xref: /freebsd/sys/netinet/ip_output.c (revision 100353cfbf882e23c911300ebd0cb458bd3ee975)
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_llatbl.h>
66 #include <net/netisr.h>
67 #include <net/pfil.h>
68 #include <net/route.h>
69 #include <net/route/nhop.h>
70 #include <net/rss_config.h>
71 #include <net/vnet.h>
72 
73 #include <netinet/in.h>
74 #include <netinet/in_fib.h>
75 #include <netinet/in_kdtrace.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/ip.h>
78 #include <netinet/in_fib.h>
79 #include <netinet/in_pcb.h>
80 #include <netinet/in_rss.h>
81 #include <netinet/in_var.h>
82 #include <netinet/ip_var.h>
83 #include <netinet/ip_options.h>
84 
85 #include <netinet/udp.h>
86 #include <netinet/udp_var.h>
87 
88 #if defined(SCTP) || defined(SCTP_SUPPORT)
89 #include <netinet/sctp.h>
90 #include <netinet/sctp_crc32.h>
91 #endif
92 
93 #include <netipsec/ipsec_support.h>
94 
95 #include <machine/in_cksum.h>
96 
97 #include <security/mac/mac_framework.h>
98 
99 #ifdef MBUF_STRESS_TEST
100 static int mbuf_frag_size = 0;
101 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
102 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
103 #endif
104 
105 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
106 
107 extern int in_mcast_loop;
108 extern	struct protosw inetsw[];
109 
110 static inline int
111 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
112     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
113 {
114 	struct m_tag *fwd_tag = NULL;
115 	struct mbuf *m;
116 	struct in_addr odst;
117 	struct ip *ip;
118 	int pflags = PFIL_OUT;
119 
120 	if (flags & IP_FORWARDING)
121 		pflags |= PFIL_FWD;
122 
123 	m = *mp;
124 	ip = mtod(m, struct ip *);
125 
126 	/* Run through list of hooks for output packets. */
127 	odst.s_addr = ip->ip_dst.s_addr;
128 	switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
129 	case PFIL_DROPPED:
130 		*error = EACCES;
131 		/* FALLTHROUGH */
132 	case PFIL_CONSUMED:
133 		return 1; /* Finished */
134 	case PFIL_PASS:
135 		*error = 0;
136 	}
137 	m = *mp;
138 	ip = mtod(m, struct ip *);
139 
140 	/* See if destination IP address was changed by packet filter. */
141 	if (odst.s_addr != ip->ip_dst.s_addr) {
142 		m->m_flags |= M_SKIP_FIREWALL;
143 		/* If destination is now ourself drop to ip_input(). */
144 		if (in_localip(ip->ip_dst)) {
145 			m->m_flags |= M_FASTFWD_OURS;
146 			if (m->m_pkthdr.rcvif == NULL)
147 				m->m_pkthdr.rcvif = V_loif;
148 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
149 				m->m_pkthdr.csum_flags |=
150 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
151 				m->m_pkthdr.csum_data = 0xffff;
152 			}
153 			m->m_pkthdr.csum_flags |=
154 				CSUM_IP_CHECKED | CSUM_IP_VALID;
155 #if defined(SCTP) || defined(SCTP_SUPPORT)
156 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
157 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
158 #endif
159 			*error = netisr_queue(NETISR_IP, m);
160 			return 1; /* Finished */
161 		}
162 
163 		bzero(dst, sizeof(*dst));
164 		dst->sin_family = AF_INET;
165 		dst->sin_len = sizeof(*dst);
166 		dst->sin_addr = ip->ip_dst;
167 
168 		return -1; /* Reloop */
169 	}
170 	/* See if fib was changed by packet filter. */
171 	if ((*fibnum) != M_GETFIB(m)) {
172 		m->m_flags |= M_SKIP_FIREWALL;
173 		*fibnum = M_GETFIB(m);
174 		return -1; /* Reloop for FIB change */
175 	}
176 
177 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
178 	if (m->m_flags & M_FASTFWD_OURS) {
179 		if (m->m_pkthdr.rcvif == NULL)
180 			m->m_pkthdr.rcvif = V_loif;
181 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
182 			m->m_pkthdr.csum_flags |=
183 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
184 			m->m_pkthdr.csum_data = 0xffff;
185 		}
186 #if defined(SCTP) || defined(SCTP_SUPPORT)
187 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
188 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
189 #endif
190 		m->m_pkthdr.csum_flags |=
191 			CSUM_IP_CHECKED | CSUM_IP_VALID;
192 
193 		*error = netisr_queue(NETISR_IP, m);
194 		return 1; /* Finished */
195 	}
196 	/* Or forward to some other address? */
197 	if ((m->m_flags & M_IP_NEXTHOP) &&
198 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
199 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
200 		m->m_flags |= M_SKIP_FIREWALL;
201 		m->m_flags &= ~M_IP_NEXTHOP;
202 		m_tag_delete(m, fwd_tag);
203 
204 		return -1; /* Reloop for CHANGE of dst */
205 	}
206 
207 	return 0;
208 }
209 
210 static int
211 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
212     const struct sockaddr_in *gw, struct route *ro, bool stamp_tag)
213 {
214 #ifdef KERN_TLS
215 	struct ktls_session *tls = NULL;
216 #endif
217 	struct m_snd_tag *mst;
218 	int error;
219 
220 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
221 	mst = NULL;
222 
223 #ifdef KERN_TLS
224 	/*
225 	 * If this is an unencrypted TLS record, save a reference to
226 	 * the record.  This local reference is used to call
227 	 * ktls_output_eagain after the mbuf has been freed (thus
228 	 * dropping the mbuf's reference) in if_output.
229 	 */
230 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
231 		tls = ktls_hold(m->m_next->m_epg_tls);
232 		mst = tls->snd_tag;
233 
234 		/*
235 		 * If a TLS session doesn't have a valid tag, it must
236 		 * have had an earlier ifp mismatch, so drop this
237 		 * packet.
238 		 */
239 		if (mst == NULL) {
240 			error = EAGAIN;
241 			goto done;
242 		}
243 		/*
244 		 * Always stamp tags that include NIC ktls.
