xref: /freebsd/sys/netinet/ip_output.c (revision 9119bafbaf52e5d86e5879e54b38c5aba3efa2e9)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1990, 1993
5  *	The Regents of the University of California.  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 University 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 REGENTS 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 REGENTS 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  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_inet.h"
38 #include "opt_ipsec.h"
39 #include "opt_kern_tls.h"
40 #include "opt_mbuf_stress_test.h"
41 #include "opt_mpath.h"
42 #include "opt_ratelimit.h"
43 #include "opt_route.h"
44 #include "opt_rss.h"
45 #include "opt_sctp.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/kernel.h>
50 #include <sys/ktls.h>
51 #include <sys/lock.h>
52 #include <sys/malloc.h>
53 #include <sys/mbuf.h>
54 #include <sys/priv.h>
55 #include <sys/proc.h>
56 #include <sys/protosw.h>
57 #include <sys/rmlock.h>
58 #include <sys/sdt.h>
59 #include <sys/socket.h>
60 #include <sys/socketvar.h>
61 #include <sys/sysctl.h>
62 #include <sys/ucred.h>
63 
64 #include <net/if.h>
65 #include <net/if_var.h>
66 #include <net/if_llatbl.h>
67 #include <net/netisr.h>
68 #include <net/pfil.h>
69 #include <net/route.h>
70 #include <net/route/nhop.h>
71 #include <net/rss_config.h>
72 #include <net/vnet.h>
73 
74 #include <netinet/in.h>
75 #include <netinet/in_fib.h>
76 #include <netinet/in_kdtrace.h>
77 #include <netinet/in_systm.h>
78 #include <netinet/ip.h>
79 #include <netinet/in_fib.h>
80 #include <netinet/in_pcb.h>
81 #include <netinet/in_rss.h>
82 #include <netinet/in_var.h>
83 #include <netinet/ip_var.h>
84 #include <netinet/ip_options.h>
85 
86 #include <netinet/udp.h>
87 #include <netinet/udp_var.h>
88 
89 #if defined(SCTP) || defined(SCTP_SUPPORT)
90 #include <netinet/sctp.h>
91 #include <netinet/sctp_crc32.h>
92 #endif
93 
94 #include <netipsec/ipsec_support.h>
95 
96 #include <machine/in_cksum.h>
97 
98 #include <security/mac/mac_framework.h>
99 
100 #ifdef MBUF_STRESS_TEST
101 static int mbuf_frag_size = 0;
102 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
103 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
104 #endif
105 
106 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
107 
108 
109 extern int in_mcast_loop;
110 extern	struct protosw inetsw[];
111 
112 static inline int
113 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
114     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
115 {
116 	struct m_tag *fwd_tag = NULL;
117 	struct mbuf *m;
118 	struct in_addr odst;
119 	struct ip *ip;
120 	int pflags = PFIL_OUT;
121 
122 	if (flags & IP_FORWARDING)
123 		pflags |= PFIL_FWD;
124 
125 	m = *mp;
126 	ip = mtod(m, struct ip *);
127 
128 	/* Run through list of hooks for output packets. */
129 	odst.s_addr = ip->ip_dst.s_addr;
130 	switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
131 	case PFIL_DROPPED:
132 		*error = EACCES;
133 		/* FALLTHROUGH */
134 	case PFIL_CONSUMED:
135 		return 1; /* Finished */
136 	case PFIL_PASS:
137 		*error = 0;
138 	}
139 	m = *mp;
140 	ip = mtod(m, struct ip *);
141 
142 	/* See if destination IP address was changed by packet filter. */
143 	if (odst.s_addr != ip->ip_dst.s_addr) {
144 		m->m_flags |= M_SKIP_FIREWALL;
145 		/* If destination is now ourself drop to ip_input(). */
146 		if (in_localip(ip->ip_dst)) {
147 			m->m_flags |= M_FASTFWD_OURS;
148 			if (m->m_pkthdr.rcvif == NULL)
149 				m->m_pkthdr.rcvif = V_loif;
150 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
151 				m->m_pkthdr.csum_flags |=
152 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
153 				m->m_pkthdr.csum_data = 0xffff;
154 			}
155 			m->m_pkthdr.csum_flags |=
156 				CSUM_IP_CHECKED | CSUM_IP_VALID;
157 #if defined(SCTP) || defined(SCTP_SUPPORT)
158 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
159 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
160 #endif
161 			*error = netisr_queue(NETISR_IP, m);
162 			return 1; /* Finished */
163 		}
164 
165 		bzero(dst, sizeof(*dst));
166 		dst->sin_family = AF_INET;
167 		dst->sin_len = sizeof(*dst);
168 		dst->sin_addr = ip->ip_dst;
169 
170 		return -1; /* Reloop */
171 	}
172 	/* See if fib was changed by packet filter. */
173 	if ((*fibnum) != M_GETFIB(m)) {
174 		m->m_flags |= M_SKIP_FIREWALL;
175 		*fibnum = M_GETFIB(m);
176 		return -1; /* Reloop for FIB change */
177 	}
178 
179 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
180 	if (m->m_flags & M_FASTFWD_OURS) {
181 		if (m->m_pkthdr.rcvif == NULL)
182 			m->m_pkthdr.rcvif = V_loif;
183 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
184 			m->m_pkthdr.csum_flags |=
185 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
186 			m->m_pkthdr.csum_data = 0xffff;
187 		}
188 #if defined(SCTP) || defined(SCTP_SUPPORT)
189 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
190 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
191 #endif
192 		m->m_pkthdr.csum_flags |=
193 			CSUM_IP_CHECKED | CSUM_IP_VALID;
194 
195 		*error = netisr_queue(NETISR_IP, m);
196 		return 1; /* Finished */
197 	}
198 	/* Or forward to some other address? */
199 	if ((m->m_flags & M_IP_NEXTHOP) &&
200 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
201 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
202 		m->m_flags |= M_SKIP_FIREWALL;
203 		m->m_flags &= ~M_IP_NEXTHOP;
204 		m_tag_delete(m, fwd_tag);
205 
206 		return -1; /* Reloop for CHANGE of dst */
207 	}
208 
209 	return 0;
210 }
211 
212 static int
213 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
214     const struct sockaddr_in *gw, struct route *ro, bool stamp_tag)
215 {
216 #ifdef KERN_TLS
217 	struct ktls_session *tls = NULL;
218 #endif
219 	struct m_snd_tag *mst;
220 	int error;
221 
222 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
223 	mst = NULL;
224 
225 #ifdef KERN_TLS
226 	/*
227 	 * If this is an unencrypted TLS record, save a reference to
228 	 * the record.  This local reference is used to call
229 	 * ktls_output_eagain after the mbuf has been freed (thus
230 	 * dropping the mbuf's reference) in if_output.
