xref: /freebsd/sys/netinet/ip_output.c (revision 6e563a1b608438504d963c2d7c70e50d2e75af46)
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/sdt.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/sysctl.h>
60 #include <sys/ucred.h>
61 
62 #include <net/if.h>
63 #include <net/if_var.h>
64 #include <net/if_vlan_var.h>
65 #include <net/if_llatbl.h>
66 #include <net/ethernet.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 *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 			m_freem(m);
242 			error = EAGAIN;
243 			goto done;
244 		}
245 		/*
246 		 * Always stamp tags that include NIC ktls.
247 		 */
248 		stamp_tag = true;
249 	}
250 #endif
251 #ifdef RATELIMIT
252 	if (inp != NULL && mst == NULL) {
253 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
254 		    (inp->inp_snd_tag != NULL &&
255 		    inp->inp_snd_tag->ifp != ifp))
256 			in_pcboutput_txrtlmt(inp, ifp, m);
257 
258 		if (inp->inp_snd_tag != NULL)
259 			mst = inp->inp_snd_tag;
260 	}
261 #endif
262 	if (stamp_tag && mst != NULL) {
263 		KASSERT(m->m_pkthdr.rcvif == NULL,
264 		    ("trying to add a send tag to a forwarded packet"));
265 		if (mst->ifp != ifp) {
266 			m_freem(m);
267 			error = EAGAIN;
268 			goto done;
269 		}
270 
271 		/* stamp send tag on mbuf */
272 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
273 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
274 	}
275 
276 	error = (*ifp->if_output)(ifp, m, gw, ro);
277 
278 done:
279 	/* Check for route change invalidating send tags. */
280 #ifdef KERN_TLS
281 	if (tls != NULL) {
282 		if (error == EAGAIN)
283 			error = ktls_output_eagain(inp, tls);
284 		ktls_free(tls);
285 	}
286 #endif
287 #ifdef RATELIMIT
288 	if (error == EAGAIN)
289 		in_pcboutput_eagain(inp);
290 #endif
291 	return (error);
292 }
293 
294 /* rte<>ro_flags translation */
295 static inline void
296 rt_update_ro_flags(struct route *ro, const struct nhop_object *nh)
297 {
298 	int nh_flags = nh->nh_flags;
299 
300 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
301 
302 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
303 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
304 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
305 }
306 
307 /*
308  * IP output.  The packet in mbuf chain m contains a skeletal IP
309  * header (with len, off, ttl, proto, tos, src, dst).
310  * The mbuf chain containing the packet will be freed.
311  * The mbuf opt, if present, will not be freed.
312  * If route ro is present and has ro_rt initialized, route lookup would be
313  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
314  * then result of route lookup is stored in ro->ro_rt.
315  *
316  * In the IP forwarding case, the packet will arrive with options already
317  * inserted, so must have a NULL opt pointer.
318  */
319 int
320 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
321     struct ip_moptions *imo, struct inpcb *inp)
322 {
323 	struct ip *ip;
324 	struct ifnet *ifp = NULL;	/* keep compiler happy */
325 	struct mbuf *m0;
326 	int hlen = sizeof (struct ip);
327 	int mtu = 0;
328 	int error = 0;
329 	int vlan_pcp = -1;
330 	struct sockaddr_in *dst;
331 	const struct sockaddr *gw;
332 	struct in_ifaddr *ia = NULL;
333 	struct in_addr src;
334 	int isbroadcast;
335 	uint16_t ip_len, ip_off;
336 	struct route iproute;
337 	uint32_t fibnum;
338 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
339 	int no_route_but_check_spd = 0;
340 #endif
341 
342 	M_ASSERTPKTHDR(m);
343 	NET_EPOCH_ASSERT();
344 
345 	if (inp != NULL) {
346 		INP_LOCK_ASSERT(inp);
347 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
348 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
349 			m->m_pkthdr.flowid = inp->inp_flowid;
350 			M_HASHTYPE_SET(m, inp->inp_flowtype);
351 		}
352 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
353 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
354 			    INP_2PCP_SHIFT;
355 #ifdef NUMA
356 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
357 #endif
358 	}
359 
360 	if (opt) {
361 		int len = 0;
362 		m = ip_insertoptions(m, opt, &len);
363 		if (len != 0)
364 			hlen = len; /* ip->ip_hl is updated above */
365 	}
366 	ip = mtod(m, struct ip *);
367 	ip_len = ntohs(ip->ip_len);
368 	ip_off = ntohs(ip->ip_off);
369 
370 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
371 		ip->ip_v = IPVERSION;
372 		ip->ip_hl = hlen >> 2;
373 		ip_fillid(ip);
374 	} else {
375 		/* Header already set, fetch hlen from there */
376 		hlen = ip->ip_hl << 2;
377 	}
378 	if ((flags & IP_FORWARDING) == 0)
379 		IPSTAT_INC(ips_localout);
380 
381 	/*
382 	 * dst/gw handling:
383 	 *
384 	 * gw is readonly but can point either to dst OR rt_gateway,
385 	 * therefore we need restore gw if we're redoing lookup.
