checksum is hw-offloaded or not necessary.
914 			 */
915 			m->m_pkthdr.csum_flags |=
916 			    m0->m_pkthdr.csum_flags & M_CSUM_IPv4;
917 			m->m_pkthdr.csum_data |= mhlen << 16;
918 			KASSERT(!(ifp != NULL &&
919 			    IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) ||
920 			    (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0);
921 		}
922 		IP_STATINC(IP_STAT_OFRAGMENTS);
923 		fragments++;
924 	}
925 
926 	/*
927 	 * Update first fragment by trimming what's been copied out
928 	 * and updating header, then send each fragment (in order).
929 	 */
930 	m = m0;
931 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
932 	m->m_pkthdr.len = hlen + firstlen;
933 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
934 	ip->ip_off |= htons(IP_MF);
935 	ip->ip_sum = 0;
936 	if (sw_csum & M_CSUM_IPv4) {
937 		ip->ip_sum = in_cksum(m, hlen);
938 		m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
939 	} else {
940 		/*
941 		 * checksum is hw-offloaded or not necessary.
942 		 */
943 		KASSERT(!(ifp != NULL && IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) ||
944 		    (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0);
945 		KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >=
946 		    sizeof(struct ip));
947 	}
948 
949 sendorfree:
950 	/*
951 	 * If there is no room for all the fragments, don't queue
952 	 * any of them.
953 	 */
954 	if (ifp != NULL) {
955 		IFQ_LOCK(&ifp->if_snd);
956 		if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments &&
957 		    error == 0) {
958 			error = ENOBUFS;
959 			IP_STATINC(IP_STAT_ODROPPED);
960 			IFQ_INC_DROPS(&ifp->if_snd);
961 		}
962 		IFQ_UNLOCK(&ifp->if_snd);
963 	}
964 	if (error) {
965 		for (m = m0; m; m = m0) {
966 			m0 = m->m_nextpkt;
967 			m->m_nextpkt = NULL;
968 			m_freem(m);
969 		}
970 	}
971 
972 	return error;
973 }
974 
975 /*
976  * Determine the maximum length of the options to be inserted;
977  * we would far rather allocate too much space rather than too little.
978  */
979 u_int
ip_optlen(struct inpcb * inp)980 ip_optlen(struct inpcb *inp)
981 {
982 	struct mbuf *m = inp->inp_options;
983 
984 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) {
985 		return (m->m_len - offsetof(struct ipoption, ipopt_dst));
986 	}
987 	return 0;
988 }
989 
990 /*
991  * Insert IP options into preformed packet.
992  * Adjust IP destination as required for IP source routing,
993  * as indicated by a non-zero in_addr at the start of the options.
994  */
995 static struct mbuf *
ip_insertoptions(struct mbuf * m,struct mbuf * opt,int * phlen)996 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
997 {
998 	struct ipoption *p = mtod(opt, struct ipoption *);
999 	struct mbuf *n;
1000 	struct ip *ip = mtod(m, struct ip *);
1001 	unsigned optlen;
1002 
1003 	optlen = opt->m_len - sizeof(p->ipopt_dst);
1004 	KASSERT(optlen % 4 == 0);
1005 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
1006 		return m;		/* XXX should fail */
1007 	if (!in_nullhost(p->ipopt_dst))
1008 		ip->ip_dst = p->ipopt_dst;
1009 	if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) {
1010 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
1011 		if (n == NULL)
1012 			return m;
1013 		MCLAIM(n, m->m_owner);
1014 		m_move_pkthdr(n, m);
1015 		m->m_len -= sizeof(struct ip);
1016 		m->m_data += sizeof(struct ip);
1017 		n->m_next = m;
1018 		n->m_len = optlen + sizeof(struct ip);
1019 		n->m_data += max_linkhdr;
1020 		memcpy(mtod(n, void *), ip, sizeof(struct ip));
1021 		m = n;
1022 	} else {
1023 		m->m_data -= optlen;
1024 		m->m_len += optlen;
1025 		memmove(mtod(m, void *), ip, sizeof(struct ip));
1026 	}
1027 	m->m_pkthdr.len += optlen;
1028 	ip = mtod(m, struct ip *);
1029 	memcpy(ip + 1, p->ipopt_list, optlen);
1030 	*phlen = sizeof(struct ip) + optlen;
1031 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
1032 	return m;
1033 }
1034 
1035 /*
1036  * Copy options from ipsrc to ipdst, omitting those not copied during
1037  * fragmentation.
1038  */
1039 int
ip_optcopy(struct ip * ipsrc,struct ip * ipdst)1040 ip_optcopy(struct ip *ipsrc, struct ip *ipdst)
1041 {
1042 	u_char *cp, *dp;
1043 	int opt, optlen, cnt;
1044 
1045 	cp = (u_char *)(ipsrc + 1);
1046 	dp = (u_char *)(ipdst + 1);
1047 	cnt = (ipsrc->ip_hl << 2) - sizeof(struct ip);
1048 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1049 		opt = cp[0];
1050 		if (opt == IPOPT_EOL)
1051 			break;
1052 		if (opt == IPOPT_NOP) {
1053 			/* Preserve for IP mcast tunnel's LSRR alignment. */
1054 			*dp++ = IPOPT_NOP;
1055 			optlen = 1;
1056 			continue;
1057 		}
1058 
1059 		KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp));
1060 		optlen = cp[IPOPT_OLEN];
1061 		KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen < cnt);
1062 
1063 		/* Invalid lengths should have been caught by ip_dooptions. */
1064 		if (optlen > cnt)
1065 			optlen = cnt;
1066 		if (IPOPT_COPIED(opt)) {
1067 			bcopy((void *)cp, (void *)dp, (unsigned)optlen);
1068 			dp += optlen;
1069 		}
1070 	}
1071 
1072 	for (optlen = dp - (u_char *)(ipdst+1); optlen & 0x3; optlen++) {
1073 		*dp++ = IPOPT_EOL;
1074 	}
1075 
1076 	return optlen;
1077 }
1078 
1079 /*
1080  * IP socket option processing.