231 	 */
232 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
233 		tls = ktls_hold(m->m_next->m_epg_tls);
234 		mst = tls->snd_tag;
235 
236 		/*
237 		 * If a TLS session doesn't have a valid tag, it must
238 		 * have had an earlier ifp mismatch, so drop this
239 		 * packet.
240 		 */
241 		if (mst == NULL) {
242 			error = EAGAIN;
243 			goto done;
244 		}
245 		/*
246 		 * Always stamp tags that include NIC ktls.
247 		 */
248 		stamp_tag = true;
249 	}
250 #endif
251 #ifdef RATELIMIT
252 	if (inp != NULL && mst == NULL) {
253 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
254 		    (inp->inp_snd_tag != NULL &&
255 		    inp->inp_snd_tag->ifp != ifp))
256 			in_pcboutput_txrtlmt(inp, ifp, m);
257 
258 		if (inp->inp_snd_tag != NULL)
259 			mst = inp->inp_snd_tag;
260 	}
261 #endif
262 	if (stamp_tag && mst != NULL) {
263 		KASSERT(m->m_pkthdr.rcvif == NULL,
264 		    ("trying to add a send tag to a forwarded packet"));
265 		if (mst->ifp != ifp) {
266 			error = EAGAIN;
267 			goto done;
268 		}
269 
270 		/* stamp send tag on mbuf */
271 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
272 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
273 	}
274 
275 	error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
276 
277 done:
278 	/* Check for route change invalidating send tags. */
279 #ifdef KERN_TLS
280 	if (tls != NULL) {
281 		if (error == EAGAIN)
282 			error = ktls_output_eagain(inp, tls);
283 		ktls_free(tls);
284 	}
285 #endif
286 #ifdef RATELIMIT
287 	if (error == EAGAIN)
288 		in_pcboutput_eagain(inp);
289 #endif
290 	return (error);
291 }
292 
293 /* rte<>ro_flags translation */
294 static inline void
295 rt_update_ro_flags(struct route *ro)
296 {
297 	int nh_flags = ro->ro_nh->nh_flags;
298 
299 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
300 
301 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
302 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
303 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
304 }
305 
306 /*
307  * IP output.  The packet in mbuf chain m contains a skeletal IP
308  * header (with len, off, ttl, proto, tos, src, dst).
309  * The mbuf chain containing the packet will be freed.
310  * The mbuf opt, if present, will not be freed.
311  * If route ro is present and has ro_rt initialized, route lookup would be
312  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
313  * then result of route lookup is stored in ro->ro_rt.
314  *
315  * In the IP forwarding case, the packet will arrive with options already
316  * inserted, so must have a NULL opt pointer.
317  */
318 int
319 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
320     struct ip_moptions *imo, struct inpcb *inp)
321 {
322 	struct rm_priotracker in_ifa_tracker;
323 	struct ip *ip;
324 	struct ifnet *ifp = NULL;	/* keep compiler happy */
325 	struct mbuf *m0;
326 	int hlen = sizeof (struct ip);
327 	int mtu;
328 	int error = 0;
329 	struct sockaddr_in *dst, sin;
330 	const struct sockaddr_in *gw;
331 	struct in_ifaddr *ia;
332 	struct in_addr src;
333 	int isbroadcast;
334 	uint16_t ip_len, ip_off;
335 	uint32_t fibnum;
336 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
337 	int no_route_but_check_spd = 0;
338 #endif
339 
340 	M_ASSERTPKTHDR(m);
341 	NET_EPOCH_ASSERT();
342 
343 	if (inp != NULL) {
344 		INP_LOCK_ASSERT(inp);
345 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
346 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
347 			m->m_pkthdr.flowid = inp->inp_flowid;
348 			M_HASHTYPE_SET(m, inp->inp_flowtype);
349 		}
350 #ifdef NUMA
351 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
352 #endif
353 	}
354 
355 	if (opt) {
356 		int len = 0;
357 		m = ip_insertoptions(m, opt, &len);
358 		if (len != 0)
359 			hlen = len; /* ip->ip_hl is updated above */
360 	}
361 	ip = mtod(m, struct ip *);
362 	ip_len = ntohs(ip->ip_len);
363 	ip_off = ntohs(ip->ip_off);
364 
365 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
366 		ip->ip_v = IPVERSION;
367 		ip->ip_hl = hlen >> 2;
368 		ip_fillid(ip);
369 	} else {
370 		/* Header already set, fetch hlen from there */
371 		hlen = ip->ip_hl << 2;
372 	}
373 	if ((flags & IP_FORWARDING) == 0)
374 		IPSTAT_INC(ips_localout);
375 
376 	/*
377 	 * dst/gw handling:
378 	 *
379 	 * gw is readonly but can point either to dst OR rt_gateway,
380 	 * therefore we need restore gw if we're redoing lookup.
381 	 */
382 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
383 	if (ro != NULL)
384 		dst = (struct sockaddr_in *)&ro->ro_dst;
385 	else
386 		dst = &sin;
387 	if (ro == NULL || ro->ro_nh == NULL) {
388 		bzero(dst, sizeof(*dst));
389 		dst->sin_family = AF_INET;
390 		dst->sin_len = sizeof(*dst);
391 		dst->sin_addr = ip->ip_dst;
392 	}
393 	gw = dst;
394 again:
395 	/*
396 	 * Validate route against routing table additions;
397 	 * a better/more specific route might have been added.
398 	 */
399 	if (inp != NULL && ro != NULL && ro->ro_nh != NULL)
400 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
401 	/*
402 	 * If there is a cached route,
403 	 * check that it is to the same destination
404 	 * and is still up.  If not, free it and try again.
405 	 * The address family should also be checked in case of sharing the
406 	 * cache with IPv6.
407 	 * Also check whether routing cache needs invalidation.
408 	 */
409 	if (ro != NULL && ro->ro_nh != NULL &&
410 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
411 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
412 		RO_INVALIDATE_CACHE(ro);
413 	ia = NULL;
414 	/*
415 	 * If routing to interface only, short circuit routing lookup.
416 	 * The use of an all-ones broadcast address implies this; an
417 	 * interface is specified by the broadcast address of an interface,
418 	 * or the destination address of a ptp interface.
