xref: /freebsd/sys/netinet/ip_output.c (revision 97ca2ada80b870edbbb4f66b26e274cf8e55e0bc)
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 
32 #include <sys/cdefs.h>
33 #include "opt_inet.h"
34 #include "opt_ipsec.h"
35 #include "opt_kern_tls.h"
36 #include "opt_mbuf_stress_test.h"
37 #include "opt_ratelimit.h"
38 #include "opt_route.h"
39 #include "opt_rss.h"
40 #include "opt_sctp.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/ktls.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mbuf.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/protosw.h>
52 #include <sys/sdt.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sysctl.h>
56 #include <sys/ucred.h>
57 
58 #include <net/if.h>
59 #include <net/if_var.h>
60 #include <net/if_private.h>
61 #include <net/if_vlan_var.h>
62 #include <net/if_llatbl.h>
63 #include <net/ethernet.h>
64 #include <net/netisr.h>
65 #include <net/pfil.h>
66 #include <net/route.h>
67 #include <net/route/nhop.h>
68 #include <net/rss_config.h>
69 #include <net/vnet.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_fib.h>
73 #include <netinet/in_kdtrace.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/ip.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/in_rss.h>
79 #include <netinet/in_var.h>
80 #include <netinet/ip_var.h>
81 #include <netinet/ip_options.h>
82 
83 #include <netinet/udp.h>
84 #include <netinet/udp_var.h>
85 
86 #if defined(SCTP) || defined(SCTP_SUPPORT)
87 #include <netinet/sctp.h>
88 #include <netinet/sctp_crc32.h>
89 #endif
90 
91 #include <netipsec/ipsec_support.h>
92 
93 #include <machine/in_cksum.h>
94 
95 #include <security/mac/mac_framework.h>
96 
97 #ifdef MBUF_STRESS_TEST
98 static int mbuf_frag_size = 0;
99 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
100 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
101 #endif
102 
103 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
104 
105 extern int in_mcast_loop;
106 
107 static inline int
108 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
109     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
110 {
111 	struct m_tag *fwd_tag = NULL;
112 	struct mbuf *m;
113 	struct in_addr odst;
114 	struct ip *ip;
115 	int ret;
116 
117 	m = *mp;
118 	ip = mtod(m, struct ip *);
119 
120 	/* Run through list of hooks for output packets. */
121 	odst.s_addr = ip->ip_dst.s_addr;
122 	if (flags & IP_FORWARDING)
123 		ret = pfil_mbuf_fwd(V_inet_pfil_head, mp, ifp, inp);
124 	else
125 		ret = pfil_mbuf_out(V_inet_pfil_head, mp, ifp, inp);
126 
127 	switch (ret) {
128 	case PFIL_DROPPED:
129 		*error = EACCES;
130 		/* FALLTHROUGH */
131 	case PFIL_CONSUMED:
132 		return 1; /* Finished */
133 	case PFIL_PASS:
134 		*error = 0;
135 	}
136 	m = *mp;
137 	ip = mtod(m, struct ip *);
138 
139 	/* See if destination IP address was changed by packet filter. */
140 	if (odst.s_addr != ip->ip_dst.s_addr) {
141 		m->m_flags |= M_SKIP_FIREWALL;
142 		/* If destination is now ourself drop to ip_input(). */
143 		if (in_localip(ip->ip_dst)) {
144 			m->m_flags |= M_FASTFWD_OURS;
145 			if (m->m_pkthdr.rcvif == NULL)
146 				m->m_pkthdr.rcvif = V_loif;
147 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
148 				m->m_pkthdr.csum_flags |=
149 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
150 				m->m_pkthdr.csum_data = 0xffff;
151 			}
152 			m->m_pkthdr.csum_flags |=
153 				CSUM_IP_CHECKED | CSUM_IP_VALID;
154 #if defined(SCTP) || defined(SCTP_SUPPORT)
155 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
156 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
157 #endif
158 			*error = netisr_queue(NETISR_IP, m);
159 			return 1; /* Finished */
160 		}
161 
162 		bzero(dst, sizeof(*dst));
163 		dst->sin_family = AF_INET;
164 		dst->sin_len = sizeof(*dst);
165 		dst->sin_addr = ip->ip_dst;
166 
167 		return -1; /* Reloop */
168 	}
169 	/* See if fib was changed by packet filter. */
170 	if ((*fibnum) != M_GETFIB(m)) {
171 		m->m_flags |= M_SKIP_FIREWALL;
172 		*fibnum = M_GETFIB(m);
173 		return -1; /* Reloop for FIB change */
174 	}
175 
176 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
177 	if (m->m_flags & M_FASTFWD_OURS) {
178 		if (m->m_pkthdr.rcvif == NULL)
179 			m->m_pkthdr.rcvif = V_loif;
180 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
181 			m->m_pkthdr.csum_flags |=
182 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
183 			m->m_pkthdr.csum_data = 0xffff;