386 	 */
387 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
388 	if (ro == NULL) {
389 		ro = &iproute;
390 		bzero(ro, sizeof (*ro));
391 	}
392 	dst = (struct sockaddr_in *)&ro->ro_dst;
393 	if (ro->ro_nh == NULL) {
394 		dst->sin_family = AF_INET;
395 		dst->sin_len = sizeof(*dst);
396 		dst->sin_addr = ip->ip_dst;
397 	}
398 	gw = (const struct sockaddr *)dst;
399 again:
400 	/*
401 	 * Validate route against routing table additions;
402 	 * a better/more specific route might have been added.
403 	 */
404 	if (inp != NULL && ro->ro_nh != NULL)
405 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
406 	/*
407 	 * If there is a cached route,
408 	 * check that it is to the same destination
409 	 * and is still up.  If not, free it and try again.
410 	 * The address family should also be checked in case of sharing the
411 	 * cache with IPv6.
412 	 * Also check whether routing cache needs invalidation.
413 	 */
414 	if (ro->ro_nh != NULL &&
415 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
416 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
417 		RO_INVALIDATE_CACHE(ro);
418 	ia = NULL;
419 	/*
420 	 * If routing to interface only, short circuit routing lookup.
421 	 * The use of an all-ones broadcast address implies this; an
422 	 * interface is specified by the broadcast address of an interface,
423 	 * or the destination address of a ptp interface.
424 	 */
425 	if (flags & IP_SENDONES) {
426 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
427 						      M_GETFIB(m)))) == NULL &&
428 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
429 						    M_GETFIB(m)))) == NULL) {
430 			IPSTAT_INC(ips_noroute);
431 			error = ENETUNREACH;
432 			goto bad;
433 		}
434 		ip->ip_dst.s_addr = INADDR_BROADCAST;
435 		dst->sin_addr = ip->ip_dst;
436 		ifp = ia->ia_ifp;
437 		mtu = ifp->if_mtu;
438 		ip->ip_ttl = 1;
439 		isbroadcast = 1;
440 		src = IA_SIN(ia)->sin_addr;
441 	} else if (flags & IP_ROUTETOIF) {
442 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
443 						    M_GETFIB(m)))) == NULL &&
444 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
445 						M_GETFIB(m)))) == NULL) {
446 			IPSTAT_INC(ips_noroute);
447 			error = ENETUNREACH;
448 			goto bad;
449 		}
450 		ifp = ia->ia_ifp;
451 		mtu = ifp->if_mtu;
452 		ip->ip_ttl = 1;
453 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
454 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
455 		src = IA_SIN(ia)->sin_addr;
456 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
457 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
458 		/*
459 		 * Bypass the normal routing lookup for multicast
460 		 * packets if the interface is specified.
461 		 */
462 		ifp = imo->imo_multicast_ifp;
463 		mtu = ifp->if_mtu;
464 		IFP_TO_IA(ifp, ia);
465 		isbroadcast = 0;	/* fool gcc */
466 		/* Interface may have no addresses. */
467 		if (ia != NULL)
468 			src = IA_SIN(ia)->sin_addr;
469 		else
470 			src.s_addr = INADDR_ANY;
471 	} else if (ro != &iproute) {
472 		if (ro->ro_nh == NULL) {
473 			/*
474 			 * We want to do any cloning requested by the link
475 			 * layer, as this is probably required in all cases
476 			 * for correct operation (as it is for ARP).
477 			 */
478 			uint32_t flowid;
479 			flowid = m->m_pkthdr.flowid;
480 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
481 			    NHR_REF, flowid);
482 
483 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
484 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
485 				/*
486 				 * There is no route for this packet, but it is
487 				 * possible that a matching SPD entry exists.
488 				 */
489 				no_route_but_check_spd = 1;
490 				goto sendit;
491 #endif
492 				IPSTAT_INC(ips_noroute);
493 				error = EHOSTUNREACH;
494 				goto bad;
495 			}
496 		}
497 		struct nhop_object *nh = ro->ro_nh;
498 
499 		ia = ifatoia(nh->nh_ifa);
500 		ifp = nh->nh_ifp;
501 		counter_u64_add(nh->nh_pksent, 1);
502 		rt_update_ro_flags(ro, nh);
503 		if (nh->nh_flags & NHF_GATEWAY)
504 			gw = &nh->gw_sa;
505 		if (nh->nh_flags & NHF_HOST)
506 			isbroadcast = (nh->nh_flags & NHF_BROADCAST);
507 		else if ((ifp->if_flags & IFF_BROADCAST) && (gw->sa_family == AF_INET))
508 			isbroadcast = in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia);
509 		else
510 			isbroadcast = 0;
511 		mtu = nh->nh_mtu;
512 		src = IA_SIN(ia)->sin_addr;
513 	} else {
514 		struct nhop_object *nh;
515 
516 		nh = fib4_lookup(M_GETFIB(m), dst->sin_addr, 0, NHR_NONE,
517 		    m->m_pkthdr.flowid);
518 		if (nh == NULL) {
519 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
520 			/*
521 			 * There is no route for this packet, but it is
522 			 * possible that a matching SPD entry exists.