419 	 */
420 	if (flags & IP_SENDONES) {
421 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
422 						      M_GETFIB(m)))) == NULL &&
423 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
424 						    M_GETFIB(m)))) == NULL) {
425 			IPSTAT_INC(ips_noroute);
426 			error = ENETUNREACH;
427 			goto bad;
428 		}
429 		ip->ip_dst.s_addr = INADDR_BROADCAST;
430 		dst->sin_addr = ip->ip_dst;
431 		ifp = ia->ia_ifp;
432 		mtu = ifp->if_mtu;
433 		ip->ip_ttl = 1;
434 		isbroadcast = 1;
435 		src = IA_SIN(ia)->sin_addr;
436 	} else if (flags & IP_ROUTETOIF) {
437 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
438 						    M_GETFIB(m)))) == NULL &&
439 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
440 						M_GETFIB(m)))) == NULL) {
441 			IPSTAT_INC(ips_noroute);
442 			error = ENETUNREACH;
443 			goto bad;
444 		}
445 		ifp = ia->ia_ifp;
446 		mtu = ifp->if_mtu;
447 		ip->ip_ttl = 1;
448 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
449 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
450 		src = IA_SIN(ia)->sin_addr;
451 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
452 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
453 		/*
454 		 * Bypass the normal routing lookup for multicast
455 		 * packets if the interface is specified.
456 		 */
457 		ifp = imo->imo_multicast_ifp;
458 		mtu = ifp->if_mtu;
459 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
460 		isbroadcast = 0;	/* fool gcc */
461 		/* Interface may have no addresses. */
462 		if (ia != NULL)
463 			src = IA_SIN(ia)->sin_addr;
464 		else
465 			src.s_addr = INADDR_ANY;
466 	} else if (ro != NULL) {
467 		if (ro->ro_nh == NULL) {
468 			/*
469 			 * We want to do any cloning requested by the link
470 			 * layer, as this is probably required in all cases
471 			 * for correct operation (as it is for ARP).
472 			 */
473 			uint32_t flowid;
474 #ifdef RADIX_MPATH
475 			flowid = ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr);
476 #else
477 			flowid = m->m_pkthdr.flowid;
478 #endif
479 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
480 			    NHR_REF, flowid);
481 
482 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
483 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
484 				/*
485 				 * There is no route for this packet, but it is
486 				 * possible that a matching SPD entry exists.
487 				 */
488 				no_route_but_check_spd = 1;
489 				mtu = 0; /* Silence GCC warning. */
490 				goto sendit;
491 #endif
492 				IPSTAT_INC(ips_noroute);
493 				error = EHOSTUNREACH;
494 				goto bad;
495 			}
496 		}
497 		ia = ifatoia(ro->ro_nh->nh_ifa);
498 		ifp = ro->ro_nh->nh_ifp;
499 		counter_u64_add(ro->ro_nh->nh_pksent, 1);
500 		rt_update_ro_flags(ro);
501 		if (ro->ro_nh->nh_flags & NHF_GATEWAY)
502 			gw = &ro->ro_nh->gw4_sa;
503 		if (ro->ro_nh->nh_flags & NHF_HOST)
504 			isbroadcast = (ro->ro_nh->nh_flags & NHF_BROADCAST);
505 		else if (ifp->if_flags & IFF_BROADCAST)
506 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
507 		else
508 			isbroadcast = 0;
509 		if (ro->ro_nh->nh_flags & NHF_HOST)
510 			mtu = ro->ro_nh->nh_mtu;
511 		else
512 			mtu = ifp->if_mtu;
513 		src = IA_SIN(ia)->sin_addr;
514 	} else {
515 		struct nhop_object *nh;
516 
517 		nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0);
518 		if (nh == NULL) {
519 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
520 			/*
521 			 * There is no route for this packet, but it is
522 			 * possible that a matching SPD entry exists.
523 			 */
524 			no_route_but_check_spd = 1;
525 			mtu = 0; /* Silence GCC warning. */
526 			goto sendit;
527 #endif
528 			IPSTAT_INC(ips_noroute);
529 			error = EHOSTUNREACH;
530 			goto bad;
531 		}
532 		ifp = nh->nh_ifp;
533 		mtu = nh->nh_mtu;
534 		/*
535 		 * We are rewriting here dst to be gw actually, contradicting
536 		 * comment at the beginning of the function. However, in this
537 		 * case we are always dealing with on stack dst.
538 		 * In case if pfil(9) sends us back to beginning of the
539 		 * function, the dst would be rewritten by ip_output_pfil().
540 		 */
541 		MPASS(dst == &sin);
542 		if (nh->nh_flags & NHF_GATEWAY)
543 			dst->sin_addr = nh->gw4_sa.sin_addr;
544 		ia = ifatoia(nh->nh_ifa);
545 		src = IA_SIN(ia)->sin_addr;
546 		isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
547 		    (NHF_HOST | NHF_BROADCAST)) ||
548 		    ((ifp->if_flags & IFF_BROADCAST) &&
549 		    in_ifaddr_broadcast(dst->sin_addr, ia)));
550 	}
551 
552 	/* Catch a possible divide by zero later. */
553 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
554 	    __func__, mtu, ro,
555 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
556 
557 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
558 		m->m_flags |= M_MCAST;
559 		/*
560 		 * IP destination address is multicast.  Make sure "gw"
561 		 * still points to the address in "ro".  (It may have been
562 		 * changed to point to a gateway address, above.)
563 		 */
564 		gw = dst;
565 		/*
566 		 * See if the caller provided any multicast options
567 		 */
568 		if (imo != NULL) {
569 			ip->ip_ttl = imo->imo_multicast_ttl;
570 			if (imo->imo_multicast_vif != -1)
571 				ip->ip_src.s_addr =
572 				    ip_mcast_src ?
573 				    ip_mcast_src(imo->imo_multicast_vif) :
574 				    INADDR_ANY;
575 		} else
576 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
577 		/*
578 		 * Confirm that the outgoing interface supports multicast.
579 		 */
580 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
581 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
582 				IPSTAT_INC(ips_noroute);
583 				error = ENETUNREACH;
584 				goto bad;
585 			}
586 		}
587 		/*
588 		 * If source address not specified yet, use address
589 		 * of outgoing interface.
590 		 */
591 		if (ip->ip_src.s_addr == INADDR_ANY)
592 			ip->ip_src = src;
593 
594 		if ((imo == NULL && in_mcast_loop) ||
595 		    (imo && imo->imo_multicast_loop)) {
596 			/*
597 			 * Loop back multicast datagram if not expressly
598 			 * forbidden to do so, even if we are not a member
599 			 * of the group; ip_input() will filter it later,
600 			 * thus deferring a hash lookup and mutex acquisition
601 			 * at the expense of a cheap copy using m_copym().