245 		 */
246 		stamp_tag = true;
247 	}
248 #endif
249 #ifdef RATELIMIT
250 	if (inp != NULL && mst == NULL) {
251 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
252 		    (inp->inp_snd_tag != NULL &&
253 		    inp->inp_snd_tag->ifp != ifp))
254 			in_pcboutput_txrtlmt(inp, ifp, m);
255 
256 		if (inp->inp_snd_tag != NULL)
257 			mst = inp->inp_snd_tag;
258 	}
259 #endif
260 	if (stamp_tag && mst != NULL) {
261 		KASSERT(m->m_pkthdr.rcvif == NULL,
262 		    ("trying to add a send tag to a forwarded packet"));
263 		if (mst->ifp != ifp) {
264 			error = EAGAIN;
265 			goto done;
266 		}
267 
268 		/* stamp send tag on mbuf */
269 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
270 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
271 	}
272 
273 	error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
274 
275 done:
276 	/* Check for route change invalidating send tags. */
277 #ifdef KERN_TLS
278 	if (tls != NULL) {
279 		if (error == EAGAIN)
280 			error = ktls_output_eagain(inp, tls);
281 		ktls_free(tls);
282 	}
283 #endif
284 #ifdef RATELIMIT
285 	if (error == EAGAIN)
286 		in_pcboutput_eagain(inp);
287 #endif
288 	return (error);
289 }
290 
291 /* rte<>ro_flags translation */
292 static inline void
293 rt_update_ro_flags(struct route *ro)
294 {
295 	int nh_flags = ro->ro_nh->nh_flags;
296 
297 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
298 
299 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
300 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
301 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
302 }
303 
304 /*
305  * IP output.  The packet in mbuf chain m contains a skeletal IP
306  * header (with len, off, ttl, proto, tos, src, dst).
307  * The mbuf chain containing the packet will be freed.
308  * The mbuf opt, if present, will not be freed.
309  * If route ro is present and has ro_rt initialized, route lookup would be
310  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
311  * then result of route lookup is stored in ro->ro_rt.
312  *
313  * In the IP forwarding case, the packet will arrive with options already
314  * inserted, so must have a NULL opt pointer.
315  */
316 int
317 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
318     struct ip_moptions *imo, struct inpcb *inp)
319 {
320 	struct rm_priotracker in_ifa_tracker;
321 	struct ip *ip;
322 	struct ifnet *ifp = NULL;	/* keep compiler happy */
323 	struct mbuf *m0;
324 	int hlen = sizeof (struct ip);
325 	int mtu = 0;
326 	int error = 0;
327 	struct sockaddr_in *dst, sin;
328 	const struct sockaddr_in *gw;
329 	struct in_ifaddr *ia = NULL;
330 	struct in_addr src;
331 	int isbroadcast;
332 	uint16_t ip_len, ip_off;
333 	uint32_t fibnum;
334 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
335 	int no_route_but_check_spd = 0;
336 #endif
337 
338 	M_ASSERTPKTHDR(m);
339 	NET_EPOCH_ASSERT();
340 
341 	if (inp != NULL) {
342 		INP_LOCK_ASSERT(inp);
343 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
344 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
345 			m->m_pkthdr.flowid = inp->inp_flowid;
346 			M_HASHTYPE_SET(m, inp->inp_flowtype);
347 		}
348 #ifdef NUMA
349 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
350 #endif
351 	}
352 
353 	if (opt) {
354 		int len = 0;
355 		m = ip_insertoptions(m, opt, &len);
356 		if (len != 0)
357 			hlen = len; /* ip->ip_hl is updated above */
358 	}
359 	ip = mtod(m, struct ip *);
360 	ip_len = ntohs(ip->ip_len);
361 	ip_off = ntohs(ip->ip_off);
362 
363 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
364 		ip->ip_v = IPVERSION;
365 		ip->ip_hl = hlen >> 2;
366 		ip_fillid(ip);
367 	} else {
368 		/* Header already set, fetch hlen from there */
369 		hlen = ip->ip_hl << 2;
370 	}
371 	if ((flags & IP_FORWARDING) == 0)
372 		IPSTAT_INC(ips_localout);
373 
374 	/*
375 	 * dst/gw handling:
376 	 *
377 	 * gw is readonly but can point either to dst OR rt_gateway,
378 	 * therefore we need restore gw if we're redoing lookup.
379 	 */
380 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
381 	if (ro != NULL)
382 		dst = (struct sockaddr_in *)&ro->ro_dst;
383 	else
384 		dst = &sin;
385 	if (ro == NULL || ro->ro_nh == NULL) {
386 		bzero(dst, sizeof(*dst));
387 		dst->sin_family = AF_INET;
388 		dst->sin_len = sizeof(*dst);
389 		dst->sin_addr = ip->ip_dst;
390 	}
391 	gw = dst;
392 again:
393 	/*
394 	 * Validate route against routing table additions;
395 	 * a better/more specific route might have been added.
396 	 */
397 	if (inp != NULL && ro != NULL && ro->ro_nh != NULL)
398 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
399 	/*
400 	 * If there is a cached route,
401 	 * check that it is to the same destination
402 	 * and is still up.  If not, free it and try again.
403 	 * The address family should also be checked in case of sharing the
404 	 * cache with IPv6.
405 	 * Also check whether routing cache needs invalidation.
406 	 */
407 	if (ro != NULL && ro->ro_nh != NULL &&
408 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
409 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
410 		RO_INVALIDATE_CACHE(ro);
411 	ia = NULL;
412 	/*
413 	 * If routing to interface only, short circuit routing lookup.
414 	 * The use of an all-ones broadcast address implies this; an
415 	 * interface is specified by the broadcast address of an interface,
416 	 * or the destination address of a ptp interface.
417 	 */
418 	if (flags & IP_SENDONES) {
419 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
420 						      M_GETFIB(m)))) == NULL &&
421 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
422 						    M_GETFIB(m)))) == NULL) {
423 			IPSTAT_INC(ips_noroute);
424 			error = ENETUNREACH;
425 			goto bad;
426 		}
427 		ip->ip_dst.s_addr = INADDR_BROADCAST;
428 		dst->sin_addr = ip->ip_dst;
429 		ifp = ia->ia_ifp;
430 		mtu = ifp->if_mtu;
431 		ip->ip_ttl = 1;
432 		isbroadcast = 1;
433 		src = IA_SIN(ia)->sin_addr;
434 	} else if (flags & IP_ROUTETOIF) {
435 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
436 						    M_GETFIB(m)))) == NULL &&
437 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
438 						M_GETFIB(m)))) == NULL) {
439 			IPSTAT_INC(ips_noroute);
440 			error = ENETUNREACH;
441 			goto bad;
442 		}
443 		ifp = ia->ia_ifp;
444 		mtu = ifp->if_mtu;
445 		ip->ip_ttl = 1;
446 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
447 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
448 		src = IA_SIN(ia)->sin_addr;
449 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
450 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
451 		/*
452 		 * Bypass the normal routing lookup for multicast
453 		 * packets if the interface is specified.