184 		}
185 #if defined(SCTP) || defined(SCTP_SUPPORT)
186 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
187 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
188 #endif
189 		m->m_pkthdr.csum_flags |=
190 			CSUM_IP_CHECKED | CSUM_IP_VALID;
191 
192 		*error = netisr_queue(NETISR_IP, m);
193 		return 1; /* Finished */
194 	}
195 	/* Or forward to some other address? */
196 	if ((m->m_flags & M_IP_NEXTHOP) &&
197 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
198 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
199 		m->m_flags |= M_SKIP_FIREWALL;
200 		m->m_flags &= ~M_IP_NEXTHOP;
201 		m_tag_delete(m, fwd_tag);
202 
203 		return -1; /* Reloop for CHANGE of dst */
204 	}
205 
206 	return 0;
207 }
208 
209 static int
210 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
211     const struct sockaddr *gw, struct route *ro, bool stamp_tag)
212 {
213 #ifdef KERN_TLS
214 	struct ktls_session *tls = NULL;
215 #endif
216 	struct m_snd_tag *mst;
217 	int error;
218 
219 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
220 	mst = NULL;
221 
222 #ifdef KERN_TLS
223 	/*
224 	 * If this is an unencrypted TLS record, save a reference to
225 	 * the record.  This local reference is used to call
226 	 * ktls_output_eagain after the mbuf has been freed (thus
227 	 * dropping the mbuf's reference) in if_output.
228 	 */
229 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
230 		tls = ktls_hold(m->m_next->m_epg_tls);
231 		mst = tls->snd_tag;
232 
233 		/*
234 		 * If a TLS session doesn't have a valid tag, it must
235 		 * have had an earlier ifp mismatch, so drop this
236 		 * packet.
237 		 */
238 		if (mst == NULL) {
239 			m_freem(m);
240 			error = EAGAIN;
241 			goto done;
242 		}
243 		/*
244 		 * Always stamp tags that include NIC ktls.
245 		 */
246 		stamp_tag = true;
247 	}
248 #endif
249 #ifdef RATELIMIT
250 	if (inp != NULL && mst == NULL) {
251 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
252 		    (inp->inp_snd_tag != NULL &&
253 		    inp->inp_snd_tag->ifp != ifp))
254 			in_pcboutput_txrtlmt(inp, ifp, m);
255 
256 		if (inp->inp_snd_tag != NULL)
257 			mst = inp->inp_snd_tag;
258 	}
259 #endif
260 	if (stamp_tag && mst != NULL) {
261 		KASSERT(m->m_pkthdr.rcvif == NULL,
262 		    ("trying to add a send tag to a forwarded packet"));
263 		if (mst->ifp != ifp) {
264 			m_freem(m);
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, 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, const struct nhop_object *nh)
295 {
296 	int nh_flags = 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 ip *ip;
322 	struct ifnet *ifp = NULL;	/* keep compiler happy */
323 	struct mbuf *m0;
324 	int hlen = sizeof (struct ip);
325 	int mtu = 0;
326 	int error = 0;
327 	int vlan_pcp = -1;
328 	struct sockaddr_in *dst;
329 	const struct sockaddr *gw;
330 	struct in_ifaddr *ia = NULL;
331 	struct in_addr src;
332 	bool isbroadcast;
333 	uint16_t ip_len, ip_off;
334 	struct route iproute;
335 	uint32_t fibnum;
336 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
337 	int no_route_but_check_spd = 0;
338 #endif
339 
340 	M_ASSERTPKTHDR(m);
341 	NET_EPOCH_ASSERT();
342 
343 	if (inp != NULL) {
344 		INP_LOCK_ASSERT(inp);
345 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
346 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
347 			m->m_pkthdr.flowid = inp->inp_flowid;
348 			M_HASHTYPE_SET(m, inp->inp_flowtype);
349 		}
350 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
351 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
352 			    INP_2PCP_SHIFT;
353 #ifdef NUMA
354 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
355 #endif
356 	}
357 
358 	if (opt) {
359 		int len = 0;
360 		m = ip_insertoptions(m, opt, &len);
361 		if (len != 0)
362 			hlen = len; /* ip->ip_hl is updated above */
363 	}
364 	ip = mtod(m, struct ip *);
365 	ip_len = ntohs(ip->ip_len);
366 	ip_off = ntohs(ip->ip_off);
367 
368 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
369 		ip->ip_v = IPVERSION;
370 		ip->ip_hl = hlen >> 2;
371 		ip_fillid(ip);
372 	} else {
373 		/* Header already set, fetch hlen from there */
374 		hlen = ip->ip_hl << 2;
375 	}
376 	if ((flags & IP_FORWARDING) == 0)
377 		IPSTAT_INC(ips_localout);
378 
379 	/*
380 	 * dst/gw handling:
381 	 *
382 	 * gw is readonly but can point either to dst OR rt_gateway,
383 	 * therefore we need restore gw if we're redoing lookup.