602 			 */
603 			ip_mloopback(ifp, m, hlen);
604 		} else {
605 			/*
606 			 * If we are acting as a multicast router, perform
607 			 * multicast forwarding as if the packet had just
608 			 * arrived on the interface to which we are about
609 			 * to send.  The multicast forwarding function
610 			 * recursively calls this function, using the
611 			 * IP_FORWARDING flag to prevent infinite recursion.
612 			 *
613 			 * Multicasts that are looped back by ip_mloopback(),
614 			 * above, will be forwarded by the ip_input() routine,
615 			 * if necessary.
616 			 */
617 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
618 				/*
619 				 * If rsvp daemon is not running, do not
620 				 * set ip_moptions. This ensures that the packet
621 				 * is multicast and not just sent down one link
622 				 * as prescribed by rsvpd.
623 				 */
624 				if (!V_rsvp_on)
625 					imo = NULL;
626 				if (ip_mforward &&
627 				    ip_mforward(ip, ifp, m, imo) != 0) {
628 					m_freem(m);
629 					goto done;
630 				}
631 			}
632 		}
633 
634 		/*
635 		 * Multicasts with a time-to-live of zero may be looped-
636 		 * back, above, but must not be transmitted on a network.
637 		 * Also, multicasts addressed to the loopback interface
638 		 * are not sent -- the above call to ip_mloopback() will
639 		 * loop back a copy. ip_input() will drop the copy if
640 		 * this host does not belong to the destination group on
641 		 * the loopback interface.
642 		 */
643 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
644 			m_freem(m);
645 			goto done;
646 		}
647 
648 		goto sendit;
649 	}
650 
651 	/*
652 	 * If the source address is not specified yet, use the address
653 	 * of the outoing interface.
654 	 */
655 	if (ip->ip_src.s_addr == INADDR_ANY)
656 		ip->ip_src = src;
657 
658 	/*
659 	 * Look for broadcast address and
660 	 * verify user is allowed to send
661 	 * such a packet.
662 	 */
663 	if (isbroadcast) {
664 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
665 			error = EADDRNOTAVAIL;
666 			goto bad;
667 		}
668 		if ((flags & IP_ALLOWBROADCAST) == 0) {
669 			error = EACCES;
670 			goto bad;
671 		}
672 		/* don't allow broadcast messages to be fragmented */
673 		if (ip_len > mtu) {
674 			error = EMSGSIZE;
675 			goto bad;
676 		}
677 		m->m_flags |= M_BCAST;
678 	} else {
679 		m->m_flags &= ~M_BCAST;
680 	}
681 
682 sendit:
683 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
684 	if (IPSEC_ENABLED(ipv4)) {
685 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
686 			if (error == EINPROGRESS)
687 				error = 0;
688 			goto done;
689 		}
690 	}
691 	/*
692 	 * Check if there was a route for this packet; return error if not.
693 	 */
694 	if (no_route_but_check_spd) {
695 		IPSTAT_INC(ips_noroute);
696 		error = EHOSTUNREACH;
697 		goto bad;
698 	}
699 	/* Update variables that are affected by ipsec4_output(). */
700 	ip = mtod(m, struct ip *);
701 	hlen = ip->ip_hl << 2;
702 #endif /* IPSEC */
703 
704 	/* Jump over all PFIL processing if hooks are not active. */
705 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
706 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
707 		    &error)) {
708 		case 1: /* Finished */
709 			goto done;
710 
711 		case 0: /* Continue normally */
712 			ip = mtod(m, struct ip *);
713 			break;
714 
715 		case -1: /* Need to try again */
716 			/* Reset everything for a new round */
717 			if (ro != NULL) {
718 				RO_NHFREE(ro);
719 				ro->ro_prepend = NULL;
720 			}
721 			gw = dst;
722 			ip = mtod(m, struct ip *);
723 			goto again;
724 
725 		}
726 	}
727 
728 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
729 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
730 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
731 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
732 			IPSTAT_INC(ips_badaddr);
733 			error = EADDRNOTAVAIL;
734 			goto bad;
735 		}
736 	}
737 
738 	m->m_pkthdr.csum_flags |= CSUM_IP;
739 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
740 		m = mb_unmapped_to_ext(m);
741 		if (m == NULL) {
742 			IPSTAT_INC(ips_odropped);
743 			error = ENOBUFS;
744 			goto bad;
745 		}
746 		in_delayed_cksum(m);
747 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
748 	} else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
749 		m = mb_unmapped_to_ext(m);
750 		if (m == NULL) {
751 			IPSTAT_INC(ips_odropped);
752 			error = ENOBUFS;
753 			goto bad;
754 		}
755 	}
756 #if defined(SCTP) || defined(SCTP_SUPPORT)
757 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
758 		m = mb_unmapped_to_ext(m);
759 		if (m == NULL) {
760 			IPSTAT_INC(ips_odropped);
761 			error = ENOBUFS;
762 			goto bad;
763 		}
764 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
765 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
766 	}
767 #endif
768 
769 	/*
770 	 * If small enough for interface, or the interface will take
771 	 * care of the fragmentation for us, we can just send directly.
772 	 */
773 	if (ip_len <= mtu ||
774 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
775 		ip->ip_sum = 0;
776 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
777 			ip->ip_sum = in_cksum(m, hlen);
778 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
779 		}
780 
781 		/*
782 		 * Record statistics for this interface address.
783 		 * With CSUM_TSO the byte/packet count will be slightly
784 		 * incorrect because we count the IP+TCP headers only
785 		 * once instead of for every generated packet.
786 		 */
787 		if (!(flags & IP_FORWARDING) && ia) {
788 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
789 				counter_u64_add(ia->ia_ifa.ifa_opackets,
790 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
791 			else
792 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
793 
794 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
795 		}
796 #ifdef MBUF_STRESS_TEST
797 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
798 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
799 #endif
800 		/*
801 		 * Reset layer specific mbuf flags
802 		 * to avoid confusing lower layers.
803 		 */
804 		m_clrprotoflags(m);
805 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
806 		error = ip_output_send(inp, ifp, m, gw, ro,
807 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
808 		goto done;
809 	}
810 
811 	/* Balk when DF bit is set or the interface didn't support TSO. */
812 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
813 		error = EMSGSIZE;
814 		IPSTAT_INC(ips_cantfrag);
815 		goto bad;
816 	}
817 
818 	/*
819 	 * Too large for interface; fragment if possible. If successful,
820 	 * on return, m will point to a list of packets to be sent.
