xref: /freebsd/sys/netinet/ip_output.c (revision fe815331bb40604ba31312acf7e4619674631777)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1990, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_inet.h"
38 #include "opt_ipsec.h"
39 #include "opt_kern_tls.h"
40 #include "opt_mbuf_stress_test.h"
41 #include "opt_mpath.h"
42 #include "opt_ratelimit.h"
43 #include "opt_route.h"
44 #include "opt_rss.h"
45 #include "opt_sctp.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/kernel.h>
50 #include <sys/ktls.h>
51 #include <sys/lock.h>
52 #include <sys/malloc.h>
53 #include <sys/mbuf.h>
54 #include <sys/priv.h>
55 #include <sys/proc.h>
56 #include <sys/protosw.h>
57 #include <sys/rmlock.h>
58 #include <sys/sdt.h>
59 #include <sys/socket.h>
60 #include <sys/socketvar.h>
61 #include <sys/sysctl.h>
62 #include <sys/ucred.h>
63 
64 #include <net/if.h>
65 #include <net/if_var.h>
66 #include <net/if_llatbl.h>
67 #include <net/netisr.h>
68 #include <net/pfil.h>
69 #include <net/route.h>
70 #include <net/route/nhop.h>
71 #include <net/rss_config.h>
72 #include <net/vnet.h>
73 
74 #include <netinet/in.h>
75 #include <netinet/in_fib.h>
76 #include <netinet/in_kdtrace.h>
77 #include <netinet/in_systm.h>
78 #include <netinet/ip.h>
79 #include <netinet/in_fib.h>
80 #include <netinet/in_pcb.h>
81 #include <netinet/in_rss.h>
82 #include <netinet/in_var.h>
83 #include <netinet/ip_var.h>
84 #include <netinet/ip_options.h>
85 
86 #include <netinet/udp.h>
87 #include <netinet/udp_var.h>
88 
89 #if defined(SCTP) || defined(SCTP_SUPPORT)
90 #include <netinet/sctp.h>
91 #include <netinet/sctp_crc32.h>
92 #endif
93 
94 #include <netipsec/ipsec_support.h>
95 
96 #include <machine/in_cksum.h>
97 
98 #include <security/mac/mac_framework.h>
99 
100 #ifdef MBUF_STRESS_TEST
101 static int mbuf_frag_size = 0;
102 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
103 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
104 #endif
105 
106 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
107 
108 extern int in_mcast_loop;
109 extern	struct protosw inetsw[];
110 
111 static inline int
112 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
113     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
114 {
115 	struct m_tag *fwd_tag = NULL;
116 	struct mbuf *m;
117 	struct in_addr odst;
118 	struct ip *ip;
119 	int pflags = PFIL_OUT;
120 
121 	if (flags & IP_FORWARDING)
122 		pflags |= PFIL_FWD;
123 
124 	m = *mp;
125 	ip = mtod(m, struct ip *);
126 
127 	/* Run through list of hooks for output packets. */
128 	odst.s_addr = ip->ip_dst.s_addr;
129 	switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
130 	case PFIL_DROPPED:
131 		*error = EACCES;
132 		/* FALLTHROUGH */
133 	case PFIL_CONSUMED:
134 		return 1; /* Finished */
135 	case PFIL_PASS:
136 		*error = 0;
137 	}
138 	m = *mp;
139 	ip = mtod(m, struct ip *);
140 
141 	/* See if destination IP address was changed by packet filter. */
142 	if (odst.s_addr != ip->ip_dst.s_addr) {
143 		m->m_flags |= M_SKIP_FIREWALL;
144 		/* If destination is now ourself drop to ip_input(). */
145 		if (in_localip(ip->ip_dst)) {
146 			m->m_flags |= M_FASTFWD_OURS;
147 			if (m->m_pkthdr.rcvif == NULL)
148 				m->m_pkthdr.rcvif = V_loif;
149 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
150 				m->m_pkthdr.csum_flags |=
151 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
152 				m->m_pkthdr.csum_data = 0xffff;
153 			}
154 			m->m_pkthdr.csum_flags |=
155 				CSUM_IP_CHECKED | CSUM_IP_VALID;
156 #if defined(SCTP) || defined(SCTP_SUPPORT)
157 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
158 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
159 #endif
160 			*error = netisr_queue(NETISR_IP, m);
161 			return 1; /* Finished */
162 		}
163 
164 		bzero(dst, sizeof(*dst));
165 		dst->sin_family = AF_INET;
166 		dst->sin_len = sizeof(*dst);
167 		dst->sin_addr = ip->ip_dst;
168 
169 		return -1; /* Reloop */
170 	}
171 	/* See if fib was changed by packet filter. */
172 	if ((*fibnum) != M_GETFIB(m)) {
173 		m->m_flags |= M_SKIP_FIREWALL;
174 		*fibnum = M_GETFIB(m);
175 		return -1; /* Reloop for FIB change */
176 	}
177 
178 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
179 	if (m->m_flags & M_FASTFWD_OURS) {
180 		if (m->m_pkthdr.rcvif == NULL)
181 			m->m_pkthdr.rcvif = V_loif;
182 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
183 			m->m_pkthdr.csum_flags |=
184 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
185 			m->m_pkthdr.csum_data = 0xffff;
186 		}
187 #if defined(SCTP) || defined(SCTP_SUPPORT)
188 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
189 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
190 #endif
191 		m->m_pkthdr.csum_flags |=
192 			CSUM_IP_CHECKED | CSUM_IP_VALID;
193 
194 		*error = netisr_queue(NETISR_IP, m);
195 		return 1; /* Finished */
196 	}
197 	/* Or forward to some other address? */
198 	if ((m->m_flags & M_IP_NEXTHOP) &&
199 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
200 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
201 		m->m_flags |= M_SKIP_FIREWALL;
202 		m->m_flags &= ~M_IP_NEXTHOP;
203 		m_tag_delete(m, fwd_tag);
204 
205 		return -1; /* Reloop for CHANGE of dst */
206 	}
207 
208 	return 0;
209 }
210 
211 static int
212 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
213     const struct sockaddr_in *gw, struct route *ro, bool stamp_tag)
214 {
215 #ifdef KERN_TLS
216 	struct ktls_session *tls = NULL;
217 #endif
218 	struct m_snd_tag *mst;
219 	int error;
220 
221 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
222 	mst = NULL;
223 
224 #ifdef KERN_TLS
225 	/*
226 	 * If this is an unencrypted TLS record, save a reference to
227 	 * the record.  This local reference is used to call
228 	 * ktls_output_eagain after the mbuf has been freed (thus
229 	 * dropping the mbuf's reference) in if_output.
