xref: /freebsd/sys/netinet/ip_output.c (revision 2faf504d1ab821fe2b9df9d2afb49bb35e1334f4)
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_ratelimit.h"
42 #include "opt_route.h"
43 #include "opt_rss.h"
44 #include "opt_sctp.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
49 #include <sys/ktls.h>
50 #include <sys/lock.h>
51 #include <sys/malloc.h>
52 #include <sys/mbuf.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/protosw.h>
56 #include <sys/rmlock.h>
57 #include <sys/sdt.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
60 #include <sys/sysctl.h>
61 #include <sys/ucred.h>
62 
63 #include <net/if.h>
64 #include <net/if_var.h>
65 #include <net/if_vlan_var.h>
66 #include <net/if_llatbl.h>
67 #include <net/ethernet.h>
68 #include <net/netisr.h>
69 #include <net/pfil.h>
70 #include <net/route.h>
71 #include <net/route/nhop.h>
72 #include <net/rss_config.h>
73 #include <net/vnet.h>
74 
75 #include <netinet/in.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_kdtrace.h>
78 #include <netinet/in_systm.h>
79 #include <netinet/ip.h>
80 #include <netinet/in_fib.h>
81 #include <netinet/in_pcb.h>
82 #include <netinet/in_rss.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip_var.h>
85 #include <netinet/ip_options.h>
86 #include <netinet/ip_mroute.h>
87 
88 #include <netinet/udp.h>
89 #include <netinet/udp_var.h>
90 
91 #if defined(SCTP) || defined(SCTP_SUPPORT)
92 #include <netinet/sctp.h>
93 #include <netinet/sctp_crc32.h>
94 #endif
95 
96 #include <netipsec/ipsec_support.h>
97 
98 #include <machine/in_cksum.h>
99 
100 #include <security/mac/mac_framework.h>
101 
102 #ifdef MBUF_STRESS_TEST
103 static int mbuf_frag_size = 0;
104 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
105 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
106 #endif
107 
108 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
109 
110 extern int in_mcast_loop;
111 extern	struct protosw inetsw[];
112 
113 static inline int
114 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
115     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
116 {
117 	struct m_tag *fwd_tag = NULL;
118 	struct mbuf *m;
119 	struct in_addr odst;
120 	struct ip *ip;
121 	int pflags = PFIL_OUT;
122 
123 	if (flags & IP_FORWARDING)
124 		pflags |= PFIL_FWD;
125 
126 	m = *mp;
127 	ip = mtod(m, struct ip *);
128 
129 	/* Run through list of hooks for output packets. */
130 	odst.s_addr = ip->ip_dst.s_addr;
131 	switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
132 	case PFIL_DROPPED:
133 		*error = EACCES;
134 		/* FALLTHROUGH */
135 	case PFIL_CONSUMED:
136 		return 1; /* Finished */
137 	case PFIL_PASS:
138 		*error = 0;
139 	}
140 	m = *mp;
141 	ip = mtod(m, struct ip *);
142 
143 	/* See if destination IP address was changed by packet filter. */
144 	if (odst.s_addr != ip->ip_dst.s_addr) {
145 		m->m_flags |= M_SKIP_FIREWALL;
146 		/* If destination is now ourself drop to ip_input(). */
147 		if (in_localip(ip->ip_dst)) {
148 			m->m_flags |= M_FASTFWD_OURS;
149 			if (m->m_pkthdr.rcvif == NULL)
150 				m->m_pkthdr.rcvif = V_loif;
151 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
152 				m->m_pkthdr.csum_flags |=
153 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
154 				m->m_pkthdr.csum_data = 0xffff;
155 			}
156 			m->m_pkthdr.csum_flags |=
157 				CSUM_IP_CHECKED | CSUM_IP_VALID;
158 #if defined(SCTP) || defined(SCTP_SUPPORT)
159 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
160 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
161 #endif
162 			*error = netisr_queue(NETISR_IP, m);
163 			return 1; /* Finished */
164 		}
165 
166 		bzero(dst, sizeof(*dst));
167 		dst->sin_family = AF_INET;
168 		dst->sin_len = sizeof(*dst);
169 		dst->sin_addr = ip->ip_dst;
170 
171 		return -1; /* Reloop */
172 	}
173 	/* See if fib was changed by packet filter. */
174 	if ((*fibnum) != M_GETFIB(m)) {
175 		m->m_flags |= M_SKIP_FIREWALL;
176 		*fibnum = M_GETFIB(m);
177 		return -1; /* Reloop for FIB change */
178 	}
179 
180 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
181 	if (m->m_flags & M_FASTFWD_OURS) {
182 		if (m->m_pkthdr.rcvif == NULL)
183 			m->m_pkthdr.rcvif = V_loif;
184 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
185 			m->m_pkthdr.csum_flags |=
186 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
187 			m->m_pkthdr.csum_data = 0xffff;
188 		}
189 #if defined(SCTP) || defined(SCTP_SUPPORT)
190 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
191 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
192 #endif
193 		m->m_pkthdr.csum_flags |=
194 			CSUM_IP_CHECKED | CSUM_IP_VALID;
195 
196 		*error = netisr_queue(NETISR_IP, m);
197 		return 1; /* Finished */
198 	}
199 	/* Or forward to some other address? */
200 	if ((m->m_flags & M_IP_NEXTHOP) &&
201 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
202 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
203 		m->m_flags |= M_SKIP_FIREWALL;
204 		m->m_flags &= ~M_IP_NEXTHOP;
205 		m_tag_delete(m, fwd_tag);
206 
207 		return -1; /* Reloop for CHANGE of dst */
208 	}
209 
210 	return 0;
211 }
212 
213 static int
214 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
215     const struct sockaddr_in *gw, struct route *ro, bool stamp_tag)
216 {
217 #ifdef KERN_TLS
218 	struct ktls_session *tls = NULL;
219 #endif
220 	struct m_snd_tag *mst;
221 	int error;
222 
223 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
224 	mst = NULL;
225 
226 #ifdef KERN_TLS
227 	/*
228 	 * If this is an unencrypted TLS record, save a reference to
229 	 * the record.  This local reference is used to call
230 	 * ktls_output_eagain after the mbuf has been freed (thus
231 	 * dropping the mbuf's reference) in if_output.