821 	 */
822 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
823 	if (error)
824 		goto bad;
825 	for (; m; m = m0) {
826 		m0 = m->m_nextpkt;
827 		m->m_nextpkt = 0;
828 		if (error == 0) {
829 			/* Record statistics for this interface address. */
830 			if (ia != NULL) {
831 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
832 				counter_u64_add(ia->ia_ifa.ifa_obytes,
833 				    m->m_pkthdr.len);
834 			}
835 			/*
836 			 * Reset layer specific mbuf flags
837 			 * to avoid confusing upper layers.
838 			 */
839 			m_clrprotoflags(m);
840 
841 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
842 			    mtod(m, struct ip *), NULL);
843 			error = ip_output_send(inp, ifp, m, gw, ro, true);
844 		} else
845 			m_freem(m);
846 	}
847 
848 	if (error == 0)
849 		IPSTAT_INC(ips_fragmented);
850 
851 done:
852 	return (error);
853  bad:
854 	m_freem(m);
855 	goto done;
856 }
857 
858 /*
859  * Create a chain of fragments which fit the given mtu. m_frag points to the
860  * mbuf to be fragmented; on return it points to the chain with the fragments.
861  * Return 0 if no error. If error, m_frag may contain a partially built
862  * chain of fragments that should be freed by the caller.
863  *
864  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
865  */
866 int
867 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
868     u_long if_hwassist_flags)
869 {
870 	int error = 0;
871 	int hlen = ip->ip_hl << 2;
872 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
873 	int off;
874 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
875 	int firstlen;
876 	struct mbuf **mnext;
877 	int nfrags;
878 	uint16_t ip_len, ip_off;
879 
880 	ip_len = ntohs(ip->ip_len);
881 	ip_off = ntohs(ip->ip_off);
882 
883 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
884 		IPSTAT_INC(ips_cantfrag);
885 		return EMSGSIZE;
886 	}
887 
888 	/*
889 	 * Must be able to put at least 8 bytes per fragment.
890 	 */
891 	if (len < 8)
892 		return EMSGSIZE;
893 
894 	/*
895 	 * If the interface will not calculate checksums on
896 	 * fragmented packets, then do it here.
897 	 */
898 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
899 		m0 = mb_unmapped_to_ext(m0);
900 		if (m0 == NULL) {
901 			error = ENOBUFS;
902 			IPSTAT_INC(ips_odropped);
903 			goto done;
904 		}
905 		in_delayed_cksum(m0);
906 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
907 	}
908 #if defined(SCTP) || defined(SCTP_SUPPORT)
909 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
910 		m0 = mb_unmapped_to_ext(m0);
911 		if (m0 == NULL) {
912 			error = ENOBUFS;
913 			IPSTAT_INC(ips_odropped);
914 			goto done;
915 		}
916 		sctp_delayed_cksum(m0, hlen);
917 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
918 	}
919 #endif
920 	if (len > PAGE_SIZE) {
921 		/*
922 		 * Fragment large datagrams such that each segment
923 		 * contains a multiple of PAGE_SIZE amount of data,
924 		 * plus headers. This enables a receiver to perform
925 		 * page-flipping zero-copy optimizations.
926 		 *
927 		 * XXX When does this help given that sender and receiver
928 		 * could have different page sizes, and also mtu could
929 		 * be less than the receiver's page size ?
930 		 */
931 		int newlen;
932 
933 		off = MIN(mtu, m0->m_pkthdr.len);
934 
935 		/*
936 		 * firstlen (off - hlen) must be aligned on an
937 		 * 8-byte boundary
938 		 */
939 		if (off < hlen)
940 			goto smart_frag_failure;
941 		off = ((off - hlen) & ~7) + hlen;
942 		newlen = (~PAGE_MASK) & mtu;
943 		if ((newlen + sizeof (struct ip)) > mtu) {
944 			/* we failed, go back the default */
945 smart_frag_failure:
946 			newlen = len;
947 			off = hlen + len;
948 		}
949 		len = newlen;
950 
951 	} else {
952 		off = hlen + len;
953 	}
954 
955 	firstlen = off - hlen;
956 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
957 
958 	/*
959 	 * Loop through length of segment after first fragment,
960 	 * make new header and copy data of each part and link onto chain.
961 	 * Here, m0 is the original packet, m is the fragment being created.
962 	 * The fragments are linked off the m_nextpkt of the original
963 	 * packet, which after processing serves as the first fragment.
964 	 */
965 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
966 		struct ip *mhip;	/* ip header on the fragment */
967 		struct mbuf *m;
968 		int mhlen = sizeof (struct ip);
969 
970 		m = m_gethdr(M_NOWAIT, MT_DATA);
971 		if (m == NULL) {
972 			error = ENOBUFS;
973 			IPSTAT_INC(ips_odropped);
974 			goto done;
975 		}
976 		/*
977 		 * Make sure the complete packet header gets copied
978 		 * from the originating mbuf to the newly created
979 		 * mbuf. This also ensures that existing firewall
980 		 * classification(s), VLAN tags and so on get copied
981 		 * to the resulting fragmented packet(s):
982 		 */
983 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
984 			m_free(m);
985 			error = ENOBUFS;
986 			IPSTAT_INC(ips_odropped);
987 			goto done;
988 		}
989 		/*
990 		 * In the first mbuf, leave room for the link header, then
991 		 * copy the original IP header including options. The payload
992 		 * goes into an additional mbuf chain returned by m_copym().
993 		 */
994 		m->m_data += max_linkhdr;
995 		mhip = mtod(m, struct ip *);
996 		*mhip = *ip;
997 		if (hlen > sizeof (struct ip)) {
998 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
999 			mhip->ip_v = IPVERSION;
1000 			mhip->ip_hl = mhlen >> 2;
1001 		}
1002 		m->m_len = mhlen;
1003 		/* XXX do we need to add ip_off below ? */
1004 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1005 		if (off + len >= ip_len)
1006 			len = ip_len - off;
1007 		else
1008 			mhip->ip_off |= IP_MF;
1009 		mhip->ip_len = htons((u_short)(len + mhlen));
1010 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1011 		if (m->m_next == NULL) {	/* copy failed */
1012 			m_free(m);
1013 			error = ENOBUFS;	/* ??? */
1014 			IPSTAT_INC(ips_odropped);
1015 			goto done;
1016 		}
1017 		m->m_pkthdr.len = mhlen + len;
1018 #ifdef MAC
1019 		mac_netinet_fragment(m0, m);
1020 #endif
1021 		mhip->ip_off = htons(mhip->ip_off);
1022 		mhip->ip_sum = 0;
1023 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1024 			mhip->ip_sum = in_cksum(m, mhlen);
1025 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1026 		}
1027 		*mnext = m;
1028 		mnext = &m->m_nextpkt;
1029 	}
1030 	IPSTAT_ADD(ips_ofragments, nfrags);
1031 
1032 	/*
1033 	 * Update first fragment by trimming what's been copied out
1034 	 * and updating header.