454 		 */
455 		ifp = imo->imo_multicast_ifp;
456 		mtu = ifp->if_mtu;
457 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
458 		isbroadcast = 0;	/* fool gcc */
459 		/* Interface may have no addresses. */
460 		if (ia != NULL)
461 			src = IA_SIN(ia)->sin_addr;
462 		else
463 			src.s_addr = INADDR_ANY;
464 	} else if (ro != NULL) {
465 		if (ro->ro_nh == NULL) {
466 			/*
467 			 * We want to do any cloning requested by the link
468 			 * layer, as this is probably required in all cases
469 			 * for correct operation (as it is for ARP).
470 			 */
471 			uint32_t flowid;
472 			flowid = m->m_pkthdr.flowid;
473 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
474 			    NHR_REF, flowid);
475 
476 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
477 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
478 				/*
479 				 * There is no route for this packet, but it is
480 				 * possible that a matching SPD entry exists.
481 				 */
482 				no_route_but_check_spd = 1;
483 				goto sendit;
484 #endif
485 				IPSTAT_INC(ips_noroute);
486 				error = EHOSTUNREACH;
487 				goto bad;
488 			}
489 		}
490 		ia = ifatoia(ro->ro_nh->nh_ifa);
491 		ifp = ro->ro_nh->nh_ifp;
492 		counter_u64_add(ro->ro_nh->nh_pksent, 1);
493 		rt_update_ro_flags(ro);
494 		if (ro->ro_nh->nh_flags & NHF_GATEWAY)
495 			gw = &ro->ro_nh->gw4_sa;
496 		if (ro->ro_nh->nh_flags & NHF_HOST)
497 			isbroadcast = (ro->ro_nh->nh_flags & NHF_BROADCAST);
498 		else if (ifp->if_flags & IFF_BROADCAST)
499 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
500 		else
501 			isbroadcast = 0;
502 		if (ro->ro_nh->nh_flags & NHF_HOST)
503 			mtu = ro->ro_nh->nh_mtu;
504 		else
505 			mtu = ifp->if_mtu;
506 		src = IA_SIN(ia)->sin_addr;
507 	} else {
508 		struct nhop_object *nh;
509 
510 		nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE,
511 		    m->m_pkthdr.flowid);
512 		if (nh == NULL) {
513 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
514 			/*
515 			 * There is no route for this packet, but it is
516 			 * possible that a matching SPD entry exists.
517 			 */
518 			no_route_but_check_spd = 1;
519 			goto sendit;
520 #endif
521 			IPSTAT_INC(ips_noroute);
522 			error = EHOSTUNREACH;
523 			goto bad;
524 		}
525 		ifp = nh->nh_ifp;
526 		mtu = nh->nh_mtu;
527 		/*
528 		 * We are rewriting here dst to be gw actually, contradicting
529 		 * comment at the beginning of the function. However, in this
530 		 * case we are always dealing with on stack dst.
531 		 * In case if pfil(9) sends us back to beginning of the
532 		 * function, the dst would be rewritten by ip_output_pfil().
533 		 */
534 		MPASS(dst == &sin);
535 		if (nh->nh_flags & NHF_GATEWAY)
536 			dst->sin_addr = nh->gw4_sa.sin_addr;
537 		ia = ifatoia(nh->nh_ifa);
538 		src = IA_SIN(ia)->sin_addr;
539 		isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
540 		    (NHF_HOST | NHF_BROADCAST)) ||
541 		    ((ifp->if_flags & IFF_BROADCAST) &&
542 		    in_ifaddr_broadcast(dst->sin_addr, ia)));
543 	}
544 
545 	/* Catch a possible divide by zero later. */
546 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
547 	    __func__, mtu, ro,
548 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
549 
550 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
551 		m->m_flags |= M_MCAST;
552 		/*
553 		 * IP destination address is multicast.  Make sure "gw"
554 		 * still points to the address in "ro".  (It may have been
555 		 * changed to point to a gateway address, above.)
556 		 */
557 		gw = dst;
558 		/*
559 		 * See if the caller provided any multicast options
560 		 */
561 		if (imo != NULL) {
562 			ip->ip_ttl = imo->imo_multicast_ttl;
563 			if (imo->imo_multicast_vif != -1)
564 				ip->ip_src.s_addr =
565 				    ip_mcast_src ?
566 				    ip_mcast_src(imo->imo_multicast_vif) :
567 				    INADDR_ANY;
568 		} else
569 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
570 		/*
571 		 * Confirm that the outgoing interface supports multicast.
572 		 */
573 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
574 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
575 				IPSTAT_INC(ips_noroute);
576 				error = ENETUNREACH;
577 				goto bad;
578 			}
579 		}
580 		/*
581 		 * If source address not specified yet, use address
582 		 * of outgoing interface.
583 		 */
584 		if (ip->ip_src.s_addr == INADDR_ANY)
585 			ip->ip_src = src;
586 
587 		if ((imo == NULL && in_mcast_loop) ||
588 		    (imo && imo->imo_multicast_loop)) {
589 			/*
590 			 * Loop back multicast datagram if not expressly
591 			 * forbidden to do so, even if we are not a member
592 			 * of the group; ip_input() will filter it later,
593 			 * thus deferring a hash lookup and mutex acquisition
594 			 * at the expense of a cheap copy using m_copym().
595 			 */
596 			ip_mloopback(ifp, m, hlen);
597 		} else {
598 			/*
599 			 * If we are acting as a multicast router, perform
600 			 * multicast forwarding as if the packet had just
601 			 * arrived on the interface to which we are about
602 			 * to send.  The multicast forwarding function
603 			 * recursively calls this function, using the
604 			 * IP_FORWARDING flag to prevent infinite recursion.
605 			 *
606 			 * Multicasts that are looped back by ip_mloopback(),
607 			 * above, will be forwarded by the ip_input() routine,
608 			 * if necessary.
609 			 */
610 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
611 				/*
612 				 * If rsvp daemon is not running, do not
613 				 * set ip_moptions. This ensures that the packet
614 				 * is multicast and not just sent down one link
615 				 * as prescribed by rsvpd.