1081  */
1082 int
ip_ctloutput(int op,struct socket * so,struct sockopt * sopt)1083 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1084 {
1085 	struct inpcb *inp = sotoinpcb(so);
1086 	struct ip *ip = &in4p_ip(inp);
1087 	int inpflags = inp->inp_flags;
1088 	int optval = 0, error = 0;
1089 	struct in_pktinfo pktinfo;
1090 
1091 	KASSERT(solocked(so));
1092 
1093 	if (sopt->sopt_level != IPPROTO_IP) {
1094 		if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER)
1095 			return 0;
1096 		return ENOPROTOOPT;
1097 	}
1098 
1099 	switch (op) {
1100 	case PRCO_SETOPT:
1101 		switch (sopt->sopt_name) {
1102 		case IP_OPTIONS:
1103 #ifdef notyet
1104 		case IP_RETOPTS:
1105 #endif
1106 			error = ip_pcbopts(inp, sopt);
1107 			break;
1108 
1109 		case IP_TOS:
1110 		case IP_TTL:
1111 		case IP_MINTTL:
1112 		case IP_RECVOPTS:
1113 		case IP_RECVRETOPTS:
1114 		case IP_RECVDSTADDR:
1115 		case IP_RECVIF:
1116 		case IP_RECVPKTINFO:
1117 		case IP_RECVTTL:
1118 		case IP_BINDANY:
1119 			error = sockopt_getint(sopt, &optval);
1120 			if (error)
1121 				break;
1122 
1123 			switch (sopt->sopt_name) {
1124 			case IP_TOS:
1125 				ip->ip_tos = optval;
1126 				break;
1127 
1128 			case IP_TTL:
1129 				ip->ip_ttl = optval;
1130 				break;
1131 
1132 			case IP_MINTTL:
1133 				if (optval > 0 && optval <= MAXTTL)
1134 					in4p_ip_minttl(inp) = optval;
1135 				else
1136 					error = EINVAL;
1137 				break;
1138 #define	OPTSET(bit) \
1139 	if (optval) \
1140 		inpflags |= bit; \
1141 	else \
1142 		inpflags &= ~bit;
1143 
1144 			case IP_RECVOPTS:
1145 				OPTSET(INP_RECVOPTS);
1146 				break;
1147 
1148 			case IP_RECVPKTINFO:
1149 				OPTSET(INP_RECVPKTINFO);
1150 				break;
1151 
1152 			case IP_RECVRETOPTS:
1153 				OPTSET(INP_RECVRETOPTS);
1154 				break;
1155 
1156 			case IP_RECVDSTADDR:
1157 				OPTSET(INP_RECVDSTADDR);
1158 				break;
1159 
1160 			case IP_RECVIF:
1161 				OPTSET(INP_RECVIF);
1162 				break;
1163 
1164 			case IP_RECVTTL:
1165 				OPTSET(INP_RECVTTL);
1166 				break;
1167 
1168 			case IP_BINDANY:
1169 				error = kauth_authorize_network(
1170 				    kauth_cred_get(), KAUTH_NETWORK_BIND,
1171 				    KAUTH_REQ_NETWORK_BIND_ANYADDR, so,
1172 				    NULL, NULL);
1173 				if (error == 0) {
1174 					OPTSET(INP_BINDANY);
1175 				}
1176 				break;
1177 			}
1178 			break;
1179 		case IP_PKTINFO:
1180 			error = sockopt_getint(sopt, &optval);
1181 			if (!error) {
1182 				/* Linux compatibility */
1183 				OPTSET(INP_RECVPKTINFO);
1184 				break;
1185 			}
1186 			error = sockopt_get(sopt, &pktinfo, sizeof(pktinfo));
1187 			if (error)
1188 				break;
1189 
1190 			if (pktinfo.ipi_ifindex == 0) {
1191 				in4p_prefsrcip(inp) = pktinfo.ipi_addr;
1192 				break;
1193 			}
1194 
1195 			/* Solaris compatibility */
1196 			struct ifnet *ifp;
1197 			struct in_ifaddr *ia;
1198 			int s;
1199 
1200 			/* pick up primary address */
1201 			s = pserialize_read_enter();
1202 			ifp = if_byindex(pktinfo.ipi_ifindex);
1203 			if (ifp == NULL) {
1204 				pserialize_read_exit(s);
1205 				error = EADDRNOTAVAIL;
1206 				break;
1207 			}
1208 			ia = in_get_ia_from_ifp(ifp);
1209 			if (ia == NULL) {
1210 				pserialize_read_exit(s);
1211 				error = EADDRNOTAVAIL;
1212 				break;
1213 			}
1214 			in4p_prefsrcip(inp) = IA_SIN(ia)->sin_addr;
1215 			pserialize_read_exit(s);
1216 			break;
1217 		break;
1218 #undef OPTSET
1219 
1220 		case IP_MULTICAST_IF:
1221 		case IP_MULTICAST_TTL:
1222 		case IP_MULTICAST_LOOP:
1223 		case IP_ADD_MEMBERSHIP:
1224 		case IP_DROP_MEMBERSHIP:
1225 			error = ip_setmoptions(&inp->inp_moptions, sopt);
1226 			break;
1227 
1228 		case IP_PORTRANGE:
1229 			error = sockopt_getint(sopt, &optval);
1230 			if (error)
1231 				break;
1232 