230 	 */
231 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
232 		tls = ktls_hold(m->m_next->m_epg_tls);
233 		mst = tls->snd_tag;
234 
235 		/*
236 		 * If a TLS session doesn't have a valid tag, it must
237 		 * have had an earlier ifp mismatch, so drop this
238 		 * packet.
239 		 */
240 		if (mst == NULL) {
241 			error = EAGAIN;
242 			goto done;
243 		}
244 		/*
245 		 * Always stamp tags that include NIC ktls.
246 		 */
247 		stamp_tag = true;
248 	}
249 #endif
250 #ifdef RATELIMIT
251 	if (inp != NULL && mst == NULL) {
252 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
253 		    (inp->inp_snd_tag != NULL &&
254 		    inp->inp_snd_tag->ifp != ifp))
255 			in_pcboutput_txrtlmt(inp, ifp, m);
256 
257 		if (inp->inp_snd_tag != NULL)
258 			mst = inp->inp_snd_tag;
259 	}
260 #endif
261 	if (stamp_tag && mst != NULL) {
262 		KASSERT(m->m_pkthdr.rcvif == NULL,
263 		    ("trying to add a send tag to a forwarded packet"));
264 		if (mst->ifp != ifp) {
265 			error = EAGAIN;
266 			goto done;
267 		}
268 
269 		/* stamp send tag on mbuf */
270 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
271 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
272 	}
273 
274 	error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
275 
276 done:
277 	/* Check for route change invalidating send tags. */
278 #ifdef KERN_TLS
279 	if (tls != NULL) {
280 		if (error == EAGAIN)
281 			error = ktls_output_eagain(inp, tls);
282 		ktls_free(tls);
283 	}
284 #endif
285 #ifdef RATELIMIT
286 	if (error == EAGAIN)
287 		in_pcboutput_eagain(inp);
288 #endif
289 	return (error);
290 }
291 
292 /* rte<>ro_flags translation */
293 static inline void
294 rt_update_ro_flags(struct route *ro)
295 {
296 	int nh_flags = ro->ro_nh->nh_flags;
297 
298 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
299 
300 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
301 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
302 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
303 }
304 
305 /*
306  * IP output.  The packet in mbuf chain m contains a skeletal IP
307  * header (with len, off, ttl, proto, tos, src, dst).
308  * The mbuf chain containing the packet will be freed.
309  * The mbuf opt, if present, will not be freed.
310  * If route ro is present and has ro_rt initialized, route lookup would be
311  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
312  * then result of route lookup is stored in ro->ro_rt.
313  *
314  * In the IP forwarding case, the packet will arrive with options already
315  * inserted, so must have a NULL opt pointer.
316  */
317 int
318 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
319     struct ip_moptions *imo, struct inpcb *inp)
320 {
321 	struct rm_priotracker in_ifa_tracker;
322 	struct ip *ip;
323 	struct ifnet *ifp = NULL;	/* keep compiler happy */
324 	struct mbuf *m0;
325 	int hlen = sizeof (struct ip);
326 	int mtu = 0;
327 	int error = 0;
328 	struct sockaddr_in *dst, sin;
329 	const struct sockaddr_in *gw;
330 	struct in_ifaddr *ia = NULL;
331 	struct in_addr src;
332 	int isbroadcast;
333 	uint16_t ip_len, ip_off;
334 	uint32_t fibnum;
335 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
336 	int no_route_but_check_spd = 0;
337 #endif
338 
339 	M_ASSERTPKTHDR(m);
340 	NET_EPOCH_ASSERT();
341 
342 	if (inp != NULL) {
343 		INP_LOCK_ASSERT(inp);
344 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
345 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
346 			m->m_pkthdr.flowid = inp->inp_flowid;
347 			M_HASHTYPE_SET(m, inp->inp_flowtype);
348 		}
349 #ifdef NUMA
350 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
351 #endif
352 	}
353 
354 	if (opt) {
355 		int len = 0;
356 		m = ip_insertoptions(m, opt, &len);
357 		if (len != 0)
358 			hlen = len; /* ip->ip_hl is updated above */
359 	}
360 	ip = mtod(m, struct ip *);
361 	ip_len = ntohs(ip->ip_len);
362 	ip_off = ntohs(ip->ip_off);
363 
364 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
365 		ip->ip_v = IPVERSION;
366 		ip->ip_hl = hlen >> 2;
367 		ip_fillid(ip);
368 	} else {
369 		/* Header already set, fetch hlen from there */
370 		hlen = ip->ip_hl << 2;
371 	}
372 	if ((flags & IP_FORWARDING) == 0)
373 		IPSTAT_INC(ips_localout);
374 
375 	/*
376 	 * dst/gw handling:
377 	 *
378 	 * gw is readonly but can point either to dst OR rt_gateway,
379 	 * therefore we need restore gw if we're redoing lookup.
380 	 */
381 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
382 	if (ro != NULL)
383 		dst = (struct sockaddr_in *)&ro->ro_dst;
384 	else
385 		dst = &sin;
386 	if (ro == NULL || ro->ro_nh == NULL) {
387 		bzero(dst, sizeof(*dst));
388 		dst->sin_family = AF_INET;
389 		dst->sin_len = sizeof(*dst);
390 		dst->sin_addr = ip->ip_dst;
391 	}
392 	gw = dst;
393 again:
394 	/*
395 	 * Validate route against routing table additions;
396 	 * a better/more specific route might have been added.
397 	 */
398 	if (inp != NULL && ro != NULL && ro->ro_nh != NULL)
399 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
400 	/*
401 	 * If there is a cached route,
402 	 * check that it is to the same destination
403 	 * and is still up.  If not, free it and try again.
404 	 * The address family should also be checked in case of sharing the
405 	 * cache with IPv6.
406 	 * Also check whether routing cache needs invalidation.
407 	 */
408 	if (ro != NULL && ro->ro_nh != NULL &&
409 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
410 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
411 		RO_INVALIDATE_CACHE(ro);
412 	ia = NULL;
413 	/*
414 	 * If routing to interface only, short circuit routing lookup.
415 	 * The use of an all-ones broadcast address implies this; an
416 	 * interface is specified by the broadcast address of an interface,
417 	 * or the destination address of a ptp interface.