616 				 */
617 				if (!V_rsvp_on)
618 					imo = NULL;
619 				if (ip_mforward &&
620 				    ip_mforward(ip, ifp, m, imo) != 0) {
621 					m_freem(m);
622 					goto done;
623 				}
624 			}
625 		}
626 
627 		/*
628 		 * Multicasts with a time-to-live of zero may be looped-
629 		 * back, above, but must not be transmitted on a network.
630 		 * Also, multicasts addressed to the loopback interface
631 		 * are not sent -- the above call to ip_mloopback() will
632 		 * loop back a copy. ip_input() will drop the copy if
633 		 * this host does not belong to the destination group on
634 		 * the loopback interface.
635 		 */
636 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
637 			m_freem(m);
638 			goto done;
639 		}
640 
641 		goto sendit;
642 	}
643 
644 	/*
645 	 * If the source address is not specified yet, use the address
646 	 * of the outoing interface.
647 	 */
648 	if (ip->ip_src.s_addr == INADDR_ANY)
649 		ip->ip_src = src;
650 
651 	/*
652 	 * Look for broadcast address and
653 	 * verify user is allowed to send
654 	 * such a packet.
655 	 */
656 	if (isbroadcast) {
657 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
658 			error = EADDRNOTAVAIL;
659 			goto bad;
660 		}
661 		if ((flags & IP_ALLOWBROADCAST) == 0) {
662 			error = EACCES;
663 			goto bad;
664 		}
665 		/* don't allow broadcast messages to be fragmented */
666 		if (ip_len > mtu) {
667 			error = EMSGSIZE;
668 			goto bad;
669 		}
670 		m->m_flags |= M_BCAST;
671 	} else {
672 		m->m_flags &= ~M_BCAST;
673 	}
674 
675 sendit:
676 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
677 	if (IPSEC_ENABLED(ipv4)) {
678 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
679 			if (error == EINPROGRESS)
680 				error = 0;
681 			goto done;
682 		}
683 	}
684 	/*
685 	 * Check if there was a route for this packet; return error if not.
686 	 */
687 	if (no_route_but_check_spd) {
688 		IPSTAT_INC(ips_noroute);
689 		error = EHOSTUNREACH;
690 		goto bad;
691 	}
692 	/* Update variables that are affected by ipsec4_output(). */
693 	ip = mtod(m, struct ip *);
694 	hlen = ip->ip_hl << 2;
695 #endif /* IPSEC */
696 
697 	/* Jump over all PFIL processing if hooks are not active. */
698 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
699 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
700 		    &error)) {
701 		case 1: /* Finished */
702 			goto done;
703 
704 		case 0: /* Continue normally */
705 			ip = mtod(m, struct ip *);
706 			break;
707 
708 		case -1: /* Need to try again */
709 			/* Reset everything for a new round */
710 			if (ro != NULL) {
711 				RO_NHFREE(ro);
712 				ro->ro_prepend = NULL;
713 			}
714 			gw = dst;
715 			ip = mtod(m, struct ip *);
716 			goto again;
717 		}
718 	}
719 
720 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
721 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
722 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
723 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
724 			IPSTAT_INC(ips_badaddr);
725 			error = EADDRNOTAVAIL;
726 			goto bad;
727 		}
728 	}
729 
730 	m->m_pkthdr.csum_flags |= CSUM_IP;
731 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
732 		m = mb_unmapped_to_ext(m);
733 		if (m == NULL) {
734 			IPSTAT_INC(ips_odropped);
735 			error = ENOBUFS;
736 			goto bad;
737 		}
738 		in_delayed_cksum(m);
739 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
740 	} else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
741 		m = mb_unmapped_to_ext(m);
742 		if (m == NULL) {
743 			IPSTAT_INC(ips_odropped);
744 			error = ENOBUFS;
745 			goto bad;
746 		}
747 	}
748 #if defined(SCTP) || defined(SCTP_SUPPORT)
749 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
750 		m = mb_unmapped_to_ext(m);
751 		if (m == NULL) {
752 			IPSTAT_INC(ips_odropped);
753 			error = ENOBUFS;
754 			goto bad;
755 		}
756 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
757 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
758 	}
759 #endif
760 
761 	/*
762 	 * If small enough for interface, or the interface will take
763 	 * care of the fragmentation for us, we can just send directly.
764 	 * Note that if_vxlan could have requested TSO even though the outer
765 	 * frame is UDP.  It is correct to not fragment such datagrams and
766 	 * instead just pass them on to the driver.
767 	 */
768 	if (ip_len <= mtu ||
769 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
770 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
771 		ip->ip_sum = 0;
772 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
773 			ip->ip_sum = in_cksum(m, hlen);
774 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
775 		}
776 
777 		/*
778 		 * Record statistics for this interface address.
779 		 * With CSUM_TSO the byte/packet count will be slightly
780 		 * incorrect because we count the IP+TCP headers only
781 		 * once instead of for every generated packet.
782 		 */
783 		if (!(flags & IP_FORWARDING) && ia) {
784 			if (m->m_pkthdr.csum_flags &
785 			    (CSUM_TSO | CSUM_INNER_TSO))
786 				counter_u64_add(ia->ia_ifa.ifa_opackets,
787 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
788 			else
789 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
790 
791 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
792 		}
793 #ifdef MBUF_STRESS_TEST
794 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
795 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
796 #endif
797 		/*
798 		 * Reset layer specific mbuf flags
799 		 * to avoid confusing lower layers.
800 		 */
801 		m_clrprotoflags(m);
802 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
803 		error = ip_output_send(inp, ifp, m, gw, ro,
804 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
805 		goto done;
806 	}
807 
808 	/* Balk when DF bit is set or the interface didn't support TSO. */
809 	if ((ip_off & IP_DF) ||
810 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
811 		error = EMSGSIZE;
812 		IPSTAT_INC(ips_cantfrag);
813 		goto bad;
814 	}
815 
816 	/*
817 	 * Too large for interface; fragment if possible. If successful,
818 	 * on return, m will point to a list of packets to be sent.
819 	 */
820 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
821 	if (error)
822 		goto bad;
823 	for (; m; m = m0) {
824 		m0 = m->m_nextpkt;
825 		m->m_nextpkt = 0;
826 		if (error == 0) {
827 			/* Record statistics for this interface address. */
828 			if (ia != NULL) {
829 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
830 				counter_u64_add(ia->ia_ifa.ifa_obytes,
831 				    m->m_pkthdr.len);
832 			}
833 			/*
834 			 * Reset layer specific mbuf flags
835 			 * to avoid confusing upper layers.