1233 			switch (optval) {
1234 			case IP_PORTRANGE_DEFAULT:
1235 			case IP_PORTRANGE_HIGH:
1236 				inpflags &= ~(INP_LOWPORT);
1237 				break;
1238 
1239 			case IP_PORTRANGE_LOW:
1240 				inpflags |= INP_LOWPORT;
1241 				break;
1242 
1243 			default:
1244 				error = EINVAL;
1245 				break;
1246 			}
1247 			break;
1248 
1249 		case IP_PORTALGO:
1250 			error = sockopt_getint(sopt, &optval);
1251 			if (error)
1252 				break;
1253 
1254 			error = portalgo_algo_index_select(inp, optval);
1255 			break;
1256 
1257 #if defined(IPSEC)
1258 		case IP_IPSEC_POLICY:
1259 			if (ipsec_enabled) {
1260 				error = ipsec_set_policy(inp,
1261 				    sopt->sopt_data, sopt->sopt_size,
1262 				    curlwp->l_cred);
1263 			} else
1264 				error = ENOPROTOOPT;
1265 			break;
1266 #endif /* IPSEC */
1267 
1268 		default:
1269 			error = ENOPROTOOPT;
1270 			break;
1271 		}
1272 		break;
1273 
1274 	case PRCO_GETOPT:
1275 		switch (sopt->sopt_name) {
1276 		case IP_OPTIONS:
1277 		case IP_RETOPTS: {
1278 			struct mbuf *mopts = inp->inp_options;
1279 
1280 			if (mopts) {
1281 				struct mbuf *m;
1282 
1283 				m = m_copym(mopts, 0, M_COPYALL, M_DONTWAIT);
1284 				if (m == NULL) {
1285 					error = ENOBUFS;
1286 					break;
1287 				}
1288 				error = sockopt_setmbuf(sopt, m);
1289 			}
1290 			break;
1291 		}
1292 		case IP_TOS:
1293 		case IP_TTL:
1294 		case IP_MINTTL:
1295 		case IP_RECVOPTS:
1296 		case IP_RECVRETOPTS:
1297 		case IP_RECVDSTADDR:
1298 		case IP_RECVIF:
1299 		case IP_RECVPKTINFO:
1300 		case IP_RECVTTL:
1301 		case IP_ERRORMTU:
1302 		case IP_BINDANY:
1303 			switch (sopt->sopt_name) {
1304 			case IP_TOS:
1305 				optval = ip->ip_tos;
1306 				break;
1307 
1308 			case IP_TTL:
1309 				optval = ip->ip_ttl;
1310 				break;
1311 
1312 			case IP_MINTTL:
1313 				optval = in4p_ip_minttl(inp);
1314 				break;
1315 
1316 			case IP_ERRORMTU:
1317 				optval = in4p_errormtu(inp);
1318 				break;
1319 
1320 #define	OPTBIT(bit)	(inpflags & bit ? 1 : 0)
1321 
1322 			case IP_RECVOPTS:
1323 				optval = OPTBIT(INP_RECVOPTS);
1324 				break;
1325 
1326 			case IP_RECVPKTINFO:
1327 				optval = OPTBIT(INP_RECVPKTINFO);
1328 				break;
1329 
1330 			case IP_RECVRETOPTS:
1331 				optval = OPTBIT(INP_RECVRETOPTS);
1332 				break;
1333 
1334 			case IP_RECVDSTADDR:
1335 				optval = OPTBIT(INP_RECVDSTADDR);
1336 				break;
1337 
1338 			case IP_RECVIF:
1339 				optval = OPTBIT(INP_RECVIF);
1340 				break;
1341 
1342 			case IP_RECVTTL:
1343 				optval = OPTBIT(INP_RECVTTL);
1344 				break;
1345 
1346 			case IP_BINDANY:
1347 				optval = OPTBIT(INP_BINDANY);
1348 				break;
1349 			}
1350 			error = sockopt_setint(sopt, optval);
1351 			break;
1352 
1353 		case IP_PKTINFO:
1354 			switch (sopt->sopt_size) {
1355 			case sizeof(int):
1356 				/* Linux compatibility */
1357 				optval = OPTBIT(INP_RECVPKTINFO);
1358 				error = sockopt_setint(sopt, optval);
1359 				break;
1360 			case sizeof(struct in_pktinfo):
1361 				/* Solaris compatibility */
1362 				pktinfo.ipi_ifindex = 0;
1363 				pktinfo.ipi_addr = in4p_prefsrcip(inp);
1364 				error = sockopt_set(sopt, &pktinfo,
1365 				    sizeof(pktinfo));
1366 				break;
1367 			default:
1368 				/*
1369 				 * While size is stuck at 0, and, later, if
1370 				 * the caller doesn't use an exactly sized
1371 				 * recipient for the data, default to Linux
1372 				 * compatibility
1373 				 */
1374 				optval = OPTBIT(INP_RECVPKTINFO);
1375 				error = sockopt_setint(sopt, optval);
1376 				break;
1377 			}
1378 			break;
1379 
1380 #if 0	/* defined(IPSEC) */
1381 		case IP_IPSEC_POLICY:
1382 		{
1383 			struct mbuf *m = NULL;
1384 
1385 			/* XXX this will return EINVAL as sopt is empty */
1386 			error = ipsec_get_policy(inp, sopt->sopt_data,