384 	 */
385 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
386 	if (ro == NULL) {
387 		ro = &iproute;
388 		bzero(ro, sizeof (*ro));
389 	}
390 	dst = (struct sockaddr_in *)&ro->ro_dst;
391 	if (ro->ro_nh == NULL) {
392 		dst->sin_family = AF_INET;
393 		dst->sin_len = sizeof(*dst);
394 		dst->sin_addr = ip->ip_dst;
395 	}
396 	gw = (const struct sockaddr *)dst;
397 again:
398 	/*
399 	 * Validate route against routing table additions;
400 	 * a better/more specific route might have been added.
401 	 */
402 	if (inp != NULL && ro->ro_nh != NULL)
403 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
404 	/*
405 	 * If there is a cached route,
406 	 * check that it is to the same destination
407 	 * and is still up.  If not, free it and try again.
408 	 * The address family should also be checked in case of sharing the
409 	 * cache with IPv6.
410 	 * Also check whether routing cache needs invalidation.
411 	 */
412 	if (ro->ro_nh != NULL &&
413 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
414 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
415 		RO_INVALIDATE_CACHE(ro);
416 	ia = NULL;
417 	/*
418 	 * If routing to interface only, short circuit routing lookup.
419 	 * The use of an all-ones broadcast address implies this; an
420 	 * interface is specified by the broadcast address of an interface,
421 	 * or the destination address of a ptp interface.
422 	 */
423 	if (flags & IP_SENDONES) {
424 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
425 						      M_GETFIB(m)))) == NULL &&
426 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
427 						    M_GETFIB(m)))) == NULL) {
428 			IPSTAT_INC(ips_noroute);
429 			error = ENETUNREACH;
430 			goto bad;
431 		}
432 		ip->ip_dst.s_addr = INADDR_BROADCAST;
433 		dst->sin_addr = ip->ip_dst;
434 		ifp = ia->ia_ifp;
435 		mtu = ifp->if_mtu;
436 		ip->ip_ttl = 1;
437 		isbroadcast = true;
438 		src = IA_SIN(ia)->sin_addr;
439 	} else if (flags & IP_ROUTETOIF) {
440 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
441 						    M_GETFIB(m)))) == NULL &&
442 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
443 						M_GETFIB(m)))) == NULL) {
444 			IPSTAT_INC(ips_noroute);
445 			error = ENETUNREACH;
446 			goto bad;
447 		}
448 		ifp = ia->ia_ifp;
449 		mtu = ifp->if_mtu;
450 		ip->ip_ttl = 1;
451 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
452 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
453 		src = IA_SIN(ia)->sin_addr;
454 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
455 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
456 		/*
457 		 * Bypass the normal routing lookup for multicast
458 		 * packets if the interface is specified.
459 		 */
460 		ifp = imo->imo_multicast_ifp;
461 		mtu = ifp->if_mtu;
462 		IFP_TO_IA(ifp, ia);
463 		isbroadcast = false;
464 		/* Interface may have no addresses. */
465 		if (ia != NULL)
466 			src = IA_SIN(ia)->sin_addr;
467 		else
468 			src.s_addr = INADDR_ANY;
469 	} else if (ro != &iproute) {
470 		if (ro->ro_nh == NULL) {
471 			/*
472 			 * We want to do any cloning requested by the link
473 			 * layer, as this is probably required in all cases
474 			 * for correct operation (as it is for ARP).