418 	 */
419 	if (flags & IP_SENDONES) {
420 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
421 						      M_GETFIB(m)))) == NULL &&
422 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
423 						    M_GETFIB(m)))) == NULL) {
424 			IPSTAT_INC(ips_noroute);
425 			error = ENETUNREACH;
426 			goto bad;
427 		}
428 		ip->ip_dst.s_addr = INADDR_BROADCAST;
429 		dst->sin_addr = ip->ip_dst;
430 		ifp = ia->ia_ifp;
431 		mtu = ifp->if_mtu;
432 		ip->ip_ttl = 1;
433 		isbroadcast = 1;
434 		src = IA_SIN(ia)->sin_addr;
435 	} else if (flags & IP_ROUTETOIF) {
436 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
437 						    M_GETFIB(m)))) == NULL &&
438 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
439 						M_GETFIB(m)))) == NULL) {
440 			IPSTAT_INC(ips_noroute);
441 			error = ENETUNREACH;
442 			goto bad;
443 		}
444 		ifp = ia->ia_ifp;
445 		mtu = ifp->if_mtu;
446 		ip->ip_ttl = 1;
447 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
448 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
449 		src = IA_SIN(ia)->sin_addr;
450 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
451 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
452 		/*
453 		 * Bypass the normal routing lookup for multicast
454 		 * packets if the interface is specified.
455 		 */
456 		ifp = imo->imo_multicast_ifp;
457 		mtu = ifp->if_mtu;
458 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
459 		isbroadcast = 0;	/* fool gcc */
460 		/* Interface may have no addresses. */
461 		if (ia != NULL)
462 			src = IA_SIN(ia)->sin_addr;
463 		else
464 			src.s_addr = INADDR_ANY;
465 	} else if (ro != NULL) {
466 		if (ro->ro_nh == NULL) {
467 			/*
468 			 * We want to do any cloning requested by the link
469 			 * layer, as this is probably required in all cases
470 			 * for correct operation (as it is for ARP).
471 			 */
472 			uint32_t flowid;
473 #ifdef RADIX_MPATH
474 			flowid = ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr);
475 #else
476 			flowid = m->m_pkthdr.flowid;
477 #endif
478 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
479 			    NHR_REF, flowid);
480 
481 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
482 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
483 				/*
484 				 * There is no route for this packet, but it is
485 				 * possible that a matching SPD entry exists.
486 				 */
487 				no_route_but_check_spd = 1;
488 				goto sendit;
489 #endif
490 				IPSTAT_INC(ips_noroute);
491 				error = EHOSTUNREACH;
492 				goto bad;
493 			}
494 		}
495 		ia = ifatoia(ro->ro_nh->nh_ifa);
496 		ifp = ro->ro_nh->nh_ifp;
497 		counter_u64_add(ro->ro_nh->nh_pksent, 1);
498 		rt_update_ro_flags(ro);
499 		if (ro->ro_nh->nh_flags & NHF_GATEWAY)
500 			gw = &ro->ro_nh->gw4_sa;
501 		if (ro->ro_nh->nh_flags & NHF_HOST)
502 			isbroadcast = (ro->ro_nh->nh_flags & NHF_BROADCAST);
503 		else if (ifp->if_flags & IFF_BROADCAST)
504 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
505 		else
506 			isbroadcast = 0;
507 		if (ro->ro_nh->nh_flags & NHF_HOST)
508 			mtu = ro->ro_nh->nh_mtu;
509 		else
510 			mtu = ifp->if_mtu;
511 		src = IA_SIN(ia)->sin_addr;
512 	} else {
513 		struct nhop_object *nh;
514 
515 		nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0);
516 		if (nh == NULL) {
517 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
518 			/*
519 			 * There is no route for this packet, but it is
520 			 * possible that a matching SPD entry exists.
521 			 */
522 			no_route_but_check_spd = 1;
523 			goto sendit;
524 #endif
525 			IPSTAT_INC(ips_noroute);
526 			error = EHOSTUNREACH;
527 			goto bad;
528 		}
529 		ifp = nh->nh_ifp;
530 		mtu = nh->nh_mtu;
531 		/*
532 		 * We are rewriting here dst to be gw actually, contradicting
533 		 * comment at the beginning of the function. However, in this
534 		 * case we are always dealing with on stack dst.
535 		 * In case if pfil(9) sends us back to beginning of the
536 		 * function, the dst would be rewritten by ip_output_pfil().
537 		 */
538 		MPASS(dst == &sin);
539 		if (nh->nh_flags & NHF_GATEWAY)
540 			dst->sin_addr = nh->gw4_sa.sin_addr;
541 		ia = ifatoia(nh->nh_ifa);
542 		src = IA_SIN(ia)->sin_addr;
543 		isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
544 		    (NHF_HOST | NHF_BROADCAST)) ||
545 		    ((ifp->if_flags & IFF_BROADCAST) &&
546 		    in_ifaddr_broadcast(dst->sin_addr, ia)));
547 	}
548 
549 	/* Catch a possible divide by zero later. */
550 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
551 	    __func__, mtu, ro,
552 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
553 
554 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
555 		m->m_flags |= M_MCAST;
556 		/*
557 		 * IP destination address is multicast.  Make sure "gw"
558 		 * still points to the address in "ro".  (It may have been
559 		 * changed to point to a gateway address, above.)
560 		 */
561 		gw = dst;
562 		/*
563 		 * See if the caller provided any multicast options
564 		 */
565 		if (imo != NULL) {
566 			ip->ip_ttl = imo->imo_multicast_ttl;
567 			if (imo->imo_multicast_vif != -1)
568 				ip->ip_src.s_addr =
569 				    ip_mcast_src ?
570 				    ip_mcast_src(imo->imo_multicast_vif) :
571 				    INADDR_ANY;
572 		} else
573 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
574 		/*
575 		 * Confirm that the outgoing interface supports multicast.
576 		 */
577 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
578 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
579 				IPSTAT_INC(ips_noroute);
580 				error = ENETUNREACH;
581 				goto bad;
582 			}
583 		}
584 		/*
585 		 * If source address not specified yet, use address
586 		 * of outgoing interface.
587 		 */
588 		if (ip->ip_src.s_addr == INADDR_ANY)
589 			ip->ip_src = src;
590 
591 		if ((imo == NULL && in_mcast_loop) ||
592 		    (imo && imo->imo_multicast_loop)) {
593 			/*
594 			 * Loop back multicast datagram if not expressly
595 			 * forbidden to do so, even if we are not a member
596 			 * of the group; ip_input() will filter it later,
597 			 * thus deferring a hash lookup and mutex acquisition
598 			 * at the expense of a cheap copy using m_copym().