1387 			    sopt->sopt_size, &m);
1388 			if (error == 0)
1389 				error = sockopt_setmbuf(sopt, m);
1390 			break;
1391 		}
1392 #endif /*IPSEC*/
1393 
1394 		case IP_MULTICAST_IF:
1395 		case IP_MULTICAST_TTL:
1396 		case IP_MULTICAST_LOOP:
1397 		case IP_ADD_MEMBERSHIP:
1398 		case IP_DROP_MEMBERSHIP:
1399 			error = ip_getmoptions(inp->inp_moptions, sopt);
1400 			break;
1401 
1402 		case IP_PORTRANGE:
1403 			if (inpflags & INP_LOWPORT)
1404 				optval = IP_PORTRANGE_LOW;
1405 			else
1406 				optval = IP_PORTRANGE_DEFAULT;
1407 			error = sockopt_setint(sopt, optval);
1408 			break;
1409 
1410 		case IP_PORTALGO:
1411 			optval = inp->inp_portalgo;
1412 			error = sockopt_setint(sopt, optval);
1413 			break;
1414 
1415 		default:
1416 			error = ENOPROTOOPT;
1417 			break;
1418 		}
1419 		break;
1420 	}
1421 
1422 	if (!error) {
1423 		inp->inp_flags = inpflags;
1424 	}
1425 	return error;
1426 }
1427 
1428 static int
ip_pktinfo_prepare(const struct inpcb * inp,const struct in_pktinfo * pktinfo,struct ip_pktopts * pktopts,int * flags,kauth_cred_t cred)1429 ip_pktinfo_prepare(const struct inpcb *inp, const struct in_pktinfo *pktinfo,
1430     struct ip_pktopts *pktopts, int *flags, kauth_cred_t cred)
1431 {
1432 	struct ip_moptions *imo;
1433 	int error = 0;
1434 	bool addrset = false;
1435 
1436 	if (!in_nullhost(pktinfo->ipi_addr)) {
1437 		pktopts->ippo_laddr.sin_addr = pktinfo->ipi_addr;
1438 		/* EADDRNOTAVAIL? */
1439 		error = inpcb_bindableaddr(inp, &pktopts->ippo_laddr, cred);
1440 		if (error != 0)
1441 			return error;
1442 		addrset = true;
1443 	}
1444 
1445 	if (pktinfo->ipi_ifindex != 0) {
1446 		if (!addrset) {
1447 			struct ifnet *ifp;
1448 			struct in_ifaddr *ia;
1449 			int s;
1450 
1451 			/* pick up primary address */
1452 			s = pserialize_read_enter();
1453 			ifp = if_byindex(pktinfo->ipi_ifindex);
1454 			if (ifp == NULL) {
1455 				pserialize_read_exit(s);
1456 				return EADDRNOTAVAIL;
1457 			}
1458 			ia = in_get_ia_from_ifp(ifp);
1459 			if (ia == NULL) {
1460 				pserialize_read_exit(s);
1461 				return EADDRNOTAVAIL;
1462 			}
1463 			pktopts->ippo_laddr.sin_addr = IA_SIN(ia)->sin_addr;
1464 			pserialize_read_exit(s);
1465 		}
1466 
1467 		/*
1468 		 * If specified ipi_ifindex,
1469 		 * use copied or locally initialized ip_moptions.
1470 		 * Original ip_moptions must not be modified.
1471 		 */
1472 		imo = &pktopts->ippo_imobuf;	/* local buf in pktopts */
1473 		if (pktopts->ippo_imo != NULL) {
1474 			memcpy(imo, pktopts->ippo_imo, sizeof(*imo));
1475 		} else {
1476 			memset(imo, 0, sizeof(*imo));
1477 			imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1478 			imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1479 		}
1480 		imo->imo_multicast_if_index = pktinfo->ipi_ifindex;
1481 		pktopts->ippo_imo = imo;
1482 		*flags |= IP_ROUTETOIFINDEX;
1483 	}
1484 	return error;
1485 }
1486 
1487 /*
1488  * Set up IP outgoing packet options. Even if control is NULL,
1489  * pktopts->ippo_laddr and pktopts->ippo_imo are set and used.
1490  */
1491 int
ip_setpktopts(struct mbuf * control,struct ip_pktopts * pktopts,int * flags,struct inpcb * inp,kauth_cred_t cred)1492 ip_setpktopts(struct mbuf *control, struct ip_pktopts *pktopts, int *flags,
1493     struct inpcb *inp, kauth_cred_t cred)
1494 {
1495 	struct cmsghdr *cm;
1496 	struct in_pktinfo pktinfo;
1497 	int error;
1498 
1499 	pktopts->ippo_imo = inp->inp_moptions;
1500 
1501 	struct in_addr *ia = in_nullhost(in4p_prefsrcip(inp)) ? &in4p_laddr(inp) :
1502 	    &in4p_prefsrcip(inp);
1503 	sockaddr_in_init(&pktopts->ippo_laddr, ia, 0);
1504 
1505 	if (control == NULL)
1506 		return 0;
1507 
1508 	/*
1509 	 * XXX: Currently, we assume all the optional information is
1510 	 * stored in a single mbuf.