475 			 */
476 			uint32_t flowid;
477 			flowid = m->m_pkthdr.flowid;
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 		struct nhop_object *nh = ro->ro_nh;
496 
497 		ia = ifatoia(nh->nh_ifa);
498 		ifp = nh->nh_ifp;
499 		counter_u64_add(nh->nh_pksent, 1);
500 		rt_update_ro_flags(ro, nh);
501 		if (nh->nh_flags & NHF_GATEWAY)
502 			gw = &nh->gw_sa;
503 		if (nh->nh_flags & NHF_HOST)
504 			isbroadcast = (nh->nh_flags & NHF_BROADCAST);
505 		else if ((ifp->if_flags & IFF_BROADCAST) && (gw->sa_family == AF_INET))
506 			isbroadcast = in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia);
507 		else
508 			isbroadcast = false;
509 		mtu = nh->nh_mtu;
510 		src = IA_SIN(ia)->sin_addr;
511 	} else {
512 		struct nhop_object *nh;
513 
514 		nh = fib4_lookup(M_GETFIB(m), dst->sin_addr, 0, NHR_NONE,
515 		    m->m_pkthdr.flowid);
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 		rt_update_ro_flags(ro, nh);
532 		if (nh->nh_flags & NHF_GATEWAY)
533 			gw = &nh->gw_sa;
534 		ia = ifatoia(nh->nh_ifa);
535 		src = IA_SIN(ia)->sin_addr;
536 		isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
537 		    (NHF_HOST | NHF_BROADCAST)) ||
538 		    ((ifp->if_flags & IFF_BROADCAST) &&
539 		    (gw->sa_family == AF_INET) &&
540 		    in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia)));
541 	}
542 
543 	/* Catch a possible divide by zero later. */
544 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
545 	    __func__, mtu, ro,
546 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
547 
548 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
549 		m->m_flags |= M_MCAST;
550 		/*
551 		 * IP destination address is multicast.  Make sure "gw"
552 		 * still points to the address in "ro".  (It may have been
553 		 * changed to point to a gateway address, above.)
554 		 */
555 		gw = (const struct sockaddr *)dst;
556 		/*
557 		 * See if the caller provided any multicast options
558 		 */
559 		if (imo != NULL) {
560 			ip->ip_ttl = imo->imo_multicast_ttl;
561 			if (imo->imo_multicast_vif != -1)
562 				ip->ip_src.s_addr =
563 				    ip_mcast_src ?
564 				    ip_mcast_src(imo->imo_multicast_vif) :
565 				    INADDR_ANY;
566 		} else
567 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
568 		/*
569 		 * Confirm that the outgoing interface supports multicast.
570 		 */
571 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
572 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
573 				IPSTAT_INC(ips_noroute);
574 				error = ENETUNREACH;
575 				goto bad;
576 			}
577 		}
578 		/*
579 		 * If source address not specified yet, use address
580 		 * of outgoing interface.
581 		 */
582 		if (ip->ip_src.s_addr == INADDR_ANY)
583 			ip->ip_src = src;
584 
585 		if ((imo == NULL && in_mcast_loop) ||
586 		    (imo && imo->imo_multicast_loop)) {
587 			/*
588 			 * Loop back multicast datagram if not expressly
589 			 * forbidden to do so, even if we are not a member
590 			 * of the group; ip_input() will filter it later,
591 			 * thus deferring a hash lookup and mutex acquisition
592 			 * at the expense of a cheap copy using m_copym().
593 			 */
594 			ip_mloopback(ifp, m, hlen);
595 		} else {
596 			/*
597 			 * If we are acting as a multicast router, perform
598 			 * multicast forwarding as if the packet had just
599 			 * arrived on the interface to which we are about
600 			 * to send.  The multicast forwarding function
601 			 * recursively calls this function, using the
602 			 * IP_FORWARDING flag to prevent infinite recursion.
603 			 *
604 			 * Multicasts that are looped back by ip_mloopback(),
605 			 * above, will be forwarded by the ip_input() routine,
606 			 * if necessary.