599 			 */
600 			ip_mloopback(ifp, m, hlen);
601 		} else {
602 			/*
603 			 * If we are acting as a multicast router, perform
604 			 * multicast forwarding as if the packet had just
605 			 * arrived on the interface to which we are about
606 			 * to send.  The multicast forwarding function
607 			 * recursively calls this function, using the
608 			 * IP_FORWARDING flag to prevent infinite recursion.
609 			 *
610 			 * Multicasts that are looped back by ip_mloopback(),
611 			 * above, will be forwarded by the ip_input() routine,
612 			 * if necessary.
613 			 */
614 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
615 				/*
616 				 * If rsvp daemon is not running, do not
617 				 * set ip_moptions. This ensures that the packet
618 				 * is multicast and not just sent down one link
619 				 * as prescribed by rsvpd.
620 				 */
621 				if (!V_rsvp_on)
622 					imo = NULL;
623 				if (ip_mforward &&
624 				    ip_mforward(ip, ifp, m, imo) != 0) {
625 					m_freem(m);
626 					goto done;
627 				}
628 			}
629 		}
630 
631 		/*
632 		 * Multicasts with a time-to-live of zero may be looped-
633 		 * back, above, but must not be transmitted on a network.
634 		 * Also, multicasts addressed to the loopback interface
635 		 * are not sent -- the above call to ip_mloopback() will
636 		 * loop back a copy. ip_input() will drop the copy if
637 		 * this host does not belong to the destination group on
638 		 * the loopback interface.
639 		 */
640 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
641 			m_freem(m);
642 			goto done;
643 		}
644 
645 		goto sendit;
646 	}
647 
648 	/*
649 	 * If the source address is not specified yet, use the address
650 	 * of the outoing interface.
651 	 */
652 	if (ip->ip_src.s_addr == INADDR_ANY)
653 		ip->ip_src = src;
654 
655 	/*
656 	 * Look for broadcast address and
657 	 * verify user is allowed to send
658 	 * such a packet.
659 	 */
660 	if (isbroadcast) {
661 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
662 			error = EADDRNOTAVAIL;
663 			goto bad;
664 		}
665 		if ((flags & IP_ALLOWBROADCAST) == 0) {
666 			error = EACCES;
667 			goto bad;
668 		}
669 		/* don't allow broadcast messages to be fragmented */
670 		if (ip_len > mtu) {
671 			error = EMSGSIZE;
672 			goto bad;
673 		}
674 		m->m_flags |= M_BCAST;
675 	} else {
676 		m->m_flags &= ~M_BCAST;
677 	}
678 
679 sendit:
680 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
681 	if (IPSEC_ENABLED(ipv4)) {
682 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
683 			if (error == EINPROGRESS)
684 				error = 0;
685 			goto done;
686 		}
687 	}
688 	/*
689 	 * Check if there was a route for this packet; return error if not.
690 	 */
691 	if (no_route_but_check_spd) {
692 		IPSTAT_INC(ips_noroute);
693 		error = EHOSTUNREACH;
694 		goto bad;
695 	}
696 	/* Update variables that are affected by ipsec4_output(). */
697 	ip = mtod(m, struct ip *);
698 	hlen = ip->ip_hl << 2;
699 #endif /* IPSEC */
700 
701 	/* Jump over all PFIL processing if hooks are not active. */
702 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
703 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
704 		    &error)) {
705 		case 1: /* Finished */
706 			goto done;
707 
708 		case 0: /* Continue normally */
709 			ip = mtod(m, struct ip *);
710 			break;
711 
712 		case -1: /* Need to try again */
713 			/* Reset everything for a new round */
714 			if (ro != NULL) {
715 				RO_NHFREE(ro);
716 				ro->ro_prepend = NULL;
717 			}
718 			gw = dst;
719 			ip = mtod(m, struct ip *);
720 			goto again;
721 		}
722 	}
723 
724 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
725 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
726 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
727 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
728 			IPSTAT_INC(ips_badaddr);
729 			error = EADDRNOTAVAIL;
730 			goto bad;
731 		}
732 	}
733 
734 	m->m_pkthdr.csum_flags |= CSUM_IP;
735 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
736 		m = mb_unmapped_to_ext(m);
737 		if (m == NULL) {
738 			IPSTAT_INC(ips_odropped);
739 			error = ENOBUFS;
740 			goto bad;
741 		}
742 		in_delayed_cksum(m);
743 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
744 	} else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
745 		m = mb_unmapped_to_ext(m);
746 		if (m == NULL) {
747 			IPSTAT_INC(ips_odropped);
748 			error = ENOBUFS;
749 			goto bad;
750 		}
751 	}
752 #if defined(SCTP) || defined(SCTP_SUPPORT)
753 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
754 		m = mb_unmapped_to_ext(m);
755 		if (m == NULL) {
756 			IPSTAT_INC(ips_odropped);
757 			error = ENOBUFS;
758 			goto bad;
759 		}
760 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
761 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
762 	}
763 #endif
764 
765 	/*
766 	 * If small enough for interface, or the interface will take
767 	 * care of the fragmentation for us, we can just send directly.
768 	 * Note that if_vxlan could have requested TSO even though the outer
769 	 * frame is UDP.  It is correct to not fragment such datagrams and
770 	 * instead just pass them on to the driver.
771 	 */
772 	if (ip_len <= mtu ||
773 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
774 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
775 		ip->ip_sum = 0;
776 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
777 			ip->ip_sum = in_cksum(m, hlen);
778 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
779 		}
780 
781 		/*
782 		 * Record statistics for this interface address.
783 		 * With CSUM_TSO the byte/packet count will be slightly
784 		 * incorrect because we count the IP+TCP headers only
785 		 * once instead of for every generated packet.
786 		 */
787 		if (!(flags & IP_FORWARDING) && ia) {
788 			if (m->m_pkthdr.csum_flags &
789 			    (CSUM_TSO | CSUM_INNER_TSO))
790 				counter_u64_add(ia->ia_ifa.ifa_opackets,
791 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
792 			else
793 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
794 
795 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
796 		}
797 #ifdef MBUF_STRESS_TEST
798 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
799 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
800 #endif
801 		/*
802 		 * Reset layer specific mbuf flags
803 		 * to avoid confusing lower layers.
804 		 */
805 		m_clrprotoflags(m);
806 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
807 		error = ip_output_send(inp, ifp, m, gw, ro,
808 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
809 		goto done;
810 	}
811 
812 	/* Balk when DF bit is set or the interface didn't support TSO. */
813 	if ((ip_off & IP_DF) ||
814 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
815 		error = EMSGSIZE;
816 		IPSTAT_INC(ips_cantfrag);
817 		goto bad;
818 	}
819 
820 	/*
821 	 * Too large for interface; fragment if possible. If successful,
822 	 * on return, m will point to a list of packets to be sent.