1511 	 */
1512 	if (control->m_next)
1513 		return EINVAL;
1514 
1515 	for (; control->m_len > 0;
1516 	    control->m_data += CMSG_ALIGN(cm->cmsg_len),
1517 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
1518 		cm = mtod(control, struct cmsghdr *);
1519 		if ((control->m_len < sizeof(*cm)) ||
1520 		    (cm->cmsg_len == 0) ||
1521 		    (cm->cmsg_len > control->m_len)) {
1522 			return EINVAL;
1523 		}
1524 		if (cm->cmsg_level != IPPROTO_IP)
1525 			continue;
1526 
1527 		switch (cm->cmsg_type) {
1528 		case IP_PKTINFO:
1529 			if (cm->cmsg_len != CMSG_LEN(sizeof(pktinfo)))
1530 				return EINVAL;
1531 			memcpy(&pktinfo, CMSG_DATA(cm), sizeof(pktinfo));
1532 			error = ip_pktinfo_prepare(inp, &pktinfo, pktopts,
1533 			    flags, cred);
1534 			if (error)
1535 				return error;
1536 			break;
1537 		case IP_SENDSRCADDR: /* FreeBSD compatibility */
1538 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_addr)))
1539 				return EINVAL;
1540 			pktinfo.ipi_ifindex = 0;
1541 			pktinfo.ipi_addr =
1542 			    ((struct in_pktinfo *)CMSG_DATA(cm))->ipi_addr;
1543 			error = ip_pktinfo_prepare(inp, &pktinfo, pktopts,
1544 			    flags, cred);
1545 			if (error)
1546 				return error;
1547 			break;
1548 		default:
1549 			return ENOPROTOOPT;
1550 		}
1551 	}
1552 	return 0;
1553 }
1554 
1555 /*
1556  * Set up IP options in pcb for insertion in output packets.
1557  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1558  * with destination address if source routed.
1559  */
1560 static int
ip_pcbopts(struct inpcb * inp,const struct sockopt * sopt)1561 ip_pcbopts(struct inpcb *inp, const struct sockopt *sopt)
1562 {
1563 	struct mbuf *m;
1564 	const u_char *cp;
1565 	u_char *dp;
1566 	int cnt;
1567 
1568 	KASSERT(inp_locked(inp));
1569 
1570 	/* Turn off any old options. */
1571 	if (inp->inp_options) {
1572 		m_free(inp->inp_options);
1573 	}
1574 	inp->inp_options = NULL;
1575 	if ((cnt = sopt->sopt_size) == 0) {
1576 		/* Only turning off any previous options. */
1577 		return 0;
1578 	}
1579 	cp = sopt->sopt_data;
1580 
1581 	if (cnt % 4) {
1582 		/* Must be 4-byte aligned, because there's no padding. */
1583 		return EINVAL;
1584 	}
1585 
1586 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1587 	if (m == NULL)
1588 		return ENOBUFS;
1589 
1590 	dp = mtod(m, u_char *);
1591 	memset(dp, 0, sizeof(struct in_addr));
1592 	dp += sizeof(struct in_addr);
1593 	m->m_len = sizeof(struct in_addr);
1594 
1595 	/*
1596 	 * IP option list according to RFC791. Each option is of the form
1597 	 *
1598 	 *	[optval] [olen] [(olen - 2) data bytes]
1599 	 *
1600 	 * We validate the list and copy options to an mbuf for prepending
1601 	 * to data packets. The IP first-hop destination address will be
1602 	 * stored before actual options and is zero if unset.
1603 	 */
1604 	while (cnt > 0) {
1605 		uint8_t optval, olen, offset;
1606 
1607 		optval = cp[IPOPT_OPTVAL];
1608 
1609 		if (optval == IPOPT_EOL || optval == IPOPT_NOP) {
1610 			olen = 1;
1611 		} else {
1612 			if (cnt < IPOPT_OLEN + 1)
1613 				goto bad;
1614 
1615 			olen = cp[IPOPT_OLEN];
1616 			if (olen < IPOPT_OLEN + 1 || olen > cnt)
1617 				goto bad;
1618 		}
1619 
1620 		if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) {
1621 			/*
1622 			 * user process specifies route as:
1623 			 *	->A->B->C->D
1624 			 * D must be our final destination (but we can't
1625 			 * check that since we may not have connected yet).
1626 			 * A is first hop destination, which doesn't appear in
1627 			 * actual IP option, but is stored before the options.
1628 			 */
1629 			if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr))
1630 				goto bad;
1631 
1632 			offset = cp[IPOPT_OFFSET];
1633 			memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1,
1634 			    sizeof(struct in_addr));
1635 
1636 			cp += sizeof(struct in_addr);
1637 			cnt -= sizeof(struct in_addr);
1638 			olen -= sizeof(struct in_addr);
1639 
1640 			if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr))
1641 				goto bad;
1642 
1643 			memcpy(dp, cp, olen);
1644 			dp[IPOPT_OPTVAL] = optval;
1645 			dp[IPOPT_OLEN] = olen;
1646 			dp[IPOPT_OFFSET] = offset;
1647 			break;
1648 		} else {
1649 			if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr))
1650 				goto bad;
1651 
1652 			memcpy(dp, cp, olen);
1653 			break;
1654 		}
1655 
1656 		dp += olen;
1657 		m->m_len += olen;
1658 
1659 		if (optval == IPOPT_EOL)
1660 			break;
1661 
1662 		cp += olen;
1663 		cnt -= olen;
1664 	}
1665 
1666 	inp->inp_options = m;
1667 	return 0;
1668 
1669 bad:
1670 	(void)m_free(m);
1671 	return EINVAL;
1672 }
1673 
1674 /*
1675  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1676  * Must be called in a pserialize critical section.