607 			 */
608 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
609 				/*
610 				 * If rsvp daemon is not running, do not
611 				 * set ip_moptions. This ensures that the packet
612 				 * is multicast and not just sent down one link
613 				 * as prescribed by rsvpd.
614 				 */
615 				if (!V_rsvp_on)
616 					imo = NULL;
617 				if (ip_mforward &&
618 				    ip_mforward(ip, ifp, m, imo) != 0) {
619 					m_freem(m);
620 					goto done;
621 				}
622 			}
623 		}
624 
625 		/*
626 		 * Multicasts with a time-to-live of zero may be looped-
627 		 * back, above, but must not be transmitted on a network.
628 		 * Also, multicasts addressed to the loopback interface
629 		 * are not sent -- the above call to ip_mloopback() will
630 		 * loop back a copy. ip_input() will drop the copy if
631 		 * this host does not belong to the destination group on
632 		 * the loopback interface.
633 		 */
634 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
635 			m_freem(m);
636 			goto done;
637 		}
638 
639 		goto sendit;
640 	}
641 
642 	/*
643 	 * If the source address is not specified yet, use the address
644 	 * of the outoing interface.
645 	 */
646 	if (ip->ip_src.s_addr == INADDR_ANY)
647 		ip->ip_src = src;
648 
649 	/*
650 	 * Look for broadcast address and
651 	 * verify user is allowed to send
652 	 * such a packet.
653 	 */
654 	if (isbroadcast) {
655 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
656 			error = EADDRNOTAVAIL;
657 			goto bad;
658 		}
659 		if ((flags & IP_ALLOWBROADCAST) == 0) {
660 			error = EACCES;
661 			goto bad;
662 		}
663 		/* don't allow broadcast messages to be fragmented */
664 		if (ip_len > mtu) {
665 			error = EMSGSIZE;
666 			goto bad;
667 		}
668 		m->m_flags |= M_BCAST;
669 	} else {
670 		m->m_flags &= ~M_BCAST;
671 	}
672 
673 sendit:
674 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
675 	if (IPSEC_ENABLED(ipv4)) {
676 		struct ip ip_hdr;
677 
678 		if ((error = IPSEC_OUTPUT(ipv4, ifp, m, inp, mtu)) != 0) {
679 			if (error == EINPROGRESS)
680 				error = 0;
681 			goto done;
682 		}
683 
684 		/* Update variables that are affected by ipsec4_output(). */
685 		m_copydata(m, 0, sizeof(ip_hdr), (char *)&ip_hdr);
686 		hlen = ip_hdr.ip_hl << 2;
687 	}
688 
689 	/*
690 	 * Check if there was a route for this packet; return error if not.
691 	 */
692 	if (no_route_but_check_spd) {
693 		IPSTAT_INC(ips_noroute);
694 		error = EHOSTUNREACH;
695 		goto bad;
696 	}
697 #endif /* IPSEC */
698 
699 	/* Jump over all PFIL processing if hooks are not active. */
700 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
701 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
702 		    &error)) {
703 		case 1: /* Finished */
704 			goto done;
705 
706 		case 0: /* Continue normally */
707 			ip = mtod(m, struct ip *);
708 			ip_len = ntohs(ip->ip_len);
709 			break;
710 
711 		case -1: /* Need to try again */
712 			/* Reset everything for a new round */
713 			if (ro != NULL) {
714 				RO_NHFREE(ro);
715 				ro->ro_prepend = NULL;
716 			}
717 			gw = (const struct sockaddr *)dst;
718 			ip = mtod(m, struct ip *);
719 			goto again;
720 		}
721 	}
722 
723 	if (vlan_pcp > -1)
724 		EVL_APPLY_PRI(m, vlan_pcp);
725 
726 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
727 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
728 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
729 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
730 			IPSTAT_INC(ips_badaddr);
731 			error = EADDRNOTAVAIL;
732 			goto bad;
733 		}
734 	}
735 
736 	/* Ensure the packet data is mapped if the interface requires it. */
737 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
738 		struct mbuf *m1;
739 
740 		error = mb_unmapped_to_ext(m, &m1);
741 		if (error != 0) {
742 			if (error == EINVAL) {
743 				if_printf(ifp, "TLS packet\n");
744 				/* XXXKIB */
745 			} else if (error == ENOMEM) {
746 				error = ENOBUFS;
747 			}
748 			IPSTAT_INC(ips_odropped);
749 			goto bad;
750 		} else {
751 			m = m1;
752 		}
753 	}
754 
755 	m->m_pkthdr.csum_flags |= CSUM_IP;
756 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
757 		in_delayed_cksum(m);
758 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
759 	}
760 #if defined(SCTP) || defined(SCTP_SUPPORT)
761 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
762 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
763 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
764 	}
765 #endif
766 
767 	/*
768 	 * If small enough for interface, or the interface will take
769 	 * care of the fragmentation for us, we can just send directly.