523 			 */
524 			no_route_but_check_spd = 1;
525 			goto sendit;
526 #endif
527 			IPSTAT_INC(ips_noroute);
528 			error = EHOSTUNREACH;
529 			goto bad;
530 		}
531 		ifp = nh->nh_ifp;
532 		mtu = nh->nh_mtu;
533 		rt_update_ro_flags(ro, nh);
534 		if (nh->nh_flags & NHF_GATEWAY)
535 			gw = &nh->gw_sa;
536 		ia = ifatoia(nh->nh_ifa);
537 		src = IA_SIN(ia)->sin_addr;
538 		isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
539 		    (NHF_HOST | NHF_BROADCAST)) ||
540 		    ((ifp->if_flags & IFF_BROADCAST) &&
541 		    (gw->sa_family == AF_INET) &&
542 		    in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->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 = (const struct sockaddr *)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 = (const struct sockaddr *)dst;
715 			ip = mtod(m, struct ip *);
716 			goto again;
717 		}
718 	}
719 
720 	if (vlan_pcp > -1)
721 		EVL_APPLY_PRI(m, vlan_pcp);
722 
723 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
724 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
725 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
726 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
727 			IPSTAT_INC(ips_badaddr);
728 			error = EADDRNOTAVAIL;
729 			goto bad;
730 		}
731 	}
732 
733 	/* Ensure the packet data is mapped if the interface requires it. */
734 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
735 		m = mb_unmapped_to_ext(m);
736 		if (m == NULL) {
737 			IPSTAT_INC(ips_odropped);
738 			error = ENOBUFS;
739 			goto bad;
740 		}
741 	}
742 
743 	m->m_pkthdr.csum_flags |= CSUM_IP;
744 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
745 		in_delayed_cksum(m);
746 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
747 	}
748 #if defined(SCTP) || defined(SCTP_SUPPORT)
749 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
750 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
751 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
752 	}
753 #endif
754 
755 	/*
756 	 * If small enough for interface, or the interface will take
757 	 * care of the fragmentation for us, we can just send directly.
758 	 * Note that if_vxlan could have requested TSO even though the outer
759 	 * frame is UDP.  It is correct to not fragment such datagrams and
760 	 * instead just pass them on to the driver.
761 	 */
762 	if (ip_len <= mtu ||
763 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
764 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
765 		ip->ip_sum = 0;
766 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
767 			ip->ip_sum = in_cksum(m, hlen);
768 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
769 		}
770 
771 		/*
772 		 * Record statistics for this interface address.
773 		 * With CSUM_TSO the byte/packet count will be slightly
774 		 * incorrect because we count the IP+TCP headers only
775 		 * once instead of for every generated packet.
776 		 */
777 		if (!(flags & IP_FORWARDING) && ia) {
778 			if (m->m_pkthdr.csum_flags &
779 			    (CSUM_TSO | CSUM_INNER_TSO))
780 				counter_u64_add(ia->ia_ifa.ifa_opackets,
781 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
782 			else
783 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
784 
785 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
786 		}
787 #ifdef MBUF_STRESS_TEST
788 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
789 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
790 #endif
791 		/*
792 		 * Reset layer specific mbuf flags
793 		 * to avoid confusing lower layers.
794 		 */
795 		m_clrprotoflags(m);
796 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
797 		error = ip_output_send(inp, ifp, m, gw, ro,
798 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
799 		goto done;
800 	}
801 
802 	/* Balk when DF bit is set or the interface didn't support TSO. */
803 	if ((ip_off & IP_DF) ||
804 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
805 		error = EMSGSIZE;
806 		IPSTAT_INC(ips_cantfrag);
807 		goto bad;
808 	}
809 
810 	/*
811 	 * Too large for interface; fragment if possible. If successful,
812 	 * on return, m will point to a list of packets to be sent.
813 	 */
814 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
815 	if (error)
816 		goto bad;
817 	for (; m; m = m0) {
818 		m0 = m->m_nextpkt;
819 		m->m_nextpkt = 0;
820 		if (error == 0) {
821 			/* Record statistics for this interface address. */
822 			if (ia != NULL) {
823 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
824 				counter_u64_add(ia->ia_ifa.ifa_obytes,
825 				    m->m_pkthdr.len);
826 			}
827 			/*
828 			 * Reset layer specific mbuf flags
829 			 * to avoid confusing upper layers.
830 			 */
831 			m_clrprotoflags(m);
832 
833 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
834 			    mtod(m, struct ip *), NULL);
835 			error = ip_output_send(inp, ifp, m, gw, ro, true);
836 		} else
837 			m_freem(m);
838 	}
839 
840 	if (error == 0)
841 		IPSTAT_INC(ips_fragmented);
842 
843 done:
844 	return (error);
845  bad:
846 	m_freem(m);
847 	goto done;
848 }
849 
850 /*
851  * Create a chain of fragments which fit the given mtu. m_frag points to the
852  * mbuf to be fragmented; on return it points to the chain with the fragments.
853  * Return 0 if no error. If error, m_frag may contain a partially built
854  * chain of fragments that should be freed by the caller.