1677  */
1678 static struct ifnet *
ip_multicast_if(struct in_addr * a,int * ifindexp)1679 ip_multicast_if(struct in_addr *a, int *ifindexp)
1680 {
1681 	int ifindex;
1682 	struct ifnet *ifp = NULL;
1683 	struct in_ifaddr *ia;
1684 
1685 	if (ifindexp)
1686 		*ifindexp = 0;
1687 	if (ntohl(a->s_addr) >> 24 == 0) {
1688 		ifindex = ntohl(a->s_addr) & 0xffffff;
1689 		ifp = if_byindex(ifindex);
1690 		if (!ifp)
1691 			return NULL;
1692 		if (ifindexp)
1693 			*ifindexp = ifindex;
1694 	} else {
1695 		IN_ADDRHASH_READER_FOREACH(ia, a->s_addr) {
1696 			if (in_hosteq(ia->ia_addr.sin_addr, *a) &&
1697 			    (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) {
1698 				ifp = ia->ia_ifp;
1699 				if (if_is_deactivated(ifp))
1700 					ifp = NULL;
1701 				break;
1702 			}
1703 		}
1704 	}
1705 	return ifp;
1706 }
1707 
1708 static int
ip_getoptval(const struct sockopt * sopt,u_int8_t * val,u_int maxval)1709 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval)
1710 {
1711 	u_int tval;
1712 	u_char cval;
1713 	int error;
1714 
1715 	if (sopt == NULL)
1716 		return EINVAL;
1717 
1718 	switch (sopt->sopt_size) {
1719 	case sizeof(u_char):
1720 		error = sockopt_get(sopt, &cval, sizeof(u_char));
1721 		tval = cval;
1722 		break;
1723 
1724 	case sizeof(u_int):
1725 		error = sockopt_get(sopt, &tval, sizeof(u_int));
1726 		break;
1727 
1728 	default:
1729 		error = EINVAL;
1730 	}
1731 
1732 	if (error)
1733 		return error;
1734 
1735 	if (tval > maxval)
1736 		return EINVAL;
1737 
1738 	*val = tval;
1739 	return 0;
1740 }
1741 
1742 static int
ip_get_membership(const struct sockopt * sopt,struct ifnet ** ifp,struct psref * psref,struct in_addr * ia,bool add)1743 ip_get_membership(const struct sockopt *sopt, struct ifnet **ifp,
1744     struct psref *psref, struct in_addr *ia, bool add)
1745 {
1746 	int error;
1747 	struct ip_mreq mreq;
1748 
1749 	error = sockopt_get(sopt, &mreq, sizeof(mreq));
1750 	if (error)
1751 		return error;
1752 
1753 	if (!IN_MULTICAST(mreq.imr_multiaddr.s_addr))
1754 		return EINVAL;
1755 
1756 	memcpy(ia, &mreq.imr_multiaddr, sizeof(*ia));
1757 
1758 	if (in_nullhost(mreq.imr_interface)) {
1759 		union {
1760 			struct sockaddr		dst;
1761 			struct sockaddr_in	dst4;
1762 		} u;
1763 		struct route ro;
1764 
1765 		if (!add) {
1766 			*ifp = NULL;
1767 			return 0;
1768 		}
1769 		/*
1770 		 * If no interface address was provided, use the interface of
1771 		 * the route to the given multicast address.
1772 		 */
1773 		struct rtentry *rt;
1774 		memset(&ro, 0, sizeof(ro));
1775 
1776 		sockaddr_in_init(&u.dst4, ia, 0);
1777 		error = rtcache_setdst(&ro, &u.dst);
1778 		if (error != 0)
1779 			return error;
1780 		*ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL;
1781 		if (*ifp != NULL) {
1782 			if (if_is_deactivated(*ifp))
1783 				*ifp = NULL;
1784 			else
1785 				if_acquire(*ifp, psref);
1786 		}
1787 		rtcache_unref(rt, &ro);
1788 		rtcache_free(&ro);
1789 	} else {
1790 		int s = pserialize_read_enter();
1791 		*ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1792 		if (!add && *ifp == NULL) {
1793 			pserialize_read_exit(s);
1794 			return EADDRNOTAVAIL;
1795 		}
1796 		if (*ifp != NULL) {
1797 			if (if_is_deactivated(*ifp))
1798 				*ifp = NULL;
1799 			else
1800 				if_acquire(*ifp, psref);
1801 		}
1802 		pserialize_read_exit(s);
1803 	}
1804 	return 0;
1805 }
1806 
1807 /*
1808  * Add a multicast group membership.
1809  * Group must be a valid IP multicast address.
1810  */
1811 static int
ip_add_membership(struct ip_moptions * imo,const struct sockopt * sopt)1812 ip_add_membership(struct ip_moptions *imo, const struct sockopt *sopt)
1813 {
1814 	struct ifnet *ifp = NULL;	// XXX: gcc [ppc]
1815 	struct in_addr ia;
1816 	int i, error, bound;
1817 	struct psref psref;
1818 
1819 	/* imo is protected by solock or referenced only by the caller */
1820 
1821 	bound = curlwp_bind();
1822 	if (sopt->sopt_size == sizeof(struct ip_mreq))
1823 		error = ip_get_membership(sopt, &ifp, &psref, &ia, true);
1824 	else {
1825 #ifdef INET6
1826 		error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia));
1827 #else
1828 		error = EINVAL;
1829 #endif
1830 	}
1831 
1832 	if (error)
1833 		goto out;
1834 
1835 	/*
1836 	 * See if we found an interface, and confirm that it
1837 	 * supports multicast.
1838 	 */
1839 	if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1840 		error = EADDRNOTAVAIL;
1841 		goto out;
1842 	}
1843 
1844 	/*
1845 	 * See if the membership already exists or if all the
1846 	 * membership slots are full.
1847 	 */
1848 	for (i = 0; i < imo->imo_num_memberships; ++i) {
1849 		if (imo->imo_membership[i]->inm_ifp == ifp &&
1850 		    in_hosteq(imo->imo_membership[i]->inm_addr, ia))
1851 			break;
1852 	}
1853 	if (i < imo->imo_num_memberships) {
1854 		error = EADDRINUSE;
1855 		goto out;
1856 	}
1857 
1858 	if (i == IP_MAX_MEMBERSHIPS) {
1859 		error = ETOOMANYREFS;
1860 		goto out;
1861 	}
1862 
1863 	/*
1864 	 * Everything looks good; add a new record to the multicast
1865 	 * address list for the given interface.