1035 	 */
1036 	m_adj(m0, hlen + firstlen - ip_len);
1037 	m0->m_pkthdr.len = hlen + firstlen;
1038 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1039 	ip->ip_off = htons(ip_off | IP_MF);
1040 	ip->ip_sum = 0;
1041 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1042 		ip->ip_sum = in_cksum(m0, hlen);
1043 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1044 	}
1045 
1046 done:
1047 	*m_frag = m0;
1048 	return error;
1049 }
1050 
1051 void
1052 in_delayed_cksum(struct mbuf *m)
1053 {
1054 	struct ip *ip;
1055 	struct udphdr *uh;
1056 	uint16_t cklen, csum, offset;
1057 
1058 	ip = mtod(m, struct ip *);
1059 	offset = ip->ip_hl << 2 ;
1060 
1061 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1062 		/* if udp header is not in the first mbuf copy udplen */
1063 		if (offset + sizeof(struct udphdr) > m->m_len) {
1064 			m_copydata(m, offset + offsetof(struct udphdr,
1065 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1066 			cklen = ntohs(cklen);
1067 		} else {
1068 			uh = (struct udphdr *)mtodo(m, offset);
1069 			cklen = ntohs(uh->uh_ulen);
1070 		}
1071 		csum = in_cksum_skip(m, cklen + offset, offset);
1072 		if (csum == 0)
1073 			csum = 0xffff;
1074 	} else {
1075 		cklen = ntohs(ip->ip_len);
1076 		csum = in_cksum_skip(m, cklen, offset);
1077 	}
1078 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1079 
1080 	if (offset + sizeof(csum) > m->m_len)
1081 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1082 	else
1083 		*(u_short *)mtodo(m, offset) = csum;
1084 }
1085 
1086 /*
1087  * IP socket option processing.
1088  */
1089 int
1090 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1091 {
1092 	struct inpcb *inp = sotoinpcb(so);
1093 	int	error, optval;
1094 #ifdef	RSS
1095 	uint32_t rss_bucket;
1096 	int retval;
1097 #endif
1098 
1099 	error = optval = 0;
1100 	if (sopt->sopt_level != IPPROTO_IP) {
1101 		error = EINVAL;
1102 
1103 		if (sopt->sopt_level == SOL_SOCKET &&
1104 		    sopt->sopt_dir == SOPT_SET) {
1105 			switch (sopt->sopt_name) {
1106 			case SO_REUSEADDR:
1107 				INP_WLOCK(inp);
1108 				if ((so->so_options & SO_REUSEADDR) != 0)
1109 					inp->inp_flags2 |= INP_REUSEADDR;
1110 				else
1111 					inp->inp_flags2 &= ~INP_REUSEADDR;
1112 				INP_WUNLOCK(inp);
1113 				error = 0;
1114 				break;
1115 			case SO_REUSEPORT:
1116 				INP_WLOCK(inp);
1117 				if ((so->so_options & SO_REUSEPORT) != 0)
1118 					inp->inp_flags2 |= INP_REUSEPORT;
1119 				else
1120 					inp->inp_flags2 &= ~INP_REUSEPORT;
1121 				INP_WUNLOCK(inp);
1122 				error = 0;
1123 				break;
1124 			case SO_REUSEPORT_LB:
1125 				INP_WLOCK(inp);
1126 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1127 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1128 				else
1129 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1130 				INP_WUNLOCK(inp);
1131 				error = 0;
1132 				break;
1133 			case SO_SETFIB:
1134 				INP_WLOCK(inp);
1135 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1136 				INP_WUNLOCK(inp);
1137 				error = 0;
1138 				break;
1139 			case SO_MAX_PACING_RATE:
1140 #ifdef RATELIMIT
1141 				INP_WLOCK(inp);
1142 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1143 				INP_WUNLOCK(inp);
1144 				error = 0;
1145 #else
1146 				error = EOPNOTSUPP;
1147 #endif
1148 				break;
1149 			default:
1150 				break;
1151 			}
1152 		}
1153 		return (error);
1154 	}
1155 
1156 	switch (sopt->sopt_dir) {
1157 	case SOPT_SET:
1158 		switch (sopt->sopt_name) {
1159 		case IP_OPTIONS:
1160 #ifdef notyet
1161 		case IP_RETOPTS:
1162 #endif
1163 		{
1164 			struct mbuf *m;
1165 			if (sopt->sopt_valsize > MLEN) {
1166 				error = EMSGSIZE;
1167 				break;
1168 			}
1169 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1170 			if (m == NULL) {
1171 				error = ENOBUFS;
1172 				break;
1173 			}
1174 			m->m_len = sopt->sopt_valsize;
1175 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1176 					    m->m_len);
1177 			if (error) {
1178 				m_free(m);
1179 				break;
1180 			}
1181 			INP_WLOCK(inp);
1182 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1183 			INP_WUNLOCK(inp);
1184 			return (error);
1185 		}
1186 
1187 		case IP_BINDANY:
1188 			if (sopt->sopt_td != NULL) {
1189 				error = priv_check(sopt->sopt_td,
1190 				    PRIV_NETINET_BINDANY);
1191 				if (error)
1192 					break;
1193 			}
1194 			/* FALLTHROUGH */
1195 		case IP_BINDMULTI:
1196 #ifdef	RSS
1197 		case IP_RSS_LISTEN_BUCKET:
1198 #endif
1199 		case IP_TOS:
1200 		case IP_TTL:
1201 		case IP_MINTTL:
1202 		case IP_RECVOPTS:
1203 		case IP_RECVRETOPTS:
1204 		case IP_ORIGDSTADDR:
1205 		case IP_RECVDSTADDR:
1206 		case IP_RECVTTL:
1207 		case IP_RECVIF:
1208 		case IP_ONESBCAST:
1209 		case IP_DONTFRAG:
1210 		case IP_RECVTOS:
1211 		case IP_RECVFLOWID:
1212 #ifdef	RSS
1213 		case IP_RECVRSSBUCKETID:
1214 #endif
1215 			error = sooptcopyin(sopt, &optval, sizeof optval,
1216 					    sizeof optval);