232 	 */
233 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
234 		tls = ktls_hold(m->m_next->m_epg_tls);
235 		mst = tls->snd_tag;
236 
237 		/*
238 		 * If a TLS session doesn't have a valid tag, it must
239 		 * have had an earlier ifp mismatch, so drop this
240 		 * packet.
241 		 */
242 		if (mst == NULL) {
243 			error = EAGAIN;
244 			goto done;
245 		}
246 		/*
247 		 * Always stamp tags that include NIC ktls.
248 		 */
249 		stamp_tag = true;
250 	}
251 #endif
252 #ifdef RATELIMIT
253 	if (inp != NULL && mst == NULL) {
254 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
255 		    (inp->inp_snd_tag != NULL &&
256 		    inp->inp_snd_tag->ifp != ifp))
257 			in_pcboutput_txrtlmt(inp, ifp, m);
258 
259 		if (inp->inp_snd_tag != NULL)
260 			mst = inp->inp_snd_tag;
261 	}
262 #endif
263 	if (stamp_tag && mst != NULL) {
264 		KASSERT(m->m_pkthdr.rcvif == NULL,
265 		    ("trying to add a send tag to a forwarded packet"));
266 		if (mst->ifp != ifp) {
267 			error = EAGAIN;
268 			goto done;
269 		}
270 
271 		/* stamp send tag on mbuf */
272 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
273 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
274 	}
275 
276 	error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
277 
278 done:
279 	/* Check for route change invalidating send tags. */
280 #ifdef KERN_TLS
281 	if (tls != NULL) {
282 		if (error == EAGAIN)
283 			error = ktls_output_eagain(inp, tls);
284 		ktls_free(tls);
285 	}
286 #endif
287 #ifdef RATELIMIT
288 	if (error == EAGAIN)
289 		in_pcboutput_eagain(inp);
290 #endif
291 	return (error);
292 }
293 
294 /* rte<>ro_flags translation */
295 static inline void
296 rt_update_ro_flags(struct route *ro, const struct nhop_object *nh)
297 {
298 	int nh_flags = nh->nh_flags;
299 
300 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
301 
302 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
303 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
304 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
305 }
306 
307 /*
308  * IP output.  The packet in mbuf chain m contains a skeletal IP
309  * header (with len, off, ttl, proto, tos, src, dst).
310  * The mbuf chain containing the packet will be freed.
311  * The mbuf opt, if present, will not be freed.
312  * If route ro is present and has ro_rt initialized, route lookup would be
313  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
314  * then result of route lookup is stored in ro->ro_rt.
315  *
316  * In the IP forwarding case, the packet will arrive with options already
317  * inserted, so must have a NULL opt pointer.
318  */
319 int
320 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
321     struct ip_moptions *imo, struct inpcb *inp)
322 {
323 	MROUTER_RLOCK_TRACKER;
324 	struct rm_priotracker in_ifa_tracker;
325 	struct ip *ip;
326 	struct ifnet *ifp = NULL;	/* keep compiler happy */
327 	struct mbuf *m0;
328 	int hlen = sizeof (struct ip);
329 	int mtu = 0;
330 	int error = 0;
331 	int vlan_pcp = -1;
332 	struct sockaddr_in *dst, sin;
333 	const struct sockaddr_in *gw;
334 	struct in_ifaddr *ia = NULL;
335 	struct in_addr src;
336 	int isbroadcast;
337 	uint16_t ip_len, ip_off;
338 	uint32_t fibnum;
339 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
340 	int no_route_but_check_spd = 0;
341 #endif
342 
343 	M_ASSERTPKTHDR(m);
344 	NET_EPOCH_ASSERT();
345 
346 	if (inp != NULL) {
347 		INP_LOCK_ASSERT(inp);
348 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
349 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
350 			m->m_pkthdr.flowid = inp->inp_flowid;
351 			M_HASHTYPE_SET(m, inp->inp_flowtype);
352 		}
353 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
354 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
355 			    INP_2PCP_SHIFT;
356 #ifdef NUMA
357 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
358 #endif
359 	}
360 
361 	if (opt) {
362 		int len = 0;
363 		m = ip_insertoptions(m, opt, &len);
364 		if (len != 0)
365 			hlen = len; /* ip->ip_hl is updated above */
366 	}
367 	ip = mtod(m, struct ip *);
368 	ip_len = ntohs(ip->ip_len);
369 	ip_off = ntohs(ip->ip_off);
370 
371 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
372 		ip->ip_v = IPVERSION;
373 		ip->ip_hl = hlen >> 2;
374 		ip_fillid(ip);
375 	} else {
376 		/* Header already set, fetch hlen from there */
377 		hlen = ip->ip_hl << 2;
378 	}
379 	if ((flags & IP_FORWARDING) == 0)
380 		IPSTAT_INC(ips_localout);
381 
382 	/*
383 	 * dst/gw handling:
384 	 *
385 	 * gw is readonly but can point either to dst OR rt_gateway,
386 	 * therefore we need restore gw if we're redoing lookup.
387 	 */
388 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
389 	if (ro != NULL)
390 		dst = (struct sockaddr_in *)&ro->ro_dst;
391 	else
392 		dst = &sin;
393 	if (ro == NULL || ro->ro_nh == NULL) {
394 		bzero(dst, sizeof(*dst));
395 		dst->sin_family = AF_INET;
396 		dst->sin_len = sizeof(*dst);
397 		dst->sin_addr = ip->ip_dst;
398 	}
399 	gw = dst;
400 again:
401 	/*
402 	 * Validate route against routing table additions;
403 	 * a better/more specific route might have been added.
404 	 */
405 	if (inp != NULL && ro != NULL && ro->ro_nh != NULL)
406 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
407 	/*
408 	 * If there is a cached route,
409 	 * check that it is to the same destination
410 	 * and is still up.  If not, free it and try again.
411 	 * The address family should also be checked in case of sharing the
412 	 * cache with IPv6.
413 	 * Also check whether routing cache needs invalidation.
414 	 */
415 	if (ro != NULL && ro->ro_nh != NULL &&
416 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
417 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
418 		RO_INVALIDATE_CACHE(ro);
419 	ia = NULL;
420 	/*
421 	 * If routing to interface only, short circuit routing lookup.