836 			 */
837 			m_clrprotoflags(m);
838 
839 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
840 			    mtod(m, struct ip *), NULL);
841 			error = ip_output_send(inp, ifp, m, gw, ro, true);
842 		} else
843 			m_freem(m);
844 	}
845 
846 	if (error == 0)
847 		IPSTAT_INC(ips_fragmented);
848 
849 done:
850 	return (error);
851  bad:
852 	m_freem(m);
853 	goto done;
854 }
855 
856 /*
857  * Create a chain of fragments which fit the given mtu. m_frag points to the
858  * mbuf to be fragmented; on return it points to the chain with the fragments.
859  * Return 0 if no error. If error, m_frag may contain a partially built
860  * chain of fragments that should be freed by the caller.
861  *
862  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
863  */
864 int
865 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
866     u_long if_hwassist_flags)
867 {
868 	int error = 0;
869 	int hlen = ip->ip_hl << 2;
870 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
871 	int off;
872 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
873 	int firstlen;
874 	struct mbuf **mnext;
875 	int nfrags;
876 	uint16_t ip_len, ip_off;
877 
878 	ip_len = ntohs(ip->ip_len);
879 	ip_off = ntohs(ip->ip_off);
880 
881 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
882 		IPSTAT_INC(ips_cantfrag);
883 		return EMSGSIZE;
884 	}
885 
886 	/*
887 	 * Must be able to put at least 8 bytes per fragment.
888 	 */
889 	if (len < 8)
890 		return EMSGSIZE;
891 
892 	/*
893 	 * If the interface will not calculate checksums on
894 	 * fragmented packets, then do it here.
895 	 */
896 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
897 		m0 = mb_unmapped_to_ext(m0);
898 		if (m0 == NULL) {
899 			error = ENOBUFS;
900 			IPSTAT_INC(ips_odropped);
901 			goto done;
902 		}
903 		in_delayed_cksum(m0);
904 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
905 	}
906 #if defined(SCTP) || defined(SCTP_SUPPORT)
907 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
908 		m0 = mb_unmapped_to_ext(m0);
909 		if (m0 == NULL) {
910 			error = ENOBUFS;
911 			IPSTAT_INC(ips_odropped);
912 			goto done;
913 		}
914 		sctp_delayed_cksum(m0, hlen);
915 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
916 	}
917 #endif
918 	if (len > PAGE_SIZE) {
919 		/*
920 		 * Fragment large datagrams such that each segment
921 		 * contains a multiple of PAGE_SIZE amount of data,
922 		 * plus headers. This enables a receiver to perform
923 		 * page-flipping zero-copy optimizations.
924 		 *
925 		 * XXX When does this help given that sender and receiver
926 		 * could have different page sizes, and also mtu could
927 		 * be less than the receiver's page size ?
928 		 */
929 		int newlen;
930 
931 		off = MIN(mtu, m0->m_pkthdr.len);
932 
933 		/*
934 		 * firstlen (off - hlen) must be aligned on an
935 		 * 8-byte boundary
936 		 */
937 		if (off < hlen)
938 			goto smart_frag_failure;
939 		off = ((off - hlen) & ~7) + hlen;
940 		newlen = (~PAGE_MASK) & mtu;
941 		if ((newlen + sizeof (struct ip)) > mtu) {
942 			/* we failed, go back the default */
943 smart_frag_failure:
944 			newlen = len;
945 			off = hlen + len;
946 		}
947 		len = newlen;
948 
949 	} else {
950 		off = hlen + len;
951 	}
952 
953 	firstlen = off - hlen;
954 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
955 
956 	/*
957 	 * Loop through length of segment after first fragment,
958 	 * make new header and copy data of each part and link onto chain.
959 	 * Here, m0 is the original packet, m is the fragment being created.
960 	 * The fragments are linked off the m_nextpkt of the original
961 	 * packet, which after processing serves as the first fragment.
962 	 */
963 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
964 		struct ip *mhip;	/* ip header on the fragment */
965 		struct mbuf *m;
966 		int mhlen = sizeof (struct ip);
967 
968 		m = m_gethdr(M_NOWAIT, MT_DATA);
969 		if (m == NULL) {
970 			error = ENOBUFS;
971 			IPSTAT_INC(ips_odropped);
972 			goto done;
973 		}
974 		/*
975 		 * Make sure the complete packet header gets copied
976 		 * from the originating mbuf to the newly created
977 		 * mbuf. This also ensures that existing firewall
978 		 * classification(s), VLAN tags and so on get copied
979 		 * to the resulting fragmented packet(s):
980 		 */
981 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
982 			m_free(m);
983 			error = ENOBUFS;
984 			IPSTAT_INC(ips_odropped);
985 			goto done;
986 		}
987 		/*
988 		 * In the first mbuf, leave room for the link header, then
989 		 * copy the original IP header including options. The payload
990 		 * goes into an additional mbuf chain returned by m_copym().
991 		 */
992 		m->m_data += max_linkhdr;
993 		mhip = mtod(m, struct ip *);
994 		*mhip = *ip;
995 		if (hlen > sizeof (struct ip)) {
996 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
997 			mhip->ip_v = IPVERSION;
998 			mhip->ip_hl = mhlen >> 2;
999 		}
1000 		m->m_len = mhlen;
1001 		/* XXX do we need to add ip_off below ? */
1002 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1003 		if (off + len >= ip_len)
1004 			len = ip_len - off;
1005 		else
1006 			mhip->ip_off |= IP_MF;
1007 		mhip->ip_len = htons((u_short)(len + mhlen));
1008 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1009 		if (m->m_next == NULL) {	/* copy failed */
1010 			m_free(m);
1011 			error = ENOBUFS;	/* ??? */
1012 			IPSTAT_INC(ips_odropped);
1013 			goto done;
1014 		}
1015 		m->m_pkthdr.len = mhlen + len;
1016 #ifdef MAC
1017 		mac_netinet_fragment(m0, m);
1018 #endif
1019 		mhip->ip_off = htons(mhip->ip_off);
1020 		mhip->ip_sum = 0;
1021 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1022 			mhip->ip_sum = in_cksum(m, mhlen);
1023 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1024 		}
1025 		*mnext = m;
1026 		mnext = &m->m_nextpkt;
1027 	}
1028 	IPSTAT_ADD(ips_ofragments, nfrags);
1029 
1030 	/*
1031 	 * Update first fragment by trimming what's been copied out
1032 	 * and updating header.