823 	 */
824 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
825 	if (error)
826 		goto bad;
827 	for (; m; m = m0) {
828 		m0 = m->m_nextpkt;
829 		m->m_nextpkt = 0;
830 		if (error == 0) {
831 			/* Record statistics for this interface address. */
832 			if (ia != NULL) {
833 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
834 				counter_u64_add(ia->ia_ifa.ifa_obytes,
835 				    m->m_pkthdr.len);
836 			}
837 			/*
838 			 * Reset layer specific mbuf flags
839 			 * to avoid confusing upper layers.
840 			 */
841 			m_clrprotoflags(m);
842 
843 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
844 			    mtod(m, struct ip *), NULL);
845 			error = ip_output_send(inp, ifp, m, gw, ro, true);
846 		} else
847 			m_freem(m);
848 	}
849 
850 	if (error == 0)
851 		IPSTAT_INC(ips_fragmented);
852 
853 done:
854 	return (error);
855  bad:
856 	m_freem(m);
857 	goto done;
858 }
859 
860 /*
861  * Create a chain of fragments which fit the given mtu. m_frag points to the
862  * mbuf to be fragmented; on return it points to the chain with the fragments.
863  * Return 0 if no error. If error, m_frag may contain a partially built
864  * chain of fragments that should be freed by the caller.
865  *
866  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
867  */
868 int
869 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
870     u_long if_hwassist_flags)
871 {
872 	int error = 0;
873 	int hlen = ip->ip_hl << 2;
874 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
875 	int off;
876 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
877 	int firstlen;
878 	struct mbuf **mnext;
879 	int nfrags;
880 	uint16_t ip_len, ip_off;
881 
882 	ip_len = ntohs(ip->ip_len);
883 	ip_off = ntohs(ip->ip_off);
884 
885 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
886 		IPSTAT_INC(ips_cantfrag);
887 		return EMSGSIZE;
888 	}
889 
890 	/*
891 	 * Must be able to put at least 8 bytes per fragment.
892 	 */
893 	if (len < 8)
894 		return EMSGSIZE;
895 
896 	/*
897 	 * If the interface will not calculate checksums on
898 	 * fragmented packets, then do it here.
899 	 */
900 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
901 		m0 = mb_unmapped_to_ext(m0);
902 		if (m0 == NULL) {
903 			error = ENOBUFS;
904 			IPSTAT_INC(ips_odropped);
905 			goto done;
906 		}
907 		in_delayed_cksum(m0);
908 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
909 	}
910 #if defined(SCTP) || defined(SCTP_SUPPORT)
911 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
912 		m0 = mb_unmapped_to_ext(m0);
913 		if (m0 == NULL) {
914 			error = ENOBUFS;
915 			IPSTAT_INC(ips_odropped);
916 			goto done;
917 		}
918 		sctp_delayed_cksum(m0, hlen);
919 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
920 	}
921 #endif
922 	if (len > PAGE_SIZE) {
923 		/*
924 		 * Fragment large datagrams such that each segment
925 		 * contains a multiple of PAGE_SIZE amount of data,
926 		 * plus headers. This enables a receiver to perform
927 		 * page-flipping zero-copy optimizations.
928 		 *
929 		 * XXX When does this help given that sender and receiver
930 		 * could have different page sizes, and also mtu could
931 		 * be less than the receiver's page size ?
932 		 */
933 		int newlen;
934 
935 		off = MIN(mtu, m0->m_pkthdr.len);
936 
937 		/*
938 		 * firstlen (off - hlen) must be aligned on an
939 		 * 8-byte boundary
940 		 */
941 		if (off < hlen)
942 			goto smart_frag_failure;
943 		off = ((off - hlen) & ~7) + hlen;
944 		newlen = (~PAGE_MASK) & mtu;
945 		if ((newlen + sizeof (struct ip)) > mtu) {
946 			/* we failed, go back the default */
947 smart_frag_failure:
948 			newlen = len;
949 			off = hlen + len;
950 		}
951 		len = newlen;
952 
953 	} else {
954 		off = hlen + len;
955 	}
956 
957 	firstlen = off - hlen;
958 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
959 
960 	/*
961 	 * Loop through length of segment after first fragment,
962 	 * make new header and copy data of each part and link onto chain.
963 	 * Here, m0 is the original packet, m is the fragment being created.
964 	 * The fragments are linked off the m_nextpkt of the original
965 	 * packet, which after processing serves as the first fragment.
966 	 */
967 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
968 		struct ip *mhip;	/* ip header on the fragment */
969 		struct mbuf *m;
970 		int mhlen = sizeof (struct ip);
971 
972 		m = m_gethdr(M_NOWAIT, MT_DATA);
973 		if (m == NULL) {
974 			error = ENOBUFS;
975 			IPSTAT_INC(ips_odropped);
976 			goto done;
977 		}
978 		/*
979 		 * Make sure the complete packet header gets copied
980 		 * from the originating mbuf to the newly created
981 		 * mbuf. This also ensures that existing firewall
982 		 * classification(s), VLAN tags and so on get copied
983 		 * to the resulting fragmented packet(s):
984 		 */
985 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
986 			m_free(m);
987 			error = ENOBUFS;
988 			IPSTAT_INC(ips_odropped);
989 			goto done;
990 		}
991 		/*
992 		 * In the first mbuf, leave room for the link header, then
993 		 * copy the original IP header including options. The payload
994 		 * goes into an additional mbuf chain returned by m_copym().
995 		 */
996 		m->m_data += max_linkhdr;
997 		mhip = mtod(m, struct ip *);
998 		*mhip = *ip;
999 		if (hlen > sizeof (struct ip)) {
1000 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1001 			mhip->ip_v = IPVERSION;
1002 			mhip->ip_hl = mhlen >> 2;
1003 		}
1004 		m->m_len = mhlen;
1005 		/* XXX do we need to add ip_off below ? */
1006 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1007 		if (off + len >= ip_len)
1008 			len = ip_len - off;
1009 		else
1010 			mhip->ip_off |= IP_MF;
1011 		mhip->ip_len = htons((u_short)(len + mhlen));
1012 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1013 		if (m->m_next == NULL) {	/* copy failed */
1014 			m_free(m);
1015 			error = ENOBUFS;	/* ??? */
1016 			IPSTAT_INC(ips_odropped);
1017 			goto done;
1018 		}
1019 		m->m_pkthdr.len = mhlen + len;
1020 #ifdef MAC
1021 		mac_netinet_fragment(m0, m);
1022 #endif
1023 		mhip->ip_off = htons(mhip->ip_off);
1024 		mhip->ip_sum = 0;
1025 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1026 			mhip->ip_sum = in_cksum(m, mhlen);
1027 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1028 		}
1029 		*mnext = m;
1030 		mnext = &m->m_nextpkt;
1031 	}
1032 	IPSTAT_ADD(ips_ofragments, nfrags);
1033 
1034 	/*
1035 	 * Update first fragment by trimming what's been copied out
1036 	 * and updating header.