422 	 * The use of an all-ones broadcast address implies this; an
423 	 * interface is specified by the broadcast address of an interface,
424 	 * or the destination address of a ptp interface.
425 	 */
426 	if (flags & IP_SENDONES) {
427 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
428 						      M_GETFIB(m)))) == NULL &&
429 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
430 						    M_GETFIB(m)))) == NULL) {
431 			IPSTAT_INC(ips_noroute);
432 			error = ENETUNREACH;
433 			goto bad;
434 		}
435 		ip->ip_dst.s_addr = INADDR_BROADCAST;
436 		dst->sin_addr = ip->ip_dst;
437 		ifp = ia->ia_ifp;
438 		mtu = ifp->if_mtu;
439 		ip->ip_ttl = 1;
440 		isbroadcast = 1;
441 		src = IA_SIN(ia)->sin_addr;
442 	} else if (flags & IP_ROUTETOIF) {
443 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
444 						    M_GETFIB(m)))) == NULL &&
445 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
446 						M_GETFIB(m)))) == NULL) {
447 			IPSTAT_INC(ips_noroute);
448 			error = ENETUNREACH;
449 			goto bad;
450 		}
451 		ifp = ia->ia_ifp;
452 		mtu = ifp->if_mtu;
453 		ip->ip_ttl = 1;
454 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
455 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
456 		src = IA_SIN(ia)->sin_addr;
457 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
458 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
459 		/*
460 		 * Bypass the normal routing lookup for multicast
461 		 * packets if the interface is specified.
462 		 */
463 		ifp = imo->imo_multicast_ifp;
464 		mtu = ifp->if_mtu;
465 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
466 		isbroadcast = 0;	/* fool gcc */
467 		/* Interface may have no addresses. */
468 		if (ia != NULL)
469 			src = IA_SIN(ia)->sin_addr;
470 		else
471 			src.s_addr = INADDR_ANY;
472 	} else if (ro != NULL) {
473 		if (ro->ro_nh == NULL) {
474 			/*
475 			 * We want to do any cloning requested by the link
476 			 * layer, as this is probably required in all cases
477 			 * for correct operation (as it is for ARP).
478 			 */
479 			uint32_t flowid;
480 			flowid = m->m_pkthdr.flowid;
481 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
482 			    NHR_REF, flowid);
483 
484 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
485 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
486 				/*
487 				 * There is no route for this packet, but it is
488 				 * possible that a matching SPD entry exists.
489 				 */
490 				no_route_but_check_spd = 1;
491 				goto sendit;
492 #endif
493 				IPSTAT_INC(ips_noroute);
494 				error = EHOSTUNREACH;
495 				goto bad;
496 			}
497 		}
498 		struct nhop_object *nh = ro->ro_nh;
499 
500 		ia = ifatoia(nh->nh_ifa);
501 		ifp = nh->nh_ifp;
502 		counter_u64_add(nh->nh_pksent, 1);
503 		rt_update_ro_flags(ro, nh);
504 		if (nh->nh_flags & NHF_GATEWAY)
505 			gw = &nh->gw4_sa;
506 		if (nh->nh_flags & NHF_HOST)
507 			isbroadcast = (nh->nh_flags & NHF_BROADCAST);
508 		else if (ifp->if_flags & IFF_BROADCAST)
509 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
510 		else
511 			isbroadcast = 0;
512 		mtu = nh->nh_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,
518 		    m->m_pkthdr.flowid);
519 		if (nh == NULL) {
520 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
521 			/*
522 			 * There is no route for this packet, but it is
523 			 * possible that a matching SPD entry exists.
524 			 */
525 			no_route_but_check_spd = 1;
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 			MROUTER_RLOCK();
618 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
619 				/*
620 				 * If rsvp daemon is not running, do not
621 				 * set ip_moptions. This ensures that the packet
622 				 * is multicast and not just sent down one link
623 				 * as prescribed by rsvpd.
624 				 */
625 				if (!V_rsvp_on)
626 					imo = NULL;
627 				if (ip_mforward &&
628 				    ip_mforward(ip, ifp, m, imo) != 0) {
629 					MROUTER_RUNLOCK();
630 					m_freem(m);
631 					goto done;
632 				}
633 			}
634 			MROUTER_RUNLOCK();
635 		}
636 
637 		/*
638 		 * Multicasts with a time-to-live of zero may be looped-
639 		 * back, above, but must not be transmitted on a network.
640 		 * Also, multicasts addressed to the loopback interface
641 		 * are not sent -- the above call to ip_mloopback() will
642 		 * loop back a copy. ip_input() will drop the copy if
643 		 * this host does not belong to the destination group on
644 		 * the loopback interface.
645 		 */
646 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
647 			m_freem(m);
648 			goto done;
649 		}
650 
651 		goto sendit;
652 	}
653 
654 	/*
655 	 * If the source address is not specified yet, use the address
656 	 * of the outoing interface.
657 	 */
658 	if (ip->ip_src.s_addr == INADDR_ANY)
659 		ip->ip_src = src;
660 
661 	/*
662 	 * Look for broadcast address and
663 	 * verify user is allowed to send
664 	 * such a packet.
665 	 */
666 	if (isbroadcast) {
667 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
668 			error = EADDRNOTAVAIL;
669 			goto bad;
670 		}
671 		if ((flags & IP_ALLOWBROADCAST) == 0) {
672 			error = EACCES;
673 			goto bad;
674 		}
675 		/* don't allow broadcast messages to be fragmented */
676 		if (ip_len > mtu) {
677 			error = EMSGSIZE;
678 			goto bad;
679 		}
680 		m->m_flags |= M_BCAST;
681 	} else {
682 		m->m_flags &= ~M_BCAST;
683 	}
684 
685 sendit:
686 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
687 	if (IPSEC_ENABLED(ipv4)) {
688 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
689 			if (error == EINPROGRESS)
690 				error = 0;
691 			goto done;
692 		}
693 	}
694 	/*
695 	 * Check if there was a route for this packet; return error if not.