770 	 * Note that if_vxlan could have requested TSO even though the outer
771 	 * frame is UDP.  It is correct to not fragment such datagrams and
772 	 * instead just pass them on to the driver.
773 	 */
774 	if (ip_len <= mtu ||
775 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
776 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
777 		ip->ip_sum = 0;
778 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
779 			ip->ip_sum = in_cksum(m, hlen);
780 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
781 		}
782 
783 		/*
784 		 * Record statistics for this interface address.
785 		 * With CSUM_TSO the byte/packet count will be slightly
786 		 * incorrect because we count the IP+TCP headers only
787 		 * once instead of for every generated packet.
788 		 */
789 		if (!(flags & IP_FORWARDING) && ia) {
790 			if (m->m_pkthdr.csum_flags &
791 			    (CSUM_TSO | CSUM_INNER_TSO))
792 				counter_u64_add(ia->ia_ifa.ifa_opackets,
793 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
794 			else
795 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
796 
797 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
798 		}
799 #ifdef MBUF_STRESS_TEST
800 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
801 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
802 #endif
803 		/*
804 		 * Reset layer specific mbuf flags
805 		 * to avoid confusing lower layers.
806 		 */
807 		m_clrprotoflags(m);
808 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
809 		error = ip_output_send(inp, ifp, m, gw, ro,
810 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
811 		goto done;
812 	}
813 
814 	/* Balk when DF bit is set or the interface didn't support TSO. */
815 	if ((ip_off & IP_DF) ||
816 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
817 		error = EMSGSIZE;
818 		IPSTAT_INC(ips_cantfrag);
819 		goto bad;
820 	}
821 
822 	/*
823 	 * Too large for interface; fragment if possible. If successful,
824 	 * on return, m will point to a list of packets to be sent.
825 	 */
826 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
827 	if (error)
828 		goto bad;
829 	for (; m; m = m0) {
830 		m0 = m->m_nextpkt;
831 		m->m_nextpkt = 0;
832 		if (error == 0) {
833 			/* Record statistics for this interface address. */
834 			if (ia != NULL) {
835 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
836 				counter_u64_add(ia->ia_ifa.ifa_obytes,
837 				    m->m_pkthdr.len);
838 			}
839 			/*
840 			 * Reset layer specific mbuf flags
841 			 * to avoid confusing upper layers.
842 			 */
843 			m_clrprotoflags(m);
844 
845 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
846 			    mtod(m, struct ip *), NULL);
847 			error = ip_output_send(inp, ifp, m, gw, ro, true);
848 		} else
849 			m_freem(m);
850 	}
851 
852 	if (error == 0)
853 		IPSTAT_INC(ips_fragmented);
854 
855 done:
856 	return (error);
857  bad:
858 	m_freem(m);
859 	goto done;
860 }
861 
862 /*
863  * Create a chain of fragments which fit the given mtu. m_frag points to the
864  * mbuf to be fragmented; on return it points to the chain with the fragments.
865  * Return 0 if no error. If error, m_frag may contain a partially built
866  * chain of fragments that should be freed by the caller.
867  *
868  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
869  */
870 int
871 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
872     u_long if_hwassist_flags)
873 {
874 	int error = 0;
875 	int hlen = ip->ip_hl << 2;
876 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
877 	int off;
878 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
879 	int firstlen;
880 	struct mbuf **mnext;
881 	int nfrags;
882 	uint16_t ip_len, ip_off;
883 
884 	ip_len = ntohs(ip->ip_len);
885 	ip_off = ntohs(ip->ip_off);
886 
887 	/*
888 	 * Packet shall not have "Don't Fragment" flag and have at least 8
889 	 * bytes of payload.