855  *
856  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
857  */
858 int
859 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
860     u_long if_hwassist_flags)
861 {
862 	int error = 0;
863 	int hlen = ip->ip_hl << 2;
864 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
865 	int off;
866 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
867 	int firstlen;
868 	struct mbuf **mnext;
869 	int nfrags;
870 	uint16_t ip_len, ip_off;
871 
872 	ip_len = ntohs(ip->ip_len);
873 	ip_off = ntohs(ip->ip_off);
874 
875 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
876 		IPSTAT_INC(ips_cantfrag);
877 		return EMSGSIZE;
878 	}
879 
880 	/*
881 	 * Must be able to put at least 8 bytes per fragment.
882 	 */
883 	if (len < 8)
884 		return EMSGSIZE;
885 
886 	/*
887 	 * If the interface will not calculate checksums on
888 	 * fragmented packets, then do it here.
889 	 */
890 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
891 		in_delayed_cksum(m0);
892 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
893 	}
894 #if defined(SCTP) || defined(SCTP_SUPPORT)
895 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
896 		sctp_delayed_cksum(m0, hlen);
897 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
898 	}
899 #endif
900 	if (len > PAGE_SIZE) {
901 		/*
902 		 * Fragment large datagrams such that each segment
903 		 * contains a multiple of PAGE_SIZE amount of data,
904 		 * plus headers. This enables a receiver to perform
905 		 * page-flipping zero-copy optimizations.
906 		 *
907 		 * XXX When does this help given that sender and receiver
908 		 * could have different page sizes, and also mtu could
909 		 * be less than the receiver's page size ?
910 		 */
911 		int newlen;
912 
913 		off = MIN(mtu, m0->m_pkthdr.len);
914 
915 		/*
916 		 * firstlen (off - hlen) must be aligned on an
917 		 * 8-byte boundary
918 		 */
919 		if (off < hlen)
920 			goto smart_frag_failure;
921 		off = ((off - hlen) & ~7) + hlen;
922 		newlen = (~PAGE_MASK) & mtu;
923 		if ((newlen + sizeof (struct ip)) > mtu) {
924 			/* we failed, go back the default */
925 smart_frag_failure:
926 			newlen = len;
927 			off = hlen + len;
928 		}
929 		len = newlen;
930 
931 	} else {
932 		off = hlen + len;
933 	}
934 
935 	firstlen = off - hlen;
936 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
937 
938 	/*
939 	 * Loop through length of segment after first fragment,
940 	 * make new header and copy data of each part and link onto chain.
941 	 * Here, m0 is the original packet, m is the fragment being created.
942 	 * The fragments are linked off the m_nextpkt of the original
943 	 * packet, which after processing serves as the first fragment.
944 	 */
945 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
946 		struct ip *mhip;	/* ip header on the fragment */
947 		struct mbuf *m;
948 		int mhlen = sizeof (struct ip);
949 
950 		m = m_gethdr(M_NOWAIT, MT_DATA);
951 		if (m == NULL) {
952 			error = ENOBUFS;
953 			IPSTAT_INC(ips_odropped);
954 			goto done;
955 		}
956 		/*
957 		 * Make sure the complete packet header gets copied
958 		 * from the originating mbuf to the newly created
959 		 * mbuf. This also ensures that existing firewall
960 		 * classification(s), VLAN tags and so on get copied
961 		 * to the resulting fragmented packet(s):
962 		 */
963 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
964 			m_free(m);
965 			error = ENOBUFS;
966 			IPSTAT_INC(ips_odropped);
967 			goto done;
968 		}
969 		/*
970 		 * In the first mbuf, leave room for the link header, then
971 		 * copy the original IP header including options. The payload
972 		 * goes into an additional mbuf chain returned by m_copym().
973 		 */
974 		m->m_data += max_linkhdr;
975 		mhip = mtod(m, struct ip *);
976 		*mhip = *ip;
977 		if (hlen > sizeof (struct ip)) {
978 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
979 			mhip->ip_v = IPVERSION;
980 			mhip->ip_hl = mhlen >> 2;
981 		}
982 		m->m_len = mhlen;
983 		/* XXX do we need to add ip_off below ? */
984 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
985 		if (off + len >= ip_len)
986 			len = ip_len - off;
987 		else
988 			mhip->ip_off |= IP_MF;
989 		mhip->ip_len = htons((u_short)(len + mhlen));
990 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
991 		if (m->m_next == NULL) {	/* copy failed */
992 			m_free(m);
993 			error = ENOBUFS;	/* ??? */
994 			IPSTAT_INC(ips_odropped);
995 			goto done;
996 		}
997 		m->m_pkthdr.len = mhlen + len;
998 #ifdef MAC
999 		mac_netinet_fragment(m0, m);
1000 #endif
1001 		mhip->ip_off = htons(mhip->ip_off);
1002 		mhip->ip_sum = 0;
1003 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1004 			mhip->ip_sum = in_cksum(m, mhlen);
1005 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1006 		}
1007 		*mnext = m;
1008 		mnext = &m->m_nextpkt;
1009 	}
1010 	IPSTAT_ADD(ips_ofragments, nfrags);
1011 
1012 	/*
1013 	 * Update first fragment by trimming what's been copied out
1014 	 * and updating header.