1217 			if (error)
1218 				break;
1219 
1220 			switch (sopt->sopt_name) {
1221 			case IP_TOS:
1222 				inp->inp_ip_tos = optval;
1223 				break;
1224 
1225 			case IP_TTL:
1226 				inp->inp_ip_ttl = optval;
1227 				break;
1228 
1229 			case IP_MINTTL:
1230 				if (optval >= 0 && optval <= MAXTTL)
1231 					inp->inp_ip_minttl = optval;
1232 				else
1233 					error = EINVAL;
1234 				break;
1235 
1236 #define	OPTSET(bit) do {						\
1237 	INP_WLOCK(inp);							\
1238 	if (optval)							\
1239 		inp->inp_flags |= bit;					\
1240 	else								\
1241 		inp->inp_flags &= ~bit;					\
1242 	INP_WUNLOCK(inp);						\
1243 } while (0)
1244 
1245 #define	OPTSET2(bit, val) do {						\
1246 	INP_WLOCK(inp);							\
1247 	if (val)							\
1248 		inp->inp_flags2 |= bit;					\
1249 	else								\
1250 		inp->inp_flags2 &= ~bit;				\
1251 	INP_WUNLOCK(inp);						\
1252 } while (0)
1253 
1254 			case IP_RECVOPTS:
1255 				OPTSET(INP_RECVOPTS);
1256 				break;
1257 
1258 			case IP_RECVRETOPTS:
1259 				OPTSET(INP_RECVRETOPTS);
1260 				break;
1261 
1262 			case IP_RECVDSTADDR:
1263 				OPTSET(INP_RECVDSTADDR);
1264 				break;
1265 
1266 			case IP_ORIGDSTADDR:
1267 				OPTSET2(INP_ORIGDSTADDR, optval);
1268 				break;
1269 
1270 			case IP_RECVTTL:
1271 				OPTSET(INP_RECVTTL);
1272 				break;
1273 
1274 			case IP_RECVIF:
1275 				OPTSET(INP_RECVIF);
1276 				break;
1277 
1278 			case IP_ONESBCAST:
1279 				OPTSET(INP_ONESBCAST);
1280 				break;
1281 			case IP_DONTFRAG:
1282 				OPTSET(INP_DONTFRAG);
1283 				break;
1284 			case IP_BINDANY:
1285 				OPTSET(INP_BINDANY);
1286 				break;
1287 			case IP_RECVTOS:
1288 				OPTSET(INP_RECVTOS);
1289 				break;
1290 			case IP_BINDMULTI:
1291 				OPTSET2(INP_BINDMULTI, optval);
1292 				break;
1293 			case IP_RECVFLOWID:
1294 				OPTSET2(INP_RECVFLOWID, optval);
1295 				break;
1296 #ifdef	RSS
1297 			case IP_RSS_LISTEN_BUCKET:
1298 				if ((optval >= 0) &&
1299 				    (optval < rss_getnumbuckets())) {
1300 					inp->inp_rss_listen_bucket = optval;
1301 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1302 				} else {
1303 					error = EINVAL;
1304 				}
1305 				break;
1306 			case IP_RECVRSSBUCKETID:
1307 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1308 				break;
1309 #endif
1310 			}
1311 			break;
1312 #undef OPTSET
1313 #undef OPTSET2
1314 
1315 		/*
1316 		 * Multicast socket options are processed by the in_mcast
1317 		 * module.
1318 		 */
1319 		case IP_MULTICAST_IF:
1320 		case IP_MULTICAST_VIF:
1321 		case IP_MULTICAST_TTL:
1322 		case IP_MULTICAST_LOOP:
1323 		case IP_ADD_MEMBERSHIP:
1324 		case IP_DROP_MEMBERSHIP:
1325 		case IP_ADD_SOURCE_MEMBERSHIP:
1326 		case IP_DROP_SOURCE_MEMBERSHIP:
1327 		case IP_BLOCK_SOURCE:
1328 		case IP_UNBLOCK_SOURCE:
1329 		case IP_MSFILTER:
1330 		case MCAST_JOIN_GROUP:
1331 		case MCAST_LEAVE_GROUP:
1332 		case MCAST_JOIN_SOURCE_GROUP:
1333 		case MCAST_LEAVE_SOURCE_GROUP:
1334 		case MCAST_BLOCK_SOURCE:
1335 		case MCAST_UNBLOCK_SOURCE:
1336 			error = inp_setmoptions(inp, sopt);
1337 			break;
1338 
1339 		case IP_PORTRANGE:
1340 			error = sooptcopyin(sopt, &optval, sizeof optval,
1341 					    sizeof optval);
1342 			if (error)
1343 				break;
1344 
1345 			INP_WLOCK(inp);
1346 			switch (optval) {
1347 			case IP_PORTRANGE_DEFAULT:
1348 				inp->inp_flags &= ~(INP_LOWPORT);
1349 				inp->inp_flags &= ~(INP_HIGHPORT);
1350 				break;
1351 
1352 			case IP_PORTRANGE_HIGH:
1353 				inp->inp_flags &= ~(INP_LOWPORT);
1354 				inp->inp_flags |= INP_HIGHPORT;
1355 				break;
1356 
1357 			case IP_PORTRANGE_LOW:
1358 				inp->inp_flags &= ~(INP_HIGHPORT);
1359 				inp->inp_flags |= INP_LOWPORT;
1360 				break;
1361 
1362 			default:
1363 				error = EINVAL;
1364 				break;
1365 			}
1366 			INP_WUNLOCK(inp);
1367 			break;
1368 
1369 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1370 		case IP_IPSEC_POLICY:
1371 			if (IPSEC_ENABLED(ipv4)) {
1372 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1373 				break;
1374 			}
1375 			/* FALLTHROUGH */
1376 #endif /* IPSEC */
1377 
1378 		default:
1379 			error = ENOPROTOOPT;
1380 			break;
1381 		}
1382 		break;
1383 
1384 	case SOPT_GET:
1385 		switch (sopt->sopt_name) {
1386 		case IP_OPTIONS:
1387 		case IP_RETOPTS:
1388 			INP_RLOCK(inp);
1389 			if (inp->inp_options) {
1390 				struct mbuf *options;
1391 
1392 				options = m_copym(inp->inp_options, 0,
1393 				    M_COPYALL, M_NOWAIT);
1394 				INP_RUNLOCK(inp);
1395 				if (options != NULL) {
1396 					error = sooptcopyout(sopt,
1397 							     mtod(options, char *),
1398 							     options->m_len);
1399 					m_freem(options);
1400 				} else
1401 					error = ENOMEM;
1402 			} else {
1403 				INP_RUNLOCK(inp);
1404 				sopt->sopt_valsize = 0;
1405 			}
1406 			break;
1407 
1408 		case IP_TOS:
1409 		case IP_TTL:
1410 		case IP_MINTTL:
1411 		case IP_RECVOPTS:
1412 		case IP_RECVRETOPTS:
1413 		case IP_ORIGDSTADDR:
1414 		case IP_RECVDSTADDR:
1415 		case IP_RECVTTL:
1416 		case IP_RECVIF:
1417 		case IP_PORTRANGE:
1418 		case IP_ONESBCAST:
1419 		case IP_DONTFRAG:
1420 		case IP_BINDANY:
1421 		case IP_RECVTOS:
1422 		case IP_BINDMULTI:
1423 		case IP_FLOWID:
1424 		case IP_FLOWTYPE:
1425 		case IP_RECVFLOWID:
1426 #ifdef	RSS
1427 		case IP_RSSBUCKETID:
1428 		case IP_RECVRSSBUCKETID:
1429 #endif
1430 			switch (sopt->sopt_name) {
1431 
1432 			case IP_TOS:
1433 				optval = inp->inp_ip_tos;
1434 				break;
1435 
1436 			case IP_TTL:
1437 				optval = inp->inp_ip_ttl;
1438 				break;
1439 
1440 			case IP_MINTTL:
1441 				optval = inp->inp_ip_minttl;
1442 				break;
1443 
1444 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1445 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1446 
1447 			case IP_RECVOPTS:
1448 				optval = OPTBIT(INP_RECVOPTS);
1449 				break;
1450 
1451 			case IP_RECVRETOPTS:
1452 				optval = OPTBIT(INP_RECVRETOPTS);
1453 				break;
1454 
1455 			case IP_RECVDSTADDR:
1456 				optval = OPTBIT(INP_RECVDSTADDR);
1457 				break;
1458 
1459 			case IP_ORIGDSTADDR:
1460 				optval = OPTBIT2(INP_ORIGDSTADDR);
1461 				break;
1462 
1463 			case IP_RECVTTL:
1464 				optval = OPTBIT(INP_RECVTTL);
1465 				break;
1466 
1467 			case IP_RECVIF:
1468 				optval = OPTBIT(INP_RECVIF);
1469 				break;
1470 
1471 			case IP_PORTRANGE:
1472 				if (inp->inp_flags & INP_HIGHPORT)
1473 					optval = IP_PORTRANGE_HIGH;
1474 				else if (inp->inp_flags & INP_LOWPORT)
1475 					optval = IP_PORTRANGE_LOW;
1476 				else
1477 					optval = 0;
1478 				break;
1479 
1480 			case IP_ONESBCAST:
1481 				optval = OPTBIT(INP_ONESBCAST);
1482 				break;
1483 			case IP_DONTFRAG:
1484 				optval = OPTBIT(INP_DONTFRAG);
1485 				break;
1486 			case IP_BINDANY:
1487 				optval = OPTBIT(INP_BINDANY);
1488 				break;
1489 			case IP_RECVTOS:
1490 				optval = OPTBIT(INP_RECVTOS);
1491 				break;
1492 			case IP_FLOWID:
1493 				optval = inp->inp_flowid;
1494 				break;
1495 			case IP_FLOWTYPE:
1496 				optval = inp->inp_flowtype;
1497 				break;
1498 			case IP_RECVFLOWID:
1499 				optval = OPTBIT2(INP_RECVFLOWID);
1500 				break;
1501 #ifdef	RSS
1502 			case IP_RSSBUCKETID:
1503 				retval = rss_hash2bucket(inp->inp_flowid,
1504 				    inp->inp_flowtype,
1505 				    &rss_bucket);
1506 				if (retval == 0)
1507 					optval = rss_bucket;
1508 				else
1509 					error = EINVAL;
1510 				break;
1511 			case IP_RECVRSSBUCKETID:
1512 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1513 				break;
1514 #endif
1515 			case IP_BINDMULTI:
1516 				optval = OPTBIT2(INP_BINDMULTI);
1517 				break;
1518 			}
1519 			error = sooptcopyout(sopt, &optval, sizeof optval);
1520 			break;
1521 
1522 		/*
1523 		 * Multicast socket options are processed by the in_mcast
1524 		 * module.
1525 		 */
1526 		case IP_MULTICAST_IF:
1527 		case IP_MULTICAST_VIF:
1528 		case IP_MULTICAST_TTL:
1529 		case IP_MULTICAST_LOOP:
1530 		case IP_MSFILTER:
1531 			error = inp_getmoptions(inp, sopt);
1532 			break;
1533 
1534 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1535 		case IP_IPSEC_POLICY:
1536 			if (IPSEC_ENABLED(ipv4)) {
1537 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1538 				break;
1539 			}
1540 			/* FALLTHROUGH */
1541 #endif /* IPSEC */
1542 
1543 		default:
1544 			error = ENOPROTOOPT;
1545 			break;
1546 		}
1547 		break;
1548 	}
1549 	return (error);
1550 }
1551 
1552 /*
1553  * Routine called from ip_output() to loop back a copy of an IP multicast
1554  * packet to the input queue of a specified interface.  Note that this
1555  * calls the output routine of the loopback "driver", but with an interface
1556  * pointer that might NOT be a loopback interface -- evil, but easier than
1557  * replicating that code here.
1558  */
1559 static void
1560 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1561 {
1562 	struct ip *ip;
1563 	struct mbuf *copym;
1564 
1565 	/*
1566 	 * Make a deep copy of the packet because we're going to
1567 	 * modify the pack in order to generate checksums.
1568 	 */
1569 	copym = m_dup(m, M_NOWAIT);
1570 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1571 		copym = m_pullup(copym, hlen);
1572 	if (copym != NULL) {
1573 		/* If needed, compute the checksum and mark it as valid. */
1574 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1575 			in_delayed_cksum(copym);
1576 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1577 			copym->m_pkthdr.csum_flags |=
1578 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1579 			copym->m_pkthdr.csum_data = 0xffff;
1580 		}
1581 		/*
1582 		 * We don't bother to fragment if the IP length is greater
1583 		 * than the interface's MTU.  Can this possibly matter?
1584 		 */
1585 		ip = mtod(copym, struct ip *);
1586 		ip->ip_sum = 0;
1587 		ip->ip_sum = in_cksum(copym, hlen);
1588 		if_simloop(ifp, copym, AF_INET, 0);
1589 	}
1590 }
1591