890 	 */
891 	if (__predict_false((ip_off & IP_DF) || len < 8)) {
892 		IPSTAT_INC(ips_cantfrag);
893 		return (EMSGSIZE);
894 	}
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 		in_delayed_cksum(m0);
902 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
903 	}
904 #if defined(SCTP) || defined(SCTP_SUPPORT)
905 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
906 		sctp_delayed_cksum(m0, hlen);
907 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
908 	}
909 #endif
910 	if (len > PAGE_SIZE) {
911 		/*
912 		 * Fragment large datagrams such that each segment
913 		 * contains a multiple of PAGE_SIZE amount of data,
914 		 * plus headers. This enables a receiver to perform
915 		 * page-flipping zero-copy optimizations.
916 		 *
917 		 * XXX When does this help given that sender and receiver
918 		 * could have different page sizes, and also mtu could
919 		 * be less than the receiver's page size ?
920 		 */
921 		int newlen;
922 
923 		off = MIN(mtu, m0->m_pkthdr.len);
924 
925 		/*
926 		 * firstlen (off - hlen) must be aligned on an
927 		 * 8-byte boundary
928 		 */
929 		if (off < hlen)
930 			goto smart_frag_failure;
931 		off = ((off - hlen) & ~7) + hlen;
932 		newlen = (~PAGE_MASK) & mtu;
933 		if ((newlen + sizeof (struct ip)) > mtu) {
934 			/* we failed, go back the default */
935 smart_frag_failure:
936 			newlen = len;
937 			off = hlen + len;
938 		}
939 		len = newlen;
940 
941 	} else {
942 		off = hlen + len;
943 	}
944 
945 	firstlen = off - hlen;
946 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
947 
948 	/*
949 	 * Loop through length of segment after first fragment,
950 	 * make new header and copy data of each part and link onto chain.
951 	 * Here, m0 is the original packet, m is the fragment being created.
952 	 * The fragments are linked off the m_nextpkt of the original
953 	 * packet, which after processing serves as the first fragment.
954 	 */
955 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
956 		struct ip *mhip;	/* ip header on the fragment */
957 		struct mbuf *m;
958 		int mhlen = sizeof (struct ip);
959 
960 		m = m_gethdr(M_NOWAIT, MT_DATA);
961 		if (m == NULL) {
962 			error = ENOBUFS;
963 			IPSTAT_INC(ips_odropped);
964 			goto done;
965 		}
966 		/*
967 		 * Make sure the complete packet header gets copied
968 		 * from the originating mbuf to the newly created
969 		 * mbuf. This also ensures that existing firewall
970 		 * classification(s), VLAN tags and so on get copied
971 		 * to the resulting fragmented packet(s):
972 		 */
973 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
974 			m_free(m);
975 			error = ENOBUFS;
976 			IPSTAT_INC(ips_odropped);
977 			goto done;
978 		}
979 		/*
980 		 * In the first mbuf, leave room for the link header, then
981 		 * copy the original IP header including options. The payload
982 		 * goes into an additional mbuf chain returned by m_copym().
983 		 */
984 		m->m_data += max_linkhdr;
985 		mhip = mtod(m, struct ip *);
986 		*mhip = *ip;
987 		if (hlen > sizeof (struct ip)) {
988 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
989 			mhip->ip_v = IPVERSION;
990 			mhip->ip_hl = mhlen >> 2;
991 		}
992 		m->m_len = mhlen;
993 		/* XXX do we need to add ip_off below ? */
994 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
995 		if (off + len >= ip_len)
996 			len = ip_len - off;
997 		else
998 			mhip->ip_off |= IP_MF;
999 		mhip->ip_len = htons((u_short)(len + mhlen));
1000 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1001 		if (m->m_next == NULL) {	/* copy failed */
1002 			m_free(m);
1003 			error = ENOBUFS;	/* ??? */
1004 			IPSTAT_INC(ips_odropped);
1005 			goto done;
1006 		}
1007 		m->m_pkthdr.len = mhlen + len;
1008 #ifdef MAC
1009 		mac_netinet_fragment(m0, m);
1010 #endif
1011 		mhip->ip_off = htons(mhip->ip_off);
1012 		mhip->ip_sum = 0;
1013 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1014 			mhip->ip_sum = in_cksum(m, mhlen);
1015 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1016 		}
1017 		*mnext = m;
1018 		mnext = &m->m_nextpkt;
1019 	}
1020 	IPSTAT_ADD(ips_ofragments, nfrags);
1021 
1022 	/*
1023 	 * Update first fragment by trimming what's been copied out
1024 	 * and updating header.