696 	 */
697 	if (no_route_but_check_spd) {
698 		IPSTAT_INC(ips_noroute);
699 		error = EHOSTUNREACH;
700 		goto bad;
701 	}
702 	/* Update variables that are affected by ipsec4_output(). */
703 	ip = mtod(m, struct ip *);
704 	hlen = ip->ip_hl << 2;
705 #endif /* IPSEC */
706 
707 	/* Jump over all PFIL processing if hooks are not active. */
708 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
709 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
710 		    &error)) {
711 		case 1: /* Finished */
712 			goto done;
713 
714 		case 0: /* Continue normally */
715 			ip = mtod(m, struct ip *);
716 			break;
717 
718 		case -1: /* Need to try again */
719 			/* Reset everything for a new round */
720 			if (ro != NULL) {
721 				RO_NHFREE(ro);
722 				ro->ro_prepend = NULL;
723 			}
724 			gw = dst;
725 			ip = mtod(m, struct ip *);
726 			goto again;
727 		}
728 	}
729 
730 	if (vlan_pcp > -1)
731 		EVL_APPLY_PRI(m, vlan_pcp);
732 
733 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
734 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
735 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
736 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
737 			IPSTAT_INC(ips_badaddr);
738 			error = EADDRNOTAVAIL;
739 			goto bad;
740 		}
741 	}
742 
743 	m->m_pkthdr.csum_flags |= CSUM_IP;
744 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
745 		m = mb_unmapped_to_ext(m);
746 		if (m == NULL) {
747 			IPSTAT_INC(ips_odropped);
748 			error = ENOBUFS;
749 			goto bad;
750 		}
751 		in_delayed_cksum(m);
752 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
753 	} else if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
754 		m = mb_unmapped_to_ext(m);
755 		if (m == NULL) {
756 			IPSTAT_INC(ips_odropped);
757 			error = ENOBUFS;
758 			goto bad;
759 		}
760 	}
761 #if defined(SCTP) || defined(SCTP_SUPPORT)
762 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
763 		m = mb_unmapped_to_ext(m);
764 		if (m == NULL) {
765 			IPSTAT_INC(ips_odropped);
766 			error = ENOBUFS;
767 			goto bad;
768 		}
769 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
770 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
771 	}
772 #endif
773 
774 	/*
775 	 * If small enough for interface, or the interface will take
776 	 * care of the fragmentation for us, we can just send directly.
777 	 * Note that if_vxlan could have requested TSO even though the outer
778 	 * frame is UDP.  It is correct to not fragment such datagrams and
779 	 * instead just pass them on to the driver.
780 	 */
781 	if (ip_len <= mtu ||
782 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
783 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
784 		ip->ip_sum = 0;
785 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
786 			ip->ip_sum = in_cksum(m, hlen);
787 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
788 		}
789 
790 		/*
791 		 * Record statistics for this interface address.
792 		 * With CSUM_TSO the byte/packet count will be slightly
793 		 * incorrect because we count the IP+TCP headers only
794 		 * once instead of for every generated packet.
795 		 */
796 		if (!(flags & IP_FORWARDING) && ia) {
797 			if (m->m_pkthdr.csum_flags &
798 			    (CSUM_TSO | CSUM_INNER_TSO))
799 				counter_u64_add(ia->ia_ifa.ifa_opackets,
800 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
801 			else
802 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
803 
804 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
805 		}
806 #ifdef MBUF_STRESS_TEST
807 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
808 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
809 #endif
810 		/*
811 		 * Reset layer specific mbuf flags
812 		 * to avoid confusing lower layers.
813 		 */
814 		m_clrprotoflags(m);
815 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
816 		error = ip_output_send(inp, ifp, m, gw, ro,
817 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
818 		goto done;
819 	}
820 
821 	/* Balk when DF bit is set or the interface didn't support TSO. */
822 	if ((ip_off & IP_DF) ||
823 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
824 		error = EMSGSIZE;
825 		IPSTAT_INC(ips_cantfrag);
826 		goto bad;
827 	}
828 
829 	/*
830 	 * Too large for interface; fragment if possible. If successful,
831 	 * on return, m will point to a list of packets to be sent.
832 	 */
833 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
834 	if (error)
835 		goto bad;
836 	for (; m; m = m0) {
837 		m0 = m->m_nextpkt;
838 		m->m_nextpkt = 0;
839 		if (error == 0) {
840 			/* Record statistics for this interface address. */
841 			if (ia != NULL) {
842 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
843 				counter_u64_add(ia->ia_ifa.ifa_obytes,
844 				    m->m_pkthdr.len);
845 			}
846 			/*
847 			 * Reset layer specific mbuf flags
848 			 * to avoid confusing upper layers.
849 			 */
850 			m_clrprotoflags(m);
851 
852 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
853 			    mtod(m, struct ip *), NULL);
854 			error = ip_output_send(inp, ifp, m, gw, ro, true);
855 		} else
856 			m_freem(m);
857 	}
858 
859 	if (error == 0)
860 		IPSTAT_INC(ips_fragmented);
861 
862 done:
863 	return (error);
864  bad:
865 	m_freem(m);
866 	goto done;
867 }
868 
869 /*
870  * Create a chain of fragments which fit the given mtu. m_frag points to the
871  * mbuf to be fragmented; on return it points to the chain with the fragments.
872  * Return 0 if no error. If error, m_frag may contain a partially built
873  * chain of fragments that should be freed by the caller.
874  *
875  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
876  */
877 int
878 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
879     u_long if_hwassist_flags)
880 {
881 	int error = 0;
882 	int hlen = ip->ip_hl << 2;
883 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
884 	int off;
885 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
886 	int firstlen;
887 	struct mbuf **mnext;
888 	int nfrags;
889 	uint16_t ip_len, ip_off;
890 
891 	ip_len = ntohs(ip->ip_len);
892 	ip_off = ntohs(ip->ip_off);
893 
894 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
895 		IPSTAT_INC(ips_cantfrag);
896 		return EMSGSIZE;
897 	}
898 
899 	/*
900 	 * Must be able to put at least 8 bytes per fragment.
901 	 */
902 	if (len < 8)
903 		return EMSGSIZE;
904 
905 	/*
906 	 * If the interface will not calculate checksums on
907 	 * fragmented packets, then do it here.