1015 	 */
1016 	m_adj(m0, hlen + firstlen - ip_len);
1017 	m0->m_pkthdr.len = hlen + firstlen;
1018 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1019 	ip->ip_off = htons(ip_off | IP_MF);
1020 	ip->ip_sum = 0;
1021 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1022 		ip->ip_sum = in_cksum(m0, hlen);
1023 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1024 	}
1025 
1026 done:
1027 	*m_frag = m0;
1028 	return error;
1029 }
1030 
1031 void
1032 in_delayed_cksum(struct mbuf *m)
1033 {
1034 	struct ip *ip;
1035 	struct udphdr *uh;
1036 	uint16_t cklen, csum, offset;
1037 
1038 	ip = mtod(m, struct ip *);
1039 	offset = ip->ip_hl << 2 ;
1040 
1041 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1042 		/* if udp header is not in the first mbuf copy udplen */
1043 		if (offset + sizeof(struct udphdr) > m->m_len) {
1044 			m_copydata(m, offset + offsetof(struct udphdr,
1045 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1046 			cklen = ntohs(cklen);
1047 		} else {
1048 			uh = (struct udphdr *)mtodo(m, offset);
1049 			cklen = ntohs(uh->uh_ulen);
1050 		}
1051 		csum = in_cksum_skip(m, cklen + offset, offset);
1052 		if (csum == 0)
1053 			csum = 0xffff;
1054 	} else {
1055 		cklen = ntohs(ip->ip_len);
1056 		csum = in_cksum_skip(m, cklen, offset);
1057 	}
1058 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1059 
1060 	if (offset + sizeof(csum) > m->m_len)
1061 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1062 	else
1063 		*(u_short *)mtodo(m, offset) = csum;
1064 }
1065 
1066 /*
1067  * IP socket option processing.
1068  */
1069 int
1070 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1071 {
1072 	struct inpcb *inp = sotoinpcb(so);
1073 	int	error, optval;
1074 #ifdef	RSS
1075 	uint32_t rss_bucket;
1076 	int retval;
1077 #endif
1078 
1079 	error = optval = 0;
1080 	if (sopt->sopt_level != IPPROTO_IP) {
1081 		error = EINVAL;
1082 
1083 		if (sopt->sopt_level == SOL_SOCKET &&
1084 		    sopt->sopt_dir == SOPT_SET) {
1085 			switch (sopt->sopt_name) {
1086 			case SO_REUSEADDR:
1087 				INP_WLOCK(inp);
1088 				if ((so->so_options & SO_REUSEADDR) != 0)
1089 					inp->inp_flags2 |= INP_REUSEADDR;
1090 				else
1091 					inp->inp_flags2 &= ~INP_REUSEADDR;
1092 				INP_WUNLOCK(inp);
1093 				error = 0;
1094 				break;
1095 			case SO_REUSEPORT:
1096 				INP_WLOCK(inp);
1097 				if ((so->so_options & SO_REUSEPORT) != 0)
1098 					inp->inp_flags2 |= INP_REUSEPORT;
1099 				else
1100 					inp->inp_flags2 &= ~INP_REUSEPORT;
1101 				INP_WUNLOCK(inp);
1102 				error = 0;
1103 				break;
1104 			case SO_REUSEPORT_LB:
1105 				INP_WLOCK(inp);
1106 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1107 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1108 				else
1109 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1110 				INP_WUNLOCK(inp);
1111 				error = 0;
1112 				break;
1113 			case SO_SETFIB:
1114 				INP_WLOCK(inp);
1115 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1116 				INP_WUNLOCK(inp);
1117 				error = 0;
1118 				break;
1119 			case SO_MAX_PACING_RATE:
1120 #ifdef RATELIMIT
1121 				INP_WLOCK(inp);
1122 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1123 				INP_WUNLOCK(inp);
1124 				error = 0;
1125 #else
1126 				error = EOPNOTSUPP;
1127 #endif
1128 				break;
1129 			default:
1130 				break;
1131 			}
1132 		}
1133 		return (error);
1134 	}
1135 
1136 	switch (sopt->sopt_dir) {
1137 	case SOPT_SET:
1138 		switch (sopt->sopt_name) {
1139 		case IP_OPTIONS:
1140 #ifdef notyet
1141 		case IP_RETOPTS:
1142 #endif
1143 		{
1144 			struct mbuf *m;
1145 			if (sopt->sopt_valsize > MLEN) {
1146 				error = EMSGSIZE;
1147 				break;
1148 			}
1149 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1150 			if (m == NULL) {
1151 				error = ENOBUFS;
1152 				break;
1153 			}
1154 			m->m_len = sopt->sopt_valsize;
1155 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1156 					    m->m_len);
1157 			if (error) {
1158 				m_free(m);
1159 				break;
1160 			}