1037 	 */
1038 	m_adj(m0, hlen + firstlen - ip_len);
1039 	m0->m_pkthdr.len = hlen + firstlen;
1040 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1041 	ip->ip_off = htons(ip_off | IP_MF);
1042 	ip->ip_sum = 0;
1043 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1044 		ip->ip_sum = in_cksum(m0, hlen);
1045 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1046 	}
1047 
1048 done:
1049 	*m_frag = m0;
1050 	return error;
1051 }
1052 
1053 void
1054 in_delayed_cksum(struct mbuf *m)
1055 {
1056 	struct ip *ip;
1057 	struct udphdr *uh;
1058 	uint16_t cklen, csum, offset;
1059 
1060 	ip = mtod(m, struct ip *);
1061 	offset = ip->ip_hl << 2 ;
1062 
1063 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1064 		/* if udp header is not in the first mbuf copy udplen */
1065 		if (offset + sizeof(struct udphdr) > m->m_len) {
1066 			m_copydata(m, offset + offsetof(struct udphdr,
1067 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1068 			cklen = ntohs(cklen);
1069 		} else {
1070 			uh = (struct udphdr *)mtodo(m, offset);
1071 			cklen = ntohs(uh->uh_ulen);
1072 		}
1073 		csum = in_cksum_skip(m, cklen + offset, offset);
1074 		if (csum == 0)
1075 			csum = 0xffff;
1076 	} else {
1077 		cklen = ntohs(ip->ip_len);
1078 		csum = in_cksum_skip(m, cklen, offset);
1079 	}
1080 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1081 
1082 	if (offset + sizeof(csum) > m->m_len)
1083 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1084 	else
1085 		*(u_short *)mtodo(m, offset) = csum;
1086 }
1087 
1088 /*
1089  * IP socket option processing.
1090  */
1091 int
1092 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1093 {
1094 	struct inpcb *inp = sotoinpcb(so);
1095 	int	error, optval;
1096 #ifdef	RSS
1097 	uint32_t rss_bucket;
1098 	int retval;
1099 #endif
1100 
1101 	error = optval = 0;
1102 	if (sopt->sopt_level != IPPROTO_IP) {
1103 		error = EINVAL;
1104 
1105 		if (sopt->sopt_level == SOL_SOCKET &&
1106 		    sopt->sopt_dir == SOPT_SET) {
1107 			switch (sopt->sopt_name) {
1108 			case SO_REUSEADDR:
1109 				INP_WLOCK(inp);
1110 				if ((so->so_options & SO_REUSEADDR) != 0)
1111 					inp->inp_flags2 |= INP_REUSEADDR;
1112 				else
1113 					inp->inp_flags2 &= ~INP_REUSEADDR;
1114 				INP_WUNLOCK(inp);
1115 				error = 0;
1116 				break;
1117 			case SO_REUSEPORT:
1118 				INP_WLOCK(inp);
1119 				if ((so->so_options & SO_REUSEPORT) != 0)
1120 					inp->inp_flags2 |= INP_REUSEPORT;
1121 				else
1122 					inp->inp_flags2 &= ~INP_REUSEPORT;
1123 				INP_WUNLOCK(inp);
1124 				error = 0;
1125 				break;
1126 			case SO_REUSEPORT_LB:
1127 				INP_WLOCK(inp);
1128 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1129 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1130 				else
1131 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1132 				INP_WUNLOCK(inp);
1133 				error = 0;
1134 				break;
1135 			case SO_SETFIB:
1136 				INP_WLOCK(inp);
1137 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1138 				INP_WUNLOCK(inp);
1139 				error = 0;
1140 				break;
1141 			case SO_MAX_PACING_RATE:
1142 #ifdef RATELIMIT
1143 				INP_WLOCK(inp);
1144 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1145 				INP_WUNLOCK(inp);
1146 				error = 0;
1147 #else
1148 				error = EOPNOTSUPP;
1149 #endif
1150 				break;
1151 			default:
1152 				break;
1153 			}
1154 		}
1155 		return (error);
1156 	}
1157 
1158 	switch (sopt->sopt_dir) {
1159 	case SOPT_SET:
1160 		switch (sopt->sopt_name) {
1161 		case IP_OPTIONS:
1162 #ifdef notyet
1163 		case IP_RETOPTS:
1164 #endif
1165 		{
1166 			struct mbuf *m;
1167 			if (sopt->sopt_valsize > MLEN) {
1168 				error = EMSGSIZE;
1169 				break;
1170 			}
1171 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1172 			if (m == NULL) {
1173 				error = ENOBUFS;
1174 				break;
1175 			}
1176 			m->m_len = sopt->sopt_valsize;
1177 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1178 					    m->m_len);
1179 			if (error) {
1180 				m_free(m);
1181 				break;
1182 			}
1183 			INP_WLOCK(inp);
1184 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1185 			INP_WUNLOCK(inp);
1186 			return (error);
1187 		}
1188 
1189 		case IP_BINDANY:
1190 			if (sopt->sopt_td != NULL) {
1191 				error = priv_check(sopt->sopt_td,
1192 				    PRIV_NETINET_BINDANY);
1193 				if (error)
1194 					break;
1195 			}
1196 			/* FALLTHROUGH */
1197 		case IP_BINDMULTI:
1198 #ifdef	RSS
1199 		case IP_RSS_LISTEN_BUCKET:
1200 #endif
1201 		case IP_TOS:
1202 		case IP_TTL:
1203 		case IP_MINTTL:
1204 		case IP_RECVOPTS:
1205 		case IP_RECVRETOPTS:
1206 		case IP_ORIGDSTADDR:
1207 		case IP_RECVDSTADDR:
1208 		case IP_RECVTTL:
1209 		case IP_RECVIF:
1210 		case IP_ONESBCAST:
1211 		case IP_DONTFRAG:
1212 		case IP_RECVTOS:
1213 		case IP_RECVFLOWID:
1214 #ifdef	RSS
1215 		case IP_RECVRSSBUCKETID:
1216 #endif
1217 			error = sooptcopyin(sopt, &optval, sizeof optval,
1218 					    sizeof optval);