908 	 */
909 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
910 		m0 = mb_unmapped_to_ext(m0);
911 		if (m0 == NULL) {
912 			error = ENOBUFS;
913 			IPSTAT_INC(ips_odropped);
914 			goto done;
915 		}
916 		in_delayed_cksum(m0);
917 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
918 	}
919 #if defined(SCTP) || defined(SCTP_SUPPORT)
920 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
921 		m0 = mb_unmapped_to_ext(m0);
922 		if (m0 == NULL) {
923 			error = ENOBUFS;
924 			IPSTAT_INC(ips_odropped);
925 			goto done;
926 		}
927 		sctp_delayed_cksum(m0, hlen);
928 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
929 	}
930 #endif
931 	if (len > PAGE_SIZE) {
932 		/*
933 		 * Fragment large datagrams such that each segment
934 		 * contains a multiple of PAGE_SIZE amount of data,
935 		 * plus headers. This enables a receiver to perform
936 		 * page-flipping zero-copy optimizations.
937 		 *
938 		 * XXX When does this help given that sender and receiver
939 		 * could have different page sizes, and also mtu could
940 		 * be less than the receiver's page size ?
941 		 */
942 		int newlen;
943 
944 		off = MIN(mtu, m0->m_pkthdr.len);
945 
946 		/*
947 		 * firstlen (off - hlen) must be aligned on an
948 		 * 8-byte boundary
949 		 */
950 		if (off < hlen)
951 			goto smart_frag_failure;
952 		off = ((off - hlen) & ~7) + hlen;
953 		newlen = (~PAGE_MASK) & mtu;
954 		if ((newlen + sizeof (struct ip)) > mtu) {
955 			/* we failed, go back the default */
956 smart_frag_failure:
957 			newlen = len;
958 			off = hlen + len;
959 		}
960 		len = newlen;
961 
962 	} else {
963 		off = hlen + len;
964 	}
965 
966 	firstlen = off - hlen;
967 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
968 
969 	/*
970 	 * Loop through length of segment after first fragment,
971 	 * make new header and copy data of each part and link onto chain.
972 	 * Here, m0 is the original packet, m is the fragment being created.
973 	 * The fragments are linked off the m_nextpkt of the original
974 	 * packet, which after processing serves as the first fragment.
975 	 */
976 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
977 		struct ip *mhip;	/* ip header on the fragment */
978 		struct mbuf *m;
979 		int mhlen = sizeof (struct ip);
980 
981 		m = m_gethdr(M_NOWAIT, MT_DATA);
982 		if (m == NULL) {
983 			error = ENOBUFS;
984 			IPSTAT_INC(ips_odropped);
985 			goto done;
986 		}
987 		/*
988 		 * Make sure the complete packet header gets copied
989 		 * from the originating mbuf to the newly created
990 		 * mbuf. This also ensures that existing firewall
991 		 * classification(s), VLAN tags and so on get copied
992 		 * to the resulting fragmented packet(s):
993 		 */
994 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
995 			m_free(m);
996 			error = ENOBUFS;
997 			IPSTAT_INC(ips_odropped);
998 			goto done;
999 		}
1000 		/*
1001 		 * In the first mbuf, leave room for the link header, then
1002 		 * copy the original IP header including options. The payload
1003 		 * goes into an additional mbuf chain returned by m_copym().
1004 		 */
1005 		m->m_data += max_linkhdr;
1006 		mhip = mtod(m, struct ip *);
1007 		*mhip = *ip;
1008 		if (hlen > sizeof (struct ip)) {
1009 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1010 			mhip->ip_v = IPVERSION;
1011 			mhip->ip_hl = mhlen >> 2;
1012 		}
1013 		m->m_len = mhlen;
1014 		/* XXX do we need to add ip_off below ? */
1015 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1016 		if (off + len >= ip_len)
1017 			len = ip_len - off;
1018 		else
1019 			mhip->ip_off |= IP_MF;
1020 		mhip->ip_len = htons((u_short)(len + mhlen));
1021 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1022 		if (m->m_next == NULL) {	/* copy failed */
1023 			m_free(m);
1024 			error = ENOBUFS;	/* ??? */
1025 			IPSTAT_INC(ips_odropped);
1026 			goto done;
1027 		}
1028 		m->m_pkthdr.len = mhlen + len;
1029 #ifdef MAC
1030 		mac_netinet_fragment(m0, m);
1031 #endif
1032 		mhip->ip_off = htons(mhip->ip_off);
1033 		mhip->ip_sum = 0;
1034 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1035 			mhip->ip_sum = in_cksum(m, mhlen);
1036 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1037 		}
1038 		*mnext = m;
1039 		mnext = &m->m_nextpkt;
1040 	}
1041 	IPSTAT_ADD(ips_ofragments, nfrags);
1042 
1043 	/*
1044 	 * Update first fragment by trimming what's been copied out
1045 	 * and updating header.
1046 	 */
1047 	m_adj(m0, hlen + firstlen - ip_len);
1048 	m0->m_pkthdr.len = hlen + firstlen;
1049 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1050 	ip->ip_off = htons(ip_off | IP_MF);
1051 	ip->ip_sum = 0;
1052 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1053 		ip->ip_sum = in_cksum(m0, hlen);
1054 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1055 	}
1056 
1057 done:
1058 	*m_frag = m0;
1059 	return error;
1060 }
1061 
1062 void
1063 in_delayed_cksum(struct mbuf *m)
1064 {
1065 	struct ip *ip;
1066 	struct udphdr *uh;
1067 	uint16_t cklen, csum, offset;
1068 
1069 	ip = mtod(m, struct ip *);
1070 	offset = ip->ip_hl << 2 ;
1071 
1072 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1073 		/* if udp header is not in the first mbuf copy udplen */
1074 		if (offset + sizeof(struct udphdr) > m->m_len) {
1075 			m_copydata(m, offset + offsetof(struct udphdr,
1076 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1077 			cklen = ntohs(cklen);
1078 		} else {
1079 			uh = (struct udphdr *)mtodo(m, offset);
1080 			cklen = ntohs(uh->uh_ulen);
1081 		}
1082 		csum = in_cksum_skip(m, cklen + offset, offset);
1083 		if (csum == 0)
1084 			csum = 0xffff;
1085 	} else {
1086 		cklen = ntohs(ip->ip_len);
1087 		csum = in_cksum_skip(m, cklen, offset);
1088 	}
1089 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1090 
1091 	if (offset + sizeof(csum) > m->m_len)
1092 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1093 	else
1094 		*(u_short *)mtodo(m, offset) = csum;
1095 }
1096 
1097 /*
1098  * IP socket option processing.