1161 			INP_WLOCK(inp);
1162 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1163 			INP_WUNLOCK(inp);
1164 			return (error);
1165 		}
1166 
1167 		case IP_BINDANY:
1168 			if (sopt->sopt_td != NULL) {
1169 				error = priv_check(sopt->sopt_td,
1170 				    PRIV_NETINET_BINDANY);
1171 				if (error)
1172 					break;
1173 			}
1174 			/* FALLTHROUGH */
1175 		case IP_BINDMULTI:
1176 #ifdef	RSS
1177 		case IP_RSS_LISTEN_BUCKET:
1178 #endif
1179 		case IP_TOS:
1180 		case IP_TTL:
1181 		case IP_MINTTL:
1182 		case IP_RECVOPTS:
1183 		case IP_RECVRETOPTS:
1184 		case IP_ORIGDSTADDR:
1185 		case IP_RECVDSTADDR:
1186 		case IP_RECVTTL:
1187 		case IP_RECVIF:
1188 		case IP_ONESBCAST:
1189 		case IP_DONTFRAG:
1190 		case IP_RECVTOS:
1191 		case IP_RECVFLOWID:
1192 #ifdef	RSS
1193 		case IP_RECVRSSBUCKETID:
1194 #endif
1195 		case IP_VLAN_PCP:
1196 			error = sooptcopyin(sopt, &optval, sizeof optval,
1197 					    sizeof optval);
1198 			if (error)
1199 				break;
1200 
1201 			switch (sopt->sopt_name) {
1202 			case IP_TOS:
1203 				inp->inp_ip_tos = optval;
1204 				break;
1205 
1206 			case IP_TTL:
1207 				inp->inp_ip_ttl = optval;
1208 				break;
1209 
1210 			case IP_MINTTL:
1211 				if (optval >= 0 && optval <= MAXTTL)
1212 					inp->inp_ip_minttl = optval;
1213 				else
1214 					error = EINVAL;
1215 				break;
1216 
1217 #define	OPTSET(bit) do {						\
1218 	INP_WLOCK(inp);							\
1219 	if (optval)							\
1220 		inp->inp_flags |= bit;					\
1221 	else								\
1222 		inp->inp_flags &= ~bit;					\
1223 	INP_WUNLOCK(inp);						\
1224 } while (0)
1225 
1226 #define	OPTSET2(bit, val) do {						\
1227 	INP_WLOCK(inp);							\
1228 	if (val)							\
1229 		inp->inp_flags2 |= bit;					\
1230 	else								\
1231 		inp->inp_flags2 &= ~bit;				\
1232 	INP_WUNLOCK(inp);						\
1233 } while (0)
1234 
1235 			case IP_RECVOPTS:
1236 				OPTSET(INP_RECVOPTS);
1237 				break;
1238 
1239 			case IP_RECVRETOPTS:
1240 				OPTSET(INP_RECVRETOPTS);
1241 				break;
1242 
1243 			case IP_RECVDSTADDR:
1244 				OPTSET(INP_RECVDSTADDR);
1245 				break;
1246 
1247 			case IP_ORIGDSTADDR:
1248 				OPTSET2(INP_ORIGDSTADDR, optval);
1249 				break;
1250 
1251 			case IP_RECVTTL:
1252 				OPTSET(INP_RECVTTL);
1253 				break;
1254 
1255 			case IP_RECVIF:
1256 				OPTSET(INP_RECVIF);
1257 				break;
1258 
1259 			case IP_ONESBCAST:
1260 				OPTSET(INP_ONESBCAST);
1261 				break;
1262 			case IP_DONTFRAG:
1263 				OPTSET(INP_DONTFRAG);
1264 				break;
1265 			case IP_BINDANY:
1266 				OPTSET(INP_BINDANY);
1267 				break;
1268 			case IP_RECVTOS:
1269 				OPTSET(INP_RECVTOS);
1270 				break;
1271 			case IP_BINDMULTI:
1272 				OPTSET2(INP_BINDMULTI, optval);
1273 				break;
1274 			case IP_RECVFLOWID:
1275 				OPTSET2(INP_RECVFLOWID, optval);
1276 				break;
1277 #ifdef	RSS
1278 			case IP_RSS_LISTEN_BUCKET:
1279 				if ((optval >= 0) &&
1280 				    (optval < rss_getnumbuckets())) {
1281 					inp->inp_rss_listen_bucket = optval;
1282 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1283 				} else {
1284 					error = EINVAL;
1285 				}
1286 				break;
1287 			case IP_RECVRSSBUCKETID:
1288 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1289 				break;
1290 #endif
1291 			case IP_VLAN_PCP:
1292 				if ((optval >= -1) && (optval <=
1293 				    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1294 					if (optval == -1) {
1295 						INP_WLOCK(inp);
1296 						inp->inp_flags2 &=
1297 						    ~(INP_2PCP_SET |
1298 						      INP_2PCP_MASK);
1299 						INP_WUNLOCK(inp);
1300 					} else {
1301 						INP_WLOCK(inp);
1302 						inp->inp_flags2 |=
1303 						    INP_2PCP_SET;
1304 						inp->inp_flags2 &=
1305 						    ~INP_2PCP_MASK;
1306 						inp->inp_flags2 |=
1307 						    optval << INP_2PCP_SHIFT;
1308 						INP_WUNLOCK(inp);
1309 					}
1310 				} else
1311 					error = EINVAL;
1312 				break;
1313 			}
1314 			break;
1315 #undef OPTSET
1316 #undef OPTSET2
1317 
1318 		/*
1319 		 * Multicast socket options are processed by the in_mcast
1320 		 * module.