1033 	 */
1034 	m_adj(m0, hlen + firstlen - ip_len);
1035 	m0->m_pkthdr.len = hlen + firstlen;
1036 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1037 	ip->ip_off = htons(ip_off | IP_MF);
1038 	ip->ip_sum = 0;
1039 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1040 		ip->ip_sum = in_cksum(m0, hlen);
1041 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1042 	}
1043 
1044 done:
1045 	*m_frag = m0;
1046 	return error;
1047 }
1048 
1049 void
1050 in_delayed_cksum(struct mbuf *m)
1051 {
1052 	struct ip *ip;
1053 	struct udphdr *uh;
1054 	uint16_t cklen, csum, offset;
1055 
1056 	ip = mtod(m, struct ip *);
1057 	offset = ip->ip_hl << 2 ;
1058 
1059 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1060 		/* if udp header is not in the first mbuf copy udplen */
1061 		if (offset + sizeof(struct udphdr) > m->m_len) {
1062 			m_copydata(m, offset + offsetof(struct udphdr,
1063 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1064 			cklen = ntohs(cklen);
1065 		} else {
1066 			uh = (struct udphdr *)mtodo(m, offset);
1067 			cklen = ntohs(uh->uh_ulen);
1068 		}
1069 		csum = in_cksum_skip(m, cklen + offset, offset);
1070 		if (csum == 0)
1071 			csum = 0xffff;
1072 	} else {
1073 		cklen = ntohs(ip->ip_len);
1074 		csum = in_cksum_skip(m, cklen, offset);
1075 	}
1076 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1077 
1078 	if (offset + sizeof(csum) > m->m_len)
1079 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1080 	else
1081 		*(u_short *)mtodo(m, offset) = csum;
1082 }
1083 
1084 /*
1085  * IP socket option processing.
1086  */
1087 int
1088 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1089 {
1090 	struct inpcb *inp = sotoinpcb(so);
1091 	int	error, optval;
1092 #ifdef	RSS
1093 	uint32_t rss_bucket;
1094 	int retval;
1095 #endif
1096 
1097 	error = optval = 0;
1098 	if (sopt->sopt_level != IPPROTO_IP) {
1099 		error = EINVAL;
1100 
1101 		if (sopt->sopt_level == SOL_SOCKET &&
1102 		    sopt->sopt_dir == SOPT_SET) {
1103 			switch (sopt->sopt_name) {
1104 			case SO_REUSEADDR:
1105 				INP_WLOCK(inp);
1106 				if ((so->so_options & SO_REUSEADDR) != 0)
1107 					inp->inp_flags2 |= INP_REUSEADDR;
1108 				else
1109 					inp->inp_flags2 &= ~INP_REUSEADDR;
1110 				INP_WUNLOCK(inp);
1111 				error = 0;
1112 				break;
1113 			case SO_REUSEPORT:
1114 				INP_WLOCK(inp);
1115 				if ((so->so_options & SO_REUSEPORT) != 0)
1116 					inp->inp_flags2 |= INP_REUSEPORT;
1117 				else
1118 					inp->inp_flags2 &= ~INP_REUSEPORT;
1119 				INP_WUNLOCK(inp);
1120 				error = 0;
1121 				break;
1122 			case SO_REUSEPORT_LB:
1123 				INP_WLOCK(inp);
1124 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1125 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1126 				else
1127 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1128 				INP_WUNLOCK(inp);
1129 				error = 0;
1130 				break;
1131 			case SO_SETFIB:
1132 				INP_WLOCK(inp);
1133 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1134 				INP_WUNLOCK(inp);
1135 				error = 0;
1136 				break;
1137 			case SO_MAX_PACING_RATE:
1138 #ifdef RATELIMIT
1139 				INP_WLOCK(inp);
1140 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1141 				INP_WUNLOCK(inp);
1142 				error = 0;
1143 #else
1144 				error = EOPNOTSUPP;
1145 #endif
1146 				break;
1147 			default:
1148 				break;
1149 			}
1150 		}
1151 		return (error);
1152 	}
1153 
1154 	switch (sopt->sopt_dir) {
1155 	case SOPT_SET:
1156 		switch (sopt->sopt_name) {
1157 		case IP_OPTIONS:
1158 #ifdef notyet
1159 		case IP_RETOPTS:
1160 #endif
1161 		{
1162 			struct mbuf *m;
1163 			if (sopt->sopt_valsize > MLEN) {
1164 				error = EMSGSIZE;
1165 				break;
1166 			}
1167 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1168 			if (m == NULL) {
1169 				error = ENOBUFS;
1170 				break;
1171 			}
1172 			m->m_len = sopt->sopt_valsize;
1173 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1174 					    m->m_len);
1175 			if (error) {
1176 				m_free(m);
1177 				break;
1178 			}
1179 			INP_WLOCK(inp);
1180 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1181 			INP_WUNLOCK(inp);
1182 			return (error);
1183 		}
1184 
1185 		case IP_BINDANY:
1186 			if (sopt->sopt_td != NULL) {
1187 				error = priv_check(sopt->sopt_td,
1188 				    PRIV_NETINET_BINDANY);
1189 				if (error)
1190 					break;
1191 			}
1192 			/* FALLTHROUGH */
1193 		case IP_BINDMULTI:
1194 #ifdef	RSS
1195 		case IP_RSS_LISTEN_BUCKET:
1196 #endif
1197 		case IP_TOS:
1198 		case IP_TTL:
1199 		case IP_MINTTL:
1200 		case IP_RECVOPTS:
1201 		case IP_RECVRETOPTS:
1202 		case IP_ORIGDSTADDR:
1203 		case IP_RECVDSTADDR:
1204 		case IP_RECVTTL:
1205 		case IP_RECVIF:
1206 		case IP_ONESBCAST:
1207 		case IP_DONTFRAG:
1208 		case IP_RECVTOS:
1209 		case IP_RECVFLOWID:
1210 #ifdef	RSS
1211 		case IP_RECVRSSBUCKETID:
1212 #endif
1213 			error = sooptcopyin(sopt, &optval, sizeof optval,
1214 					    sizeof optval);