1866 	 */
1867 	imo->imo_membership[i] = in_addmulti(&ia, ifp);
1868 	if (imo->imo_membership[i] == NULL) {
1869 		error = ENOBUFS;
1870 		goto out;
1871 	}
1872 
1873 	++imo->imo_num_memberships;
1874 	error = 0;
1875 out:
1876 	if_put(ifp, &psref);
1877 	curlwp_bindx(bound);
1878 	return error;
1879 }
1880 
1881 /*
1882  * Drop a multicast group membership.
1883  * Group must be a valid IP multicast address.
1884  */
1885 static int
ip_drop_membership(struct ip_moptions * imo,const struct sockopt * sopt)1886 ip_drop_membership(struct ip_moptions *imo, const struct sockopt *sopt)
1887 {
1888 	struct in_addr ia = { .s_addr = 0 };	// XXX: gcc [ppc]
1889 	struct ifnet *ifp = NULL;		// XXX: gcc [ppc]
1890 	int i, error, bound;
1891 	struct psref psref;
1892 
1893 	/* imo is protected by solock or referenced only by the caller */
1894 
1895 	bound = curlwp_bind();
1896 	if (sopt->sopt_size == sizeof(struct ip_mreq))
1897 		error = ip_get_membership(sopt, &ifp, &psref, &ia, false);
1898 	else {
1899 #ifdef INET6
1900 		error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia));
1901 #else
1902 		error = EINVAL;
1903 #endif
1904 	}
1905 
1906 	if (error)
1907 		goto out;
1908 
1909 	/*
1910 	 * Find the membership in the membership array.
1911 	 */
1912 	for (i = 0; i < imo->imo_num_memberships; ++i) {
1913 		if ((ifp == NULL ||
1914 		     imo->imo_membership[i]->inm_ifp == ifp) &&
1915 		    in_hosteq(imo->imo_membership[i]->inm_addr, ia))
1916 			break;
1917 	}
1918 	if (i == imo->imo_num_memberships) {
1919 		error = EADDRNOTAVAIL;
1920 		goto out;
1921 	}
1922 
1923 	/*
1924 	 * Give up the multicast address record to which the
1925 	 * membership points.
1926 	 */
1927 	in_delmulti(imo->imo_membership[i]);
1928 
1929 	/*
1930 	 * Remove the gap in the membership array.
1931 	 */
1932 	for (++i; i < imo->imo_num_memberships; ++i)
1933 		imo->imo_membership[i-1] = imo->imo_membership[i];
1934 	--imo->imo_num_memberships;
1935 	error = 0;
1936 out:
1937 	if_put(ifp, &psref);
1938 	curlwp_bindx(bound);
1939 	return error;
1940 }
1941 
1942 /*
1943  * Set the IP multicast options in response to user setsockopt().
1944  */
1945 int
ip_setmoptions(struct ip_moptions ** pimo,const struct sockopt * sopt)1946 ip_setmoptions(struct ip_moptions **pimo, const struct sockopt *sopt)
1947 {
1948 	struct ip_moptions *imo = *pimo;
1949 	struct in_addr addr;
1950 	struct ifnet *ifp;
1951 	int ifindex, error = 0;
1952 
1953 	/* The passed imo isn't NULL, it should be protected by solock */
1954 
1955 	if (!imo) {
1956 		/*
1957 		 * No multicast option buffer attached to the pcb;
1958 		 * allocate one and initialize to default values.
1959 		 */
1960 		imo = kmem_intr_alloc(sizeof(*imo), KM_NOSLEEP);
1961 		if (imo == NULL)
1962 			return ENOBUFS;
1963 
1964 		imo->imo_multicast_if_index = 0;
1965 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1966 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1967 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1968 		imo->imo_num_memberships = 0;
1969 		*pimo = imo;
1970 	}
1971 
1972 	switch (sopt->sopt_name) {
1973 	case IP_MULTICAST_IF: {
1974 		int s;
1975 		/*
1976 		 * Select the interface for outgoing multicast packets.
1977 		 */
1978 		error = sockopt_get(sopt, &addr, sizeof(addr));
1979 		if (error)
1980 			break;
1981 
1982 		/*
1983 		 * INADDR_ANY is used to remove a previous selection.
1984 		 * When no interface is selected, a default one is
1985 		 * chosen every time a multicast packet is sent.
1986 		 */
1987 		if (in_nullhost(addr)) {
1988 			imo->imo_multicast_if_index = 0;
1989 			break;
1990 		}
1991 		/*
1992 		 * The selected interface is identified by its local
1993 		 * IP address.  Find the interface and confirm that
1994 		 * it supports multicasting.
1995 		 */
1996 		s = pserialize_read_enter();
1997 		ifp = ip_multicast_if(&addr, &ifindex);
1998 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1999 			pserialize_read_exit(s);
2000 			error = EADDRNOTAVAIL;
2001 			break;
2002 		}
2003 		imo->imo_multicast_if_index = ifp->if_index;
2004 		pserialize_read_exit(s);
2005 		if (ifindex)
2006 			imo->imo_multicast_addr = addr;
2007 		else
2008 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
2009 		break;
2010 	    }
2011 
2012 	case IP_MULTICAST_TTL:
2013 		/*
2014 		 * Set the IP time-to-live for outgoing multicast packets.
2015 		 */
2016 		error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL);
2017 		break;
2018 
2019 	case IP_MULTICAST_LOOP:
2020 		/*
2021 		 * Set the loopback flag for outgoing multicast packets.
2022 		 * Must be zero or one.