1321 		 */
1322 		case IP_MULTICAST_IF:
1323 		case IP_MULTICAST_VIF:
1324 		case IP_MULTICAST_TTL:
1325 		case IP_MULTICAST_LOOP:
1326 		case IP_ADD_MEMBERSHIP:
1327 		case IP_DROP_MEMBERSHIP:
1328 		case IP_ADD_SOURCE_MEMBERSHIP:
1329 		case IP_DROP_SOURCE_MEMBERSHIP:
1330 		case IP_BLOCK_SOURCE:
1331 		case IP_UNBLOCK_SOURCE:
1332 		case IP_MSFILTER:
1333 		case MCAST_JOIN_GROUP:
1334 		case MCAST_LEAVE_GROUP:
1335 		case MCAST_JOIN_SOURCE_GROUP:
1336 		case MCAST_LEAVE_SOURCE_GROUP:
1337 		case MCAST_BLOCK_SOURCE:
1338 		case MCAST_UNBLOCK_SOURCE:
1339 			error = inp_setmoptions(inp, sopt);
1340 			break;
1341 
1342 		case IP_PORTRANGE:
1343 			error = sooptcopyin(sopt, &optval, sizeof optval,
1344 					    sizeof optval);
1345 			if (error)
1346 				break;
1347 
1348 			INP_WLOCK(inp);
1349 			switch (optval) {
1350 			case IP_PORTRANGE_DEFAULT:
1351 				inp->inp_flags &= ~(INP_LOWPORT);
1352 				inp->inp_flags &= ~(INP_HIGHPORT);
1353 				break;
1354 
1355 			case IP_PORTRANGE_HIGH:
1356 				inp->inp_flags &= ~(INP_LOWPORT);
1357 				inp->inp_flags |= INP_HIGHPORT;
1358 				break;
1359 
1360 			case IP_PORTRANGE_LOW:
1361 				inp->inp_flags &= ~(INP_HIGHPORT);
1362 				inp->inp_flags |= INP_LOWPORT;
1363 				break;
1364 
1365 			default:
1366 				error = EINVAL;
1367 				break;
1368 			}
1369 			INP_WUNLOCK(inp);
1370 			break;
1371 
1372 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1373 		case IP_IPSEC_POLICY:
1374 			if (IPSEC_ENABLED(ipv4)) {
1375 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1376 				break;
1377 			}
1378 			/* FALLTHROUGH */
1379 #endif /* IPSEC */
1380 
1381 		default:
1382 			error = ENOPROTOOPT;
1383 			break;
1384 		}
1385 		break;
1386 
1387 	case SOPT_GET:
1388 		switch (sopt->sopt_name) {
1389 		case IP_OPTIONS:
1390 		case IP_RETOPTS:
1391 			INP_RLOCK(inp);
1392 			if (inp->inp_options) {
1393 				struct mbuf *options;
1394 
1395 				options = m_copym(inp->inp_options, 0,
1396 				    M_COPYALL, M_NOWAIT);
1397 				INP_RUNLOCK(inp);
1398 				if (options != NULL) {
1399 					error = sooptcopyout(sopt,
1400 							     mtod(options, char *),
1401 							     options->m_len);
1402 					m_freem(options);
1403 				} else
1404 					error = ENOMEM;
1405 			} else {
1406 				INP_RUNLOCK(inp);
1407 				sopt->sopt_valsize = 0;
1408 			}
1409 			break;
1410 
1411 		case IP_TOS:
1412 		case IP_TTL:
1413 		case IP_MINTTL:
1414 		case IP_RECVOPTS:
1415 		case IP_RECVRETOPTS:
1416 		case IP_ORIGDSTADDR:
1417 		case IP_RECVDSTADDR:
1418 		case IP_RECVTTL:
1419 		case IP_RECVIF:
1420 		case IP_PORTRANGE:
1421 		case IP_ONESBCAST:
1422 		case IP_DONTFRAG:
1423 		case IP_BINDANY:
1424 		case IP_RECVTOS:
1425 		case IP_BINDMULTI:
1426 		case IP_FLOWID:
1427 		case IP_FLOWTYPE:
1428 		case IP_RECVFLOWID:
1429 #ifdef	RSS
1430 		case IP_RSSBUCKETID:
1431 		case IP_RECVRSSBUCKETID:
1432 #endif
1433 		case IP_VLAN_PCP:
1434 			switch (sopt->sopt_name) {
1435 			case IP_TOS:
1436 				optval = inp->inp_ip_tos;
1437 				break;
1438 
1439 			case IP_TTL:
1440 				optval = inp->inp_ip_ttl;
1441 				break;
1442 
1443 			case IP_MINTTL:
1444 				optval = inp->inp_ip_minttl;
1445 				break;
1446 
1447 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1448 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1449 
1450 			case IP_RECVOPTS:
1451 				optval = OPTBIT(INP_RECVOPTS);
1452 				break;
1453 
1454 			case IP_RECVRETOPTS:
1455 				optval = OPTBIT(INP_RECVRETOPTS);
1456 				break;
1457 
1458 			case IP_RECVDSTADDR:
1459 				optval = OPTBIT(INP_RECVDSTADDR);
1460 				break;
1461 
1462 			case IP_ORIGDSTADDR:
1463 				optval = OPTBIT2(INP_ORIGDSTADDR);
1464 				break;
1465 
1466 			case IP_RECVTTL:
1467 				optval = OPTBIT(INP_RECVTTL);
1468 				break;
1469 
1470 			case IP_RECVIF:
1471 				optval = OPTBIT(INP_RECVIF);