1099  */
1100 int
1101 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1102 {
1103 	struct inpcb *inp = sotoinpcb(so);
1104 	int	error, optval;
1105 #ifdef	RSS
1106 	uint32_t rss_bucket;
1107 	int retval;
1108 #endif
1109 
1110 	error = optval = 0;
1111 	if (sopt->sopt_level != IPPROTO_IP) {
1112 		error = EINVAL;
1113 
1114 		if (sopt->sopt_level == SOL_SOCKET &&
1115 		    sopt->sopt_dir == SOPT_SET) {
1116 			switch (sopt->sopt_name) {
1117 			case SO_REUSEADDR:
1118 				INP_WLOCK(inp);
1119 				if ((so->so_options & SO_REUSEADDR) != 0)
1120 					inp->inp_flags2 |= INP_REUSEADDR;
1121 				else
1122 					inp->inp_flags2 &= ~INP_REUSEADDR;
1123 				INP_WUNLOCK(inp);
1124 				error = 0;
1125 				break;
1126 			case SO_REUSEPORT:
1127 				INP_WLOCK(inp);
1128 				if ((so->so_options & SO_REUSEPORT) != 0)
1129 					inp->inp_flags2 |= INP_REUSEPORT;
1130 				else
1131 					inp->inp_flags2 &= ~INP_REUSEPORT;
1132 				INP_WUNLOCK(inp);
1133 				error = 0;
1134 				break;
1135 			case SO_REUSEPORT_LB:
1136 				INP_WLOCK(inp);
1137 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1138 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1139 				else
1140 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1141 				INP_WUNLOCK(inp);
1142 				error = 0;
1143 				break;
1144 			case SO_SETFIB:
1145 				INP_WLOCK(inp);
1146 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1147 				INP_WUNLOCK(inp);
1148 				error = 0;
1149 				break;
1150 			case SO_MAX_PACING_RATE:
1151 #ifdef RATELIMIT
1152 				INP_WLOCK(inp);
1153 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1154 				INP_WUNLOCK(inp);
1155 				error = 0;
1156 #else
1157 				error = EOPNOTSUPP;
1158 #endif
1159 				break;
1160 			default:
1161 				break;
1162 			}
1163 		}
1164 		return (error);
1165 	}
1166 
1167 	switch (sopt->sopt_dir) {
1168 	case SOPT_SET:
1169 		switch (sopt->sopt_name) {
1170 		case IP_OPTIONS:
1171 #ifdef notyet
1172 		case IP_RETOPTS:
1173 #endif
1174 		{
1175 			struct mbuf *m;
1176 			if (sopt->sopt_valsize > MLEN) {
1177 				error = EMSGSIZE;
1178 				break;
1179 			}
1180 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1181 			if (m == NULL) {
1182 				error = ENOBUFS;
1183 				break;
1184 			}
1185 			m->m_len = sopt->sopt_valsize;
1186 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1187 					    m->m_len);
1188 			if (error) {
1189 				m_free(m);
1190 				break;
1191 			}
1192 			INP_WLOCK(inp);
1193 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1194 			INP_WUNLOCK(inp);
1195 			return (error);
1196 		}
1197 
1198 		case IP_BINDANY:
1199 			if (sopt->sopt_td != NULL) {
1200 				error = priv_check(sopt->sopt_td,
1201 				    PRIV_NETINET_BINDANY);
1202 				if (error)
1203 					break;
1204 			}
1205 			/* FALLTHROUGH */
1206 		case IP_BINDMULTI:
1207 #ifdef	RSS
1208 		case IP_RSS_LISTEN_BUCKET:
1209 #endif
1210 		case IP_TOS:
1211 		case IP_TTL:
1212 		case IP_MINTTL:
1213 		case IP_RECVOPTS:
1214 		case IP_RECVRETOPTS:
1215 		case IP_ORIGDSTADDR:
1216 		case IP_RECVDSTADDR:
1217 		case IP_RECVTTL:
1218 		case IP_RECVIF:
1219 		case IP_ONESBCAST:
1220 		case IP_DONTFRAG:
1221 		case IP_RECVTOS:
1222 		case IP_RECVFLOWID:
1223 #ifdef	RSS
1224 		case IP_RECVRSSBUCKETID:
1225 #endif
1226 		case IP_VLAN_PCP:
1227 			error = sooptcopyin(sopt, &optval, sizeof optval,
1228 					    sizeof optval);
1229 			if (error)
1230 				break;
1231 
1232 			switch (sopt->sopt_name) {
1233 			case IP_TOS:
1234 				inp->inp_ip_tos = optval;
1235 				break;
1236 
1237 			case IP_TTL:
1238 				inp->inp_ip_ttl = optval;
1239 				break;
1240 
1241 			case IP_MINTTL:
1242 				if (optval >= 0 && optval <= MAXTTL)
1243 					inp->inp_ip_minttl = optval;
1244 				else
1245 					error = EINVAL;
1246 				break;
1247 
1248 #define	OPTSET(bit) do {						\
1249 	INP_WLOCK(inp);							\
1250 	if (optval)							\
1251 		inp->inp_flags |= bit;					\
1252 	else								\
1253 		inp->inp_flags &= ~bit;					\
1254 	INP_WUNLOCK(inp);						\
1255 } while (0)
1256 
1257 #define	OPTSET2(bit, val) do {						\
1258 	INP_WLOCK(inp);							\
1259 	if (val)							\
1260 		inp->inp_flags2 |= bit;					\
1261 	else								\
1262 		inp->inp_flags2 &= ~bit;				\
1263 	INP_WUNLOCK(inp);						\
1264 } while (0)
1265 
1266 			case IP_RECVOPTS:
1267 				OPTSET(INP_RECVOPTS);
1268 				break;
1269 
1270 			case IP_RECVRETOPTS:
1271 				OPTSET(INP_RECVRETOPTS);
1272 				break;
1273 
1274 			case IP_RECVDSTADDR:
1275 				OPTSET(INP_RECVDSTADDR);
1276 				break;
1277 
1278 			case IP_ORIGDSTADDR:
1279 				OPTSET2(INP_ORIGDSTADDR, optval);
1280 				break;
1281 
1282 			case IP_RECVTTL:
1283 				OPTSET(INP_RECVTTL);
1284 				break;
1285 
1286 			case IP_RECVIF:
1287 				OPTSET(INP_RECVIF);
1288 				break;
1289 
1290 			case IP_ONESBCAST:
1291 				OPTSET(INP_ONESBCAST);
1292 				break;
1293 			case IP_DONTFRAG:
1294 				OPTSET(INP_DONTFRAG);
1295 				break;
1296 			case IP_BINDANY:
1297 				OPTSET(INP_BINDANY);
1298 				break;
1299 			case IP_RECVTOS:
1300 				OPTSET(INP_RECVTOS);
1301 				break;
1302 			case IP_BINDMULTI:
1303 				OPTSET2(INP_BINDMULTI, optval);
1304 				break;
1305 			case IP_RECVFLOWID:
1306 				OPTSET2(INP_RECVFLOWID, optval);
1307 				break;
1308 #ifdef	RSS
1309 			case IP_RSS_LISTEN_BUCKET:
1310 				if ((optval >= 0) &&
1311 				    (optval < rss_getnumbuckets())) {
1312 					inp->inp_rss_listen_bucket = optval;
1313 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1314 				} else {
1315 					error = EINVAL;
1316 				}
1317 				break;
1318 			case IP_RECVRSSBUCKETID:
1319 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1320 				break;
1321 #endif
1322 			case IP_VLAN_PCP:
1323 				if ((optval >= -1) && (optval <=
1324 				    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1325 					if (optval == -1) {
1326 						INP_WLOCK(inp);
1327 						inp->inp_flags2 &=
1328 						    ~(INP_2PCP_SET |
1329 						      INP_2PCP_MASK);
1330 						INP_WUNLOCK(inp);
1331 					} else {
1332 						INP_WLOCK(inp);
1333 						inp->inp_flags2 |=
1334 						    INP_2PCP_SET;
1335 						inp->inp_flags2 &=
1336 						    ~INP_2PCP_MASK;
1337 						inp->inp_flags2 |=
1338 						    optval << INP_2PCP_SHIFT;
1339 						INP_WUNLOCK(inp);
1340 					}
1341 				} else
1342 					error = EINVAL;
1343 				break;
1344 			}
1345 			break;
1346 #undef OPTSET
1347 #undef OPTSET2
1348 
1349 		/*
1350 		 * Multicast socket options are processed by the in_mcast
1351 		 * module.
1352 		 */
1353 		case IP_MULTICAST_IF:
1354 		case IP_MULTICAST_VIF:
1355 		case IP_MULTICAST_TTL:
1356 		case IP_MULTICAST_LOOP:
1357 		case IP_ADD_MEMBERSHIP:
1358 		case IP_DROP_MEMBERSHIP:
1359 		case IP_ADD_SOURCE_MEMBERSHIP:
1360 		case IP_DROP_SOURCE_MEMBERSHIP:
1361 		case IP_BLOCK_SOURCE:
1362 		case IP_UNBLOCK_SOURCE:
1363 		case IP_MSFILTER:
1364 		case MCAST_JOIN_GROUP:
1365 		case MCAST_LEAVE_GROUP:
1366 		case MCAST_JOIN_SOURCE_GROUP:
1367 		case MCAST_LEAVE_SOURCE_GROUP:
1368 		case MCAST_BLOCK_SOURCE:
1369 		case MCAST_UNBLOCK_SOURCE:
1370 			error = inp_setmoptions(inp, sopt);
1371 			break;
1372 
1373 		case IP_PORTRANGE:
1374 			error = sooptcopyin(sopt, &optval, sizeof optval,
1375 					    sizeof optval);
1376 			if (error)
1377 				break;
1378 
1379 			INP_WLOCK(inp);
1380 			switch (optval) {
1381 			case IP_PORTRANGE_DEFAULT:
1382 				inp->inp_flags &= ~(INP_LOWPORT);
1383 				inp->inp_flags &= ~(INP_HIGHPORT);
1384 				break;
1385 
1386 			case IP_PORTRANGE_HIGH:
1387 				inp->inp_flags &= ~(INP_LOWPORT);
1388 				inp->inp_flags |= INP_HIGHPORT;
1389 				break;
1390 
1391 			case IP_PORTRANGE_LOW:
1392 				inp->inp_flags &= ~(INP_HIGHPORT);
1393 				inp->inp_flags |= INP_LOWPORT;
1394 				break;
1395 
1396 			default:
1397 				error = EINVAL;
1398 				break;
1399 			}
1400 			INP_WUNLOCK(inp);
1401 			break;
1402 
1403 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1404 		case IP_IPSEC_POLICY:
1405 			if (IPSEC_ENABLED(ipv4)) {
1406 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1407 				break;
1408 			}
1409 			/* FALLTHROUGH */
1410 #endif /* IPSEC */
1411 
1412 		default:
1413 			error = ENOPROTOOPT;
1414 			break;
1415 		}
1416 		break;
1417 
1418 	case SOPT_GET:
1419 		switch (sopt->sopt_name) {
1420 		case IP_OPTIONS:
1421 		case IP_RETOPTS:
1422 			INP_RLOCK(inp);
1423 			if (inp->inp_options) {
1424 				struct mbuf *options;
1425 
1426 				options = m_copym(inp->inp_options, 0,
1427 				    M_COPYALL, M_NOWAIT);
1428 				INP_RUNLOCK(inp);
1429 				if (options != NULL) {
1430 					error = sooptcopyout(sopt,
1431 							     mtod(options, char *),
1432 							     options->m_len);
1433 					m_freem(options);
1434 				} else
1435 					error = ENOMEM;
1436 			} else {
1437 				INP_RUNLOCK(inp);
1438 				sopt->sopt_valsize = 0;
1439 			}
1440 			break;
1441 
1442 		case IP_TOS:
1443 		case IP_TTL:
1444 		case IP_MINTTL:
1445 		case IP_RECVOPTS:
1446 		case IP_RECVRETOPTS:
1447 		case IP_ORIGDSTADDR:
1448 		case IP_RECVDSTADDR:
1449 		case IP_RECVTTL:
1450 		case IP_RECVIF:
1451 		case IP_PORTRANGE:
1452 		case IP_ONESBCAST:
1453 		case IP_DONTFRAG:
1454 		case IP_BINDANY:
1455 		case IP_RECVTOS:
1456 		case IP_BINDMULTI:
1457 		case IP_FLOWID:
1458 		case IP_FLOWTYPE:
1459 		case IP_RECVFLOWID:
1460 #ifdef	RSS
1461 		case IP_RSSBUCKETID:
1462 		case IP_RECVRSSBUCKETID:
1463 #endif
1464 		case IP_VLAN_PCP:
1465 			switch (sopt->sopt_name) {
1466 			case IP_TOS:
1467 				optval = inp->inp_ip_tos;
1468 				break;
1469 
1470 			case IP_TTL:
1471 				optval = inp->inp_ip_ttl;
1472 				break;
1473 
1474 			case IP_MINTTL:
1475 				optval = inp->inp_ip_minttl;
1476 				break;
1477 
1478 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1479 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1480 
1481 			case IP_RECVOPTS:
1482 				optval = OPTBIT(INP_RECVOPTS);
1483 				break;
1484 
1485 			case IP_RECVRETOPTS:
1486 				optval = OPTBIT(INP_RECVRETOPTS);
1487 				break;
1488 
1489 			case IP_RECVDSTADDR:
1490 				optval = OPTBIT(INP_RECVDSTADDR);
1491 				break;
1492 
1493 			case IP_ORIGDSTADDR:
1494 				optval = OPTBIT2(INP_ORIGDSTADDR);
1495 				break;
1496 
1497 			case IP_RECVTTL:
1498 				optval = OPTBIT(INP_RECVTTL);
1499 				break;
1500 
1501 			case IP_RECVIF:
1502 				optval = OPTBIT(INP_RECVIF);
1503 				break;
1504 
1505 			case IP_PORTRANGE:
1506 				if (inp->inp_flags & INP_HIGHPORT)
1507 					optval = IP_PORTRANGE_HIGH;
1508 				else if (inp->inp_flags & INP_LOWPORT)
1509 					optval = IP_PORTRANGE_LOW;
1510 				else
1511 					optval = 0;
1512 				break;
1513 
1514 			case IP_ONESBCAST:
1515 				optval = OPTBIT(INP_ONESBCAST);
1516 				break;
1517 			case IP_DONTFRAG:
1518 				optval = OPTBIT(INP_DONTFRAG);
1519 				break;
1520 			case IP_BINDANY:
1521 				optval = OPTBIT(INP_BINDANY);
1522 				break;
1523 			case IP_RECVTOS:
1524 				optval = OPTBIT(INP_RECVTOS);
1525 				break;
1526 			case IP_FLOWID:
1527 				optval = inp->inp_flowid;
1528 				break;
1529 			case IP_FLOWTYPE:
1530 				optval = inp->inp_flowtype;
1531 				break;
1532 			case IP_RECVFLOWID:
1533 				optval = OPTBIT2(INP_RECVFLOWID);
1534 				break;
1535 #ifdef	RSS
1536 			case IP_RSSBUCKETID:
1537 				retval = rss_hash2bucket(inp->inp_flowid,
1538 				    inp->inp_flowtype,
1539 				    &rss_bucket);
1540 				if (retval == 0)
1541 					optval = rss_bucket;
1542 				else
1543 					error = EINVAL;
1544 				break;
1545 			case IP_RECVRSSBUCKETID:
1546 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1547 				break;
1548 #endif
1549 			case IP_BINDMULTI:
1550 				optval = OPTBIT2(INP_BINDMULTI);
1551 				break;
1552 			case IP_VLAN_PCP:
1553 				if (OPTBIT2(INP_2PCP_SET)) {
1554 					optval = (inp->inp_flags2 &
1555 					    INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1556 				} else {
1557 					optval = -1;
1558 				}
1559 				break;
1560 			}
1561 			error = sooptcopyout(sopt, &optval, sizeof optval);
1562 			break;
1563 
1564 		/*
1565 		 * Multicast socket options are processed by the in_mcast
1566 		 * module.
1567 		 */
1568 		case IP_MULTICAST_IF:
1569 		case IP_MULTICAST_VIF:
1570 		case IP_MULTICAST_TTL:
1571 		case IP_MULTICAST_LOOP:
1572 		case IP_MSFILTER:
1573 			error = inp_getmoptions(inp, sopt);
1574 			break;
1575 
1576 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1577 		case IP_IPSEC_POLICY:
1578 			if (IPSEC_ENABLED(ipv4)) {
1579 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1580 				break;
1581 			}
1582 			/* FALLTHROUGH */
1583 #endif /* IPSEC */
1584 
1585 		default:
1586 			error = ENOPROTOOPT;
1587 			break;
1588 		}
1589 		break;
1590 	}
1591 	return (error);
1592 }
1593 
1594 /*
1595  * Routine called from ip_output() to loop back a copy of an IP multicast
1596  * packet to the input queue of a specified interface.  Note that this
1597  * calls the output routine of the loopback "driver", but with an interface
1598  * pointer that might NOT be a loopback interface -- evil, but easier than
1599  * replicating that code here.
1600  */
1601 static void
1602 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1603 {
1604 	struct ip *ip;
1605 	struct mbuf *copym;
1606 
1607 	/*
1608 	 * Make a deep copy of the packet because we're going to
1609 	 * modify the pack in order to generate checksums.
1610 	 */
1611 	copym = m_dup(m, M_NOWAIT);
1612 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1613 		copym = m_pullup(copym, hlen);
1614 	if (copym != NULL) {
1615 		/* If needed, compute the checksum and mark it as valid. */
1616 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1617 			in_delayed_cksum(copym);
1618 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1619 			copym->m_pkthdr.csum_flags |=
1620 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1621 			copym->m_pkthdr.csum_data = 0xffff;
1622 		}
1623 		/*
1624 		 * We don't bother to fragment if the IP length is greater
1625 		 * than the interface's MTU.  Can this possibly matter?
1626 		 */
1627 		ip = mtod(copym, struct ip *);
1628 		ip->ip_sum = 0;
1629 		ip->ip_sum = in_cksum(copym, hlen);
1630 		if_simloop(ifp, copym, AF_INET, 0);
1631 	}
1632 }
1633