2023 		 */
2024 		error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1);
2025 		break;
2026 
2027 	case IP_ADD_MEMBERSHIP: /* IPV6_JOIN_GROUP */
2028 		error = ip_add_membership(imo, sopt);
2029 		break;
2030 
2031 	case IP_DROP_MEMBERSHIP: /* IPV6_LEAVE_GROUP */
2032 		error = ip_drop_membership(imo, sopt);
2033 		break;
2034 
2035 	default:
2036 		error = EOPNOTSUPP;
2037 		break;
2038 	}
2039 
2040 	/*
2041 	 * If all options have default values, no need to keep the mbuf.
2042 	 */
2043 	if (imo->imo_multicast_if_index == 0 &&
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 		kmem_intr_free(imo, sizeof(*imo));
2048 		*pimo = NULL;
2049 	}
2050 
2051 	return error;
2052 }
2053 
2054 /*
2055  * Return the IP multicast options in response to user getsockopt().
2056  */
2057 int
ip_getmoptions(struct ip_moptions * imo,struct sockopt * sopt)2058 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt)
2059 {
2060 	struct in_addr addr;
2061 	uint8_t optval;
2062 	int error = 0;
2063 
2064 	/* imo is protected by solock or referenced only by the caller */
2065 
2066 	switch (sopt->sopt_name) {
2067 	case IP_MULTICAST_IF:
2068 		if (imo == NULL || imo->imo_multicast_if_index == 0)
2069 			addr = zeroin_addr;
2070 		else if (imo->imo_multicast_addr.s_addr) {
2071 			/* return the value user has set */
2072 			addr = imo->imo_multicast_addr;
2073 		} else {
2074 			struct ifnet *ifp;
2075 			struct in_ifaddr *ia = NULL;
2076 			int s = pserialize_read_enter();
2077 
2078 			ifp = if_byindex(imo->imo_multicast_if_index);
2079 			if (ifp != NULL) {
2080 				ia = in_get_ia_from_ifp(ifp);
2081 			}
2082 			addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
2083 			pserialize_read_exit(s);
2084 		}
2085 		error = sockopt_set(sopt, &addr, sizeof(addr));
2086 		break;
2087 
2088 	case IP_MULTICAST_TTL:
2089 		optval = imo ? imo->imo_multicast_ttl
2090 		    : IP_DEFAULT_MULTICAST_TTL;
2091 
2092 		error = sockopt_set(sopt, &optval, sizeof(optval));
2093 		break;
2094 
2095 	case IP_MULTICAST_LOOP:
2096 		optval = imo ? imo->imo_multicast_loop
2097 		    : IP_DEFAULT_MULTICAST_LOOP;
2098 
2099 		error = sockopt_set(sopt, &optval, sizeof(optval));
2100 		break;
2101 
2102 	default:
2103 		error = EOPNOTSUPP;
2104 	}
2105 
2106 	return error;
2107 }
2108 
2109 /*
2110  * Discard the IP multicast options.
2111  */
2112 void
ip_freemoptions(struct ip_moptions * imo)2113 ip_freemoptions(struct ip_moptions *imo)
2114 {
2115 	int i;
2116 
2117 	/* The owner of imo (inp) should be protected by solock */
2118 
2119 	if (imo != NULL) {
2120 		for (i = 0; i < imo->imo_num_memberships; ++i) {
2121 			struct in_multi *inm = imo->imo_membership[i];
2122 			in_delmulti(inm);
2123 			/* ifp should not leave thanks to solock */
2124 		}
2125 
2126 		kmem_intr_free(imo, sizeof(*imo));
2127 	}
2128 }
2129 
2130 /*
2131  * Routine called from ip_output() to loop back a copy of an IP multicast
2132  * packet to the input queue of a specified interface.  Note that this
2133  * calls the output routine of the loopback "driver", but with an interface
2134  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
2135  */
2136 static void
ip_mloopback(struct ifnet * ifp,struct mbuf * m,const struct sockaddr_in * dst)2137 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst)
2138 {
2139 	struct ip *ip;
2140 	struct mbuf *copym;
2141 
2142 	copym = m_copypacket(m, M_DONTWAIT);
2143 	if (copym != NULL &&
2144 	    (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
2145 		copym = m_pullup(copym, sizeof(struct ip));
2146 	if (copym == NULL)
2147 		return;
2148 	/*
2149 	 * We don't bother to fragment if the IP length is greater
2150 	 * than the interface's MTU.  Can this possibly matter?
2151 	 */
2152 	ip = mtod(copym, struct ip *);
2153 
2154 	if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
2155 		in_undefer_cksum_tcpudp(copym);
2156 		copym->m_pkthdr.csum_flags &=
2157 		    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
2158 	}
2159 
2160 	ip->ip_sum = 0;
2161 	ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
2162 	KERNEL_LOCK_UNLESS_NET_MPSAFE();
2163 	(void)looutput(ifp, copym, sintocsa(dst), NULL);
2164 	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
2165 }
2166 
2167 /*
2168  * Ensure sending address is valid.
2169  * Returns 0 on success, -1 if an error should be sent back or 1
2170  * if the packet could be dropped without error (protocol dependent).
2171  */
2172 static int
ip_ifaddrvalid(const struct in_ifaddr * ia)2173 ip_ifaddrvalid(const struct in_ifaddr *ia)
2174 {
2175 
2176 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
2177 		return 0;
2178 
2179 	if (ia->ia4_flags & IN_IFF_DUPLICATED)
2180 		return -1;
2181 	else if (ia->ia4_flags & (IN_IFF_TENTATIVE | IN_IFF_DETACHED))
2182 		return 1;
2183 
2184 	return 0;
2185 }
2186