1215 			if (error)
1216 				break;
1217 
1218 			switch (sopt->sopt_name) {
1219 			case IP_TOS:
1220 				inp->inp_ip_tos = optval;
1221 				break;
1222 
1223 			case IP_TTL:
1224 				inp->inp_ip_ttl = optval;
1225 				break;
1226 
1227 			case IP_MINTTL:
1228 				if (optval >= 0 && optval <= MAXTTL)
1229 					inp->inp_ip_minttl = optval;
1230 				else
1231 					error = EINVAL;
1232 				break;
1233 
1234 #define	OPTSET(bit) do {						\
1235 	INP_WLOCK(inp);							\
1236 	if (optval)							\
1237 		inp->inp_flags |= bit;					\
1238 	else								\
1239 		inp->inp_flags &= ~bit;					\
1240 	INP_WUNLOCK(inp);						\
1241 } while (0)
1242 
1243 #define	OPTSET2(bit, val) do {						\
1244 	INP_WLOCK(inp);							\
1245 	if (val)							\
1246 		inp->inp_flags2 |= bit;					\
1247 	else								\
1248 		inp->inp_flags2 &= ~bit;				\
1249 	INP_WUNLOCK(inp);						\
1250 } while (0)
1251 
1252 			case IP_RECVOPTS:
1253 				OPTSET(INP_RECVOPTS);
1254 				break;
1255 
1256 			case IP_RECVRETOPTS:
1257 				OPTSET(INP_RECVRETOPTS);
1258 				break;
1259 
1260 			case IP_RECVDSTADDR:
1261 				OPTSET(INP_RECVDSTADDR);
1262 				break;
1263 
1264 			case IP_ORIGDSTADDR:
1265 				OPTSET2(INP_ORIGDSTADDR, optval);
1266 				break;
1267 
1268 			case IP_RECVTTL:
1269 				OPTSET(INP_RECVTTL);
1270 				break;
1271 
1272 			case IP_RECVIF:
1273 				OPTSET(INP_RECVIF);
1274 				break;
1275 
1276 			case IP_ONESBCAST:
1277 				OPTSET(INP_ONESBCAST);
1278 				break;
1279 			case IP_DONTFRAG:
1280 				OPTSET(INP_DONTFRAG);
1281 				break;
1282 			case IP_BINDANY:
1283 				OPTSET(INP_BINDANY);
1284 				break;
1285 			case IP_RECVTOS:
1286 				OPTSET(INP_RECVTOS);
1287 				break;
1288 			case IP_BINDMULTI:
1289 				OPTSET2(INP_BINDMULTI, optval);
1290 				break;
1291 			case IP_RECVFLOWID:
1292 				OPTSET2(INP_RECVFLOWID, optval);
1293 				break;
1294 #ifdef	RSS
1295 			case IP_RSS_LISTEN_BUCKET:
1296 				if ((optval >= 0) &&
1297 				    (optval < rss_getnumbuckets())) {
1298 					inp->inp_rss_listen_bucket = optval;
1299 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1300 				} else {
1301 					error = EINVAL;
1302 				}
1303 				break;
1304 			case IP_RECVRSSBUCKETID:
1305 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1306 				break;
1307 #endif
1308 			}
1309 			break;
1310 #undef OPTSET
1311 #undef OPTSET2
1312 
1313 		/*
1314 		 * Multicast socket options are processed by the in_mcast
1315 		 * module.
1316 		 */
1317 		case IP_MULTICAST_IF:
1318 		case IP_MULTICAST_VIF:
1319 		case IP_MULTICAST_TTL:
1320 		case IP_MULTICAST_LOOP:
1321 		case IP_ADD_MEMBERSHIP:
1322 		case IP_DROP_MEMBERSHIP:
1323 		case IP_ADD_SOURCE_MEMBERSHIP:
1324 		case IP_DROP_SOURCE_MEMBERSHIP:
1325 		case IP_BLOCK_SOURCE:
1326 		case IP_UNBLOCK_SOURCE:
1327 		case IP_MSFILTER:
1328 		case MCAST_JOIN_GROUP:
1329 		case MCAST_LEAVE_GROUP:
1330 		case MCAST_JOIN_SOURCE_GROUP:
1331 		case MCAST_LEAVE_SOURCE_GROUP:
1332 		case MCAST_BLOCK_SOURCE:
1333 		case MCAST_UNBLOCK_SOURCE:
1334 			error = inp_setmoptions(inp, sopt);
1335 			break;
1336 
1337 		case IP_PORTRANGE:
1338 			error = sooptcopyin(sopt, &optval, sizeof optval,
1339 					    sizeof optval);
1340 			if (error)
1341 				break;
1342 
1343 			INP_WLOCK(inp);
1344 			switch (optval) {
1345 			case IP_PORTRANGE_DEFAULT:
1346 				inp->inp_flags &= ~(INP_LOWPORT);
1347 				inp->inp_flags &= ~(INP_HIGHPORT);
1348 				break;
1349 
1350 			case IP_PORTRANGE_HIGH:
1351 				inp->inp_flags &= ~(INP_LOWPORT);
1352 				inp->inp_flags |= INP_HIGHPORT;
1353 				break;
1354 
1355 			case IP_PORTRANGE_LOW:
1356 				inp->inp_flags &= ~(INP_HIGHPORT);
1357 				inp->inp_flags |= INP_LOWPORT;
1358 				break;
1359 
1360 			default:
1361 				error = EINVAL;
1362 				break;
1363 			}
1364 			INP_WUNLOCK(inp);
1365 			break;
1366 
1367 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1368 		case IP_IPSEC_POLICY:
1369 			if (IPSEC_ENABLED(ipv4)) {
1370 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1371 				break;
1372 			}
1373 			/* FALLTHROUGH */
1374 #endif /* IPSEC */
1375 
1376 		default:
1377 			error = ENOPROTOOPT;
1378 			break;
1379 		}
1380 		break;
1381 
1382 	case SOPT_GET:
1383 		switch (sopt->sopt_name) {
1384 		case IP_OPTIONS:
1385 		case IP_RETOPTS:
1386 			INP_RLOCK(inp);
1387 			if (inp->inp_options) {
1388 				struct mbuf *options;
1389 
1390 				options = m_copym(inp->inp_options, 0,
1391 				    M_COPYALL, M_NOWAIT);
1392 				INP_RUNLOCK(inp);
1393 				if (options != NULL) {
1394 					error = sooptcopyout(sopt,
1395 							     mtod(options, char *),
1396 							     options->m_len);
1397 					m_freem(options);
1398 				} else
1399 					error = ENOMEM;
1400 			} else {
1401 				INP_RUNLOCK(inp);
1402 				sopt->sopt_valsize = 0;
1403 			}
1404 			break;
1405 
1406 		case IP_TOS:
1407 		case IP_TTL:
1408 		case IP_MINTTL:
1409 		case IP_RECVOPTS:
1410 		case IP_RECVRETOPTS:
1411 		case IP_ORIGDSTADDR:
1412 		case IP_RECVDSTADDR:
1413 		case IP_RECVTTL:
1414 		case IP_RECVIF:
1415 		case IP_PORTRANGE:
1416 		case IP_ONESBCAST:
1417 		case IP_DONTFRAG:
1418 		case IP_BINDANY:
1419 		case IP_RECVTOS:
1420 		case IP_BINDMULTI:
1421 		case IP_FLOWID:
1422 		case IP_FLOWTYPE:
1423 		case IP_RECVFLOWID:
1424 #ifdef	RSS
1425 		case IP_RSSBUCKETID:
1426 		case IP_RECVRSSBUCKETID:
1427 #endif
1428 			switch (sopt->sopt_name) {
1429 			case IP_TOS:
1430 				optval = inp->inp_ip_tos;
1431 				break;
1432 
1433 			case IP_TTL:
1434 				optval = inp->inp_ip_ttl;
1435 				break;
1436 
1437 			case IP_MINTTL:
1438 				optval = inp->inp_ip_minttl;
1439 				break;
1440 
1441 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1442 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1443 
1444 			case IP_RECVOPTS:
1445 				optval = OPTBIT(INP_RECVOPTS);
1446 				break;
1447 
1448 			case IP_RECVRETOPTS:
1449 				optval = OPTBIT(INP_RECVRETOPTS);
1450 				break;
1451 
1452 			case IP_RECVDSTADDR:
1453 				optval = OPTBIT(INP_RECVDSTADDR);
1454 				break;
1455 
1456 			case IP_ORIGDSTADDR:
1457 				optval = OPTBIT2(INP_ORIGDSTADDR);
1458 				break;
1459 
1460 			case IP_RECVTTL:
1461 				optval = OPTBIT(INP_RECVTTL);
1462 				break;
1463 
1464 			case IP_RECVIF:
1465 				optval = OPTBIT(INP_RECVIF);
1466 				break;
1467 
1468 			case IP_PORTRANGE:
1469 				if (inp->inp_flags & INP_HIGHPORT)
1470 					optval = IP_PORTRANGE_HIGH;
1471 				else if (inp->inp_flags & INP_LOWPORT)
1472 					optval = IP_PORTRANGE_LOW;
1473 				else
1474 					optval = 0;
1475 				break;
1476 
1477 			case IP_ONESBCAST:
1478 				optval = OPTBIT(INP_ONESBCAST);
1479 				break;
1480 			case IP_DONTFRAG:
1481 				optval = OPTBIT(INP_DONTFRAG);
1482 				break;
1483 			case IP_BINDANY:
1484 				optval = OPTBIT(INP_BINDANY);
1485 				break;
1486 			case IP_RECVTOS:
1487 				optval = OPTBIT(INP_RECVTOS);
1488 				break;
1489 			case IP_FLOWID:
1490 				optval = inp->inp_flowid;
1491 				break;
1492 			case IP_FLOWTYPE:
1493 				optval = inp->inp_flowtype;
1494 				break;
1495 			case IP_RECVFLOWID:
1496 				optval = OPTBIT2(INP_RECVFLOWID);
1497 				break;
1498 #ifdef	RSS
1499 			case IP_RSSBUCKETID:
1500 				retval = rss_hash2bucket(inp->inp_flowid,
1501 				    inp->inp_flowtype,
1502 				    &rss_bucket);
1503 				if (retval == 0)
1504 					optval = rss_bucket;
1505 				else
1506 					error = EINVAL;
1507 				break;
1508 			case IP_RECVRSSBUCKETID:
1509 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1510 				break;
1511 #endif
1512 			case IP_BINDMULTI:
1513 				optval = OPTBIT2(INP_BINDMULTI);
1514 				break;
1515 			}
1516 			error = sooptcopyout(sopt, &optval, sizeof optval);
1517 			break;
1518 
1519 		/*
1520 		 * Multicast socket options are processed by the in_mcast
1521 		 * module.
1522 		 */
1523 		case IP_MULTICAST_IF:
1524 		case IP_MULTICAST_VIF:
1525 		case IP_MULTICAST_TTL:
1526 		case IP_MULTICAST_LOOP:
1527 		case IP_MSFILTER:
1528 			error = inp_getmoptions(inp, sopt);
1529 			break;
1530 
1531 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1532 		case IP_IPSEC_POLICY:
1533 			if (IPSEC_ENABLED(ipv4)) {
1534 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1535 				break;
1536 			}
1537 			/* FALLTHROUGH */
1538 #endif /* IPSEC */
1539 
1540 		default:
1541 			error = ENOPROTOOPT;
1542 			break;
1543 		}
1544 		break;
1545 	}
1546 	return (error);
1547 }
1548 
1549 /*
1550  * Routine called from ip_output() to loop back a copy of an IP multicast
1551  * packet to the input queue of a specified interface.  Note that this
1552  * calls the output routine of the loopback "driver", but with an interface
1553  * pointer that might NOT be a loopback interface -- evil, but easier than
1554  * replicating that code here.
1555  */
1556 static void
1557 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1558 {
1559 	struct ip *ip;
1560 	struct mbuf *copym;
1561 
1562 	/*
1563 	 * Make a deep copy of the packet because we're going to
1564 	 * modify the pack in order to generate checksums.
1565 	 */
1566 	copym = m_dup(m, M_NOWAIT);
1567 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1568 		copym = m_pullup(copym, hlen);
1569 	if (copym != NULL) {
1570 		/* If needed, compute the checksum and mark it as valid. */
1571 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1572 			in_delayed_cksum(copym);
1573 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1574 			copym->m_pkthdr.csum_flags |=
1575 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1576 			copym->m_pkthdr.csum_data = 0xffff;
1577 		}
1578 		/*
1579 		 * We don't bother to fragment if the IP length is greater
1580 		 * than the interface's MTU.  Can this possibly matter?
1581 		 */
1582 		ip = mtod(copym, struct ip *);
1583 		ip->ip_sum = 0;
1584 		ip->ip_sum = in_cksum(copym, hlen);
1585 		if_simloop(ifp, copym, AF_INET, 0);
1586 	}
1587 }
1588