1025 	 */
1026 	m_adj(m0, hlen + firstlen - ip_len);
1027 	m0->m_pkthdr.len = hlen + firstlen;
1028 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1029 	ip->ip_off = htons(ip_off | IP_MF);
1030 	ip->ip_sum = 0;
1031 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1032 		ip->ip_sum = in_cksum(m0, hlen);
1033 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1034 	}
1035 
1036 done:
1037 	*m_frag = m0;
1038 	return error;
1039 }
1040 
1041 void
1042 in_delayed_cksum(struct mbuf *m)
1043 {
1044 	struct ip *ip;
1045 	struct udphdr *uh;
1046 	uint16_t cklen, csum, offset;
1047 
1048 	ip = mtod(m, struct ip *);
1049 	offset = ip->ip_hl << 2 ;
1050 
1051 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1052 		/* if udp header is not in the first mbuf copy udplen */
1053 		if (offset + sizeof(struct udphdr) > m->m_len) {
1054 			m_copydata(m, offset + offsetof(struct udphdr,
1055 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1056 			cklen = ntohs(cklen);
1057 		} else {
1058 			uh = (struct udphdr *)mtodo(m, offset);
1059 			cklen = ntohs(uh->uh_ulen);
1060 		}
1061 		csum = in_cksum_skip(m, cklen + offset, offset);
1062 		if (csum == 0)
1063 			csum = 0xffff;
1064 	} else {
1065 		cklen = ntohs(ip->ip_len);
1066 		csum = in_cksum_skip(m, cklen, offset);
1067 	}
1068 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1069 
1070 	if (offset + sizeof(csum) > m->m_len)
1071 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1072 	else
1073 		*(u_short *)mtodo(m, offset) = csum;
1074 }
1075 
1076 /*
1077  * IP socket option processing.
1078  */
1079 int
1080 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1081 {
1082 	struct inpcb *inp = sotoinpcb(so);
1083 	int	error, optval;
1084 #ifdef	RSS
1085 	uint32_t rss_bucket;
1086 	int retval;
1087 #endif
1088 
1089 	error = optval = 0;
1090 	if (sopt->sopt_level != IPPROTO_IP) {
1091 		error = EINVAL;
1092 
1093 		if (sopt->sopt_level == SOL_SOCKET &&
1094 		    sopt->sopt_dir == SOPT_SET) {
1095 			switch (sopt->sopt_name) {
1096 			case SO_SETFIB:
1097 				error = sooptcopyin(sopt, &optval,
1098 				    sizeof(optval), sizeof(optval));
1099 				if (error != 0)
1100 					break;
1101 
1102 				INP_WLOCK(inp);
1103 				if ((inp->inp_flags & INP_BOUNDFIB) != 0 &&
1104 				    optval != so->so_fibnum) {
1105 					INP_WUNLOCK(inp);
1106 					error = EISCONN;
1107 					break;
1108 				}
1109 				error = sosetfib(inp->inp_socket, optval);
1110 				if (error == 0)
1111 					inp->inp_inc.inc_fibnum = optval;
1112 				INP_WUNLOCK(inp);
1113 				break;
1114 			case SO_MAX_PACING_RATE:
1115 #ifdef RATELIMIT
1116 				INP_WLOCK(inp);
1117 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1118 				INP_WUNLOCK(inp);
1119 				error = 0;
1120 #else
1121 				error = EOPNOTSUPP;
1122 #endif
1123 				break;
1124 			default:
1125 				break;
1126 			}
1127 		}
1128 		return (error);
1129 	}
1130 
1131 	switch (sopt->sopt_dir) {
1132 	case SOPT_SET:
1133 		switch (sopt->sopt_name) {
1134 		case IP_OPTIONS:
1135 #ifdef notyet
1136 		case IP_RETOPTS:
1137 #endif
1138 		{
1139 			struct mbuf *m;
1140 			if (sopt->sopt_valsize > MLEN) {
1141 				error = EMSGSIZE;
1142 				break;
1143 			}
1144 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1145 			if (m == NULL) {
1146 				error = ENOBUFS;
1