1219 			if (error)
1220 				break;
1221 
1222 			switch (sopt->sopt_name) {
1223 			case IP_TOS:
1224 				inp->inp_ip_tos = optval;
1225 				break;
1226 
1227 			case IP_TTL:
1228 				inp->inp_ip_ttl = optval;
1229 				break;
1230 
1231 			case IP_MINTTL:
1232 				if (optval >= 0 && optval <= MAXTTL)
1233 					inp->inp_ip_minttl = optval;
1234 				else
1235 					error = EINVAL;
1236 				break;
1237 
1238 #define	OPTSET(bit) do {						\
1239 	INP_WLOCK(inp);							\
1240 	if (optval)							\
1241 		inp->inp_flags |= bit;					\
1242 	else								\
1243 		inp->inp_flags &= ~bit;					\
1244 	INP_WUNLOCK(inp);						\
1245 } while (0)
1246 
1247 #define	OPTSET2(bit, val) do {						\
1248 	INP_WLOCK(inp);							\
1249 	if (val)							\
1250 		inp->inp_flags2 |= bit;					\
1251 	else								\
1252 		inp->inp_flags2 &= ~bit;				\
1253 	INP_WUNLOCK(inp);						\
1254 } while (0)
1255 
1256 			case IP_RECVOPTS:
1257 				OPTSET(INP_RECVOPTS);
1258 				break;
1259 
1260 			case IP_RECVRETOPTS:
1261 				OPTSET(INP_RECVRETOPTS);
1262 				break;
1263 
1264 			case IP_RECVDSTADDR:
1265 				OPTSET(INP_RECVDSTADDR);
1266 				break;
1267 
1268 			case IP_ORIGDSTADDR:
1269 				OPTSET2(INP_ORIGDSTADDR, optval);
1270 				break;
1271 
1272 			case IP_RECVTTL:
1273 				OPTSET(INP_RECVTTL);
1274 				break;
1275 
1276 			case IP_RECVIF:
1277 				OPTSET(INP_RECVIF);
1278 				break;
1279 
1280 			case IP_ONESBCAST:
1281 				OPTSET(INP_ONESBCAST);
1282 				break;
1283 			case IP_DONTFRAG:
1284 				OPTSET(INP_DONTFRAG);
1285 				break;
1286 			case IP_BINDANY:
1287 				OPTSET(INP_BINDANY);
1288 				break;
1289 			case IP_RECVTOS:
1290 				OPTSET(INP_RECVTOS);
1291 				break;
1292 			case IP_BINDMULTI:
1293 				OPTSET2(INP_BINDMULTI, optval);
1294 				break;
1295 			case IP_RECVFLOWID:
1296 				OPTSET2(INP_RECVFLOWID, optval);
1297 				break;
1298 #ifdef	RSS
1299 			case IP_RSS_LISTEN_BUCKET:
1300 				if ((optval >= 0) &&
1301 				    (optval < rss_getnumbuckets())) {
1302 					inp->inp_rss_listen_bucket = optval;
1303 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1304 				} else {
1305 					error = EINVAL;
1306 				}
1307 				break;
1308 			case IP_RECVRSSBUCKETID:
1309 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1310 				break;
1311 #endif
1312 			}
1313 			break;
1314 #undef OPTSET
1315 #undef OPTSET2
1316 
1317 		/*
1318 		 * Multicast socket options are processed by the in_mcast
1319 		 * module.
1320 		 */
1321 		case IP_MULTICAST_IF:
1322 		case IP_MULTICAST_VIF:
1323 		case IP_MULTICAST_TTL:
1324 		case IP_MULTICAST_LOOP:
1325 		case IP_ADD_MEMBERSHIP:
1326 		case IP_DROP_MEMBERSHIP:
1327 		case IP_ADD_SOURCE_MEMBERSHIP:
1328 		case IP_DROP_SOURCE_MEMBERSHIP:
1329 		case IP_BLOCK_SOURCE:
1330 		case IP_UNBLOCK_SOURCE:
1331 		case IP_MSFILTER:
1332 		case MCAST_JOIN_GROUP:
1333 		case MCAST_LEAVE_GROUP:
1334 		case MCAST_JOIN_SOURCE_GROUP:
1335 		case MCAST_LEAVE_SOURCE_GROUP:
1336 		case MCAST_BLOCK_SOURCE:
1337 		case MCAST_UNBLOCK_SOURCE:
1338 			error = inp_setmoptions(inp, sopt);
1339 			break;
1340 
1341 		case IP_PORTRANGE:
1342 			error = sooptcopyin(sopt, &optval, sizeof optval,
1343 					    sizeof optval);
1344 			if (error)
1345 				break;
1346 
1347 			INP_WLOCK(inp);
1348 			switch (optval) {
1349 			case IP_PORTRANGE_DEFAULT:
1350 				inp->inp_flags &= ~(INP_LOWPORT);
1351 				inp->inp_flags &= ~(INP_HIGHPORT);
1352 				break;
1353 
1354 			case IP_PORTRANGE_HIGH:
1355 				inp->inp_flags &= ~(INP_LOWPORT);
1356 				inp->inp_flags |= INP_HIGHPORT;
1357 				break;
1358 
1359 			case IP_PORTRANGE_LOW:
1360 				inp->inp_flags &= ~(INP_HIGHPORT);
1361 				inp->inp_flags |= INP_LOWPORT;
1362 				break;
1363 
1364 			default:
1365 				error = EINVAL;
1366 				break;
1367 			}
1368 			INP_WUNLOCK(inp);
1369 			break;
1370 
1371 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1372 		case IP_IPSEC_POLICY:
1373 			if (IPSEC_ENABLED(ipv4)) {
1374 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1375 				break;
1376 			}
1377 			/* FALLTHROUGH */
1378 #endif /* IPSEC */
1379 
1380 		default:
1381 			error = ENOPROTOOPT;
1382 			break;
1383 		}
1384 		break;
1385 
1386 	case SOPT_GET:
1387 		switch (sopt->sopt_name) {
1388 		case IP_OPTIONS:
1389 		case IP_RETOPTS:
1390 			INP_RLOCK(inp);
1391 			if (inp->inp_options) {
1392 				struct mbuf *options;
1393 
1394 				options = m_copym(inp->inp_options, 0,
1395 				    M_COPYALL, M_NOWAIT);
1396 				INP_RUNLOCK(inp);
1397 				if (options != NULL) {
1398 					error = sooptcopyout(sopt,
1399 							     mtod(options, char *),
1400 							     options->m_len);
1401 					m_freem(options);
1402 				} else
1403 					error = ENOMEM;
1404 			} else {
1405 				INP_RUNLOCK(inp);
1406 				sopt->sopt_valsize = 0;
1407 			}
1408 			break;
1409 
1410 		case IP_TOS:
1411 		case IP_TTL:
1412 		case IP_MINTTL:
1413 		case IP_RECVOPTS:
1414 		case IP_RECVRETOPTS:
1415 		case IP_ORIGDSTADDR:
1416 		case IP_RECVDSTADDR:
1417 		case IP_RECVTTL:
1418 		case IP_RECVIF:
1419 		case IP_PORTRANGE:
1420 		case IP_ONESBCAST:
1421 		case IP_DONTFRAG:
1422 		case IP_BINDANY:
1423 		case IP_RECVTOS:
1424 		case IP_BINDMULTI:
1425 		case IP_FLOWID:
1426 		case IP_FLOWTYPE:
1427 		case IP_RECVFLOWID:
1428 #ifdef	RSS
1429 		case IP_RSSBUCKETID:
1430 		case IP_RECVRSSBUCKETID:
1431 #endif
1432 			switch (sopt->sopt_name) {
1433 			case IP_TOS:
1434 				optval = inp->inp_ip_tos;
1435 				break;
1436 
1437 			case IP_TTL:
1438 				optval = inp->inp_ip_ttl;
1439 				break;
1440 
1441 			case IP_MINTTL:
1442 				optval = inp->inp_ip_minttl;
1443 				break;
1444 
1445 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1446 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1447 
1448 			case IP_RECVOPTS:
1449 				optval = OPTBIT(INP_RECVOPTS);
1450 				break;
1451 
1452 			case IP_RECVRETOPTS:
1453 				optval = OPTBIT(INP_RECVRETOPTS);
1454 				break;
1455 
1456 			case IP_RECVDSTADDR:
1457 				optval = OPTBIT(INP_RECVDSTADDR);
1458 				break;
1459 
1460 			case IP_ORIGDSTADDR:
1461 				optval = OPTBIT2(INP_ORIGDSTADDR);
1462 				break;
1463 
1464 			case IP_RECVTTL:
1465 				optval = OPTBIT(INP_RECVTTL);
1466 				break;
1467 
1468 			case IP_RECVIF:
1469 				optval = OPTBIT(INP_RECVIF);
1470 				break;
1471 
1472 			case IP_PORTRANGE:
1473 				if (inp->inp_flags & INP_HIGHPORT)
1474 					optval = IP_PORTRANGE_HIGH;
1475 				else if (inp->inp_flags & INP_LOWPORT)
1476 					optval = IP_PORTRANGE_LOW;
1477 				else
1478 					optval = 0;
1479 				break;
1480 
1481 			case IP_ONESBCAST:
1482 				optval = OPTBIT(INP_ONESBCAST);
1483 				break;
1484 			case IP_DONTFRAG:
1485 				optval = OPTBIT(INP_DONTFRAG);
1486 				break;
1487 			case IP_BINDANY:
1488 				optval = OPTBIT(INP_BINDANY);
1489 				break;
1490 			case IP_RECVTOS:
1491 				optval = OPTBIT(INP_RECVTOS);
1492 				break;
1493 			case IP_FLOWID:
1494 				optval = inp->inp_flowid;
1495 				break;
1496 			case IP_FLOWTYPE:
1497 				optval = inp->inp_flowtype;
1498 				break;
1499 			case IP_RECVFLOWID:
1500 				optval = OPTBIT2(INP_RECVFLOWID);
1501 				break;
1502 #ifdef	RSS
1503 			case IP_RSSBUCKETID:
1504 				retval = rss_hash2bucket(inp->inp_flowid,
1505 				    inp->inp_flowtype,
1506 				    &rss_bucket);
1507 				if (retval == 0)
1508 					optval = rss_bucket;
1509 				else
1510 					error = EINVAL;
1511 				break;
1512 			case IP_RECVRSSBUCKETID:
1513 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1514 				break;
1515 #endif
1516 			case IP_BINDMULTI:
1517 				optval = OPTBIT2(INP_BINDMULTI);
1518 				break;
1519 			}
1520 			error = sooptcopyout(sopt, &optval, sizeof optval);
1521 			break;
1522 
1523 		/*
1524 		 * Multicast socket options are processed by the in_mcast
1525 		 * module.
1526 		 */
1527 		case IP_MULTICAST_IF:
1528 		case IP_MULTICAST_VIF:
1529 		case IP_MULTICAST_TTL:
1530 		case IP_MULTICAST_LOOP:
1531 		case IP_MSFILTER:
1532 			error = inp_getmoptions(inp, sopt);
1533 			break;
1534 
1535 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1536 		case IP_IPSEC_POLICY:
1537 			if (IPSEC_ENABLED(ipv4)) {
1538 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1539 				break;
1540 			}
1541 			/* FALLTHROUGH */
1542 #endif /* IPSEC */
1543 
1544 		default:
1545 			error = ENOPROTOOPT;
1546 			break;
1547 		}
1548 		break;
1549 	}
1550 	return (error);
1551 }
1552 
1553 /*
1554  * Routine called from ip_output() to loop back a copy of an IP multicast
1555  * packet to the input queue of a specified interface.  Note that this
1556  * calls the output routine of the loopback "driver", but with an interface
1557  * pointer that might NOT be a loopback interface -- evil, but easier than
1558  * replicating that code here.
1559  */
1560 static void
1561 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1562 {
1563 	struct ip *ip;
1564 	struct mbuf *copym;
1565 
1566 	/*
1567 	 * Make a deep copy of the packet because we're going to
1568 	 * modify the pack in order to generate checksums.
1569 	 */
1570 	copym = m_dup(m, M_NOWAIT);
1571 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1572 		copym = m_pullup(copym, hlen);
1573 	if (copym != NULL) {
1574 		/* If needed, compute the checksum and mark it as valid. */
1575 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1576 			in_delayed_cksum(copym);
1577 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1578 			copym->m_pkthdr.csum_flags |=
1579 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1580 			copym->m_pkthdr.csum_data = 0xffff;
1581 		}
1582 		/*
1583 		 * We don't bother to fragment if the IP length is greater
1584 		 * than the interface's MTU.  Can this possibly matter?
1585 		 */
1586 		ip = mtod(copym, struct ip *);
1587 		ip->ip_sum = 0;
1588 		ip->ip_sum = in_cksum(copym, hlen);
1589 		if_simloop(ifp, copym, AF_INET, 0);
1590 	}
1591 }
1592