1472 				break;
1473 
1474 			case IP_PORTRANGE:
1475 				if (inp->inp_flags & INP_HIGHPORT)
1476 					optval = IP_PORTRANGE_HIGH;
1477 				else if (inp->inp_flags & INP_LOWPORT)
1478 					optval = IP_PORTRANGE_LOW;
1479 				else
1480 					optval = 0;
1481 				break;
1482 
1483 			case IP_ONESBCAST:
1484 				optval = OPTBIT(INP_ONESBCAST);
1485 				break;
1486 			case IP_DONTFRAG:
1487 				optval = OPTBIT(INP_DONTFRAG);
1488 				break;
1489 			case IP_BINDANY:
1490 				optval = OPTBIT(INP_BINDANY);
1491 				break;
1492 			case IP_RECVTOS:
1493 				optval = OPTBIT(INP_RECVTOS);
1494 				break;
1495 			case IP_FLOWID:
1496 				optval = inp->inp_flowid;
1497 				break;
1498 			case IP_FLOWTYPE:
1499 				optval = inp->inp_flowtype;
1500 				break;
1501 			case IP_RECVFLOWID:
1502 				optval = OPTBIT2(INP_RECVFLOWID);
1503 				break;
1504 #ifdef	RSS
1505 			case IP_RSSBUCKETID:
1506 				retval = rss_hash2bucket(inp->inp_flowid,
1507 				    inp->inp_flowtype,
1508 				    &rss_bucket);
1509 				if (retval == 0)
1510 					optval = rss_bucket;
1511 				else
1512 					error = EINVAL;
1513 				break;
1514 			case IP_RECVRSSBUCKETID:
1515 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1516 				break;
1517 #endif
1518 			case IP_BINDMULTI:
1519 				optval = OPTBIT2(INP_BINDMULTI);
1520 				break;
1521 			case IP_VLAN_PCP:
1522 				if (OPTBIT2(INP_2PCP_SET)) {
1523 					optval = (inp->inp_flags2 &
1524 					    INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1525 				} else {
1526 					optval = -1;
1527 				}
1528 				break;
1529 			}
1530 			error = sooptcopyout(sopt, &optval, sizeof optval);
1531 			break;
1532 
1533 		/*
1534 		 * Multicast socket options are processed by the in_mcast
1535 		 * module.
1536 		 */
1537 		case IP_MULTICAST_IF:
1538 		case IP_MULTICAST_VIF:
1539 		case IP_MULTICAST_TTL:
1540 		case IP_MULTICAST_LOOP:
1541 		case IP_MSFILTER:
1542 			error = inp_getmoptions(inp, sopt);
1543 			break;
1544 
1545 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1546 		case IP_IPSEC_POLICY:
1547 			if (IPSEC_ENABLED(ipv4)) {
1548 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1549 				break;
1550 			}
1551 			/* FALLTHROUGH */
1552 #endif /* IPSEC */
1553 
1554 		default:
1555 			error = ENOPROTOOPT;
1556 			break;
1557 		}
1558 		break;
1559 	}
1560 	return (error);
1561 }
1562 
1563 /*
1564  * Routine called from ip_output() to loop back a copy of an IP multicast
1565  * packet to the input queue of a specified interface.  Note that this
1566  * calls the output routine of the loopback "driver", but with an interface
1567  * pointer that might NOT be a loopback interface -- evil, but easier than
1568  * replicating that code here.
1569  */
1570 static void
1571 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1572 {
1573 	struct ip *ip;
1574 	struct mbuf *copym;
1575 
1576 	/*
1577 	 * Make a deep copy of the packet because we're going to
1578 	 * modify the pack in order to generate checksums.
1579 	 */
1580 	copym = m_dup(m, M_NOWAIT);
1581 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1582 		copym = m_pullup(copym, hlen);
1583 	if (copym != NULL) {
1584 		/* If needed, compute the checksum and mark it as valid. */
1585 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1586 			in_delayed_cksum(copym);
1587 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1588 			copym->m_pkthdr.csum_flags |=
1589 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1590 			copym->m_pkthdr.csum_data = 0xffff;
1591 		}
1592 		/*
1593 		 * We don't bother to fragment if the IP length is greater
1594 		 * than the interface's MTU.  Can this possibly matter?
1595 		 */
1596 		ip = mtod(copym, struct ip *);
1597 		ip->ip_sum = 0;
1598 		ip->ip_sum = in_cksum(copym, hlen);
1599 		if_simloop(ifp, copym, AF_INET, 0);
1600 	}
1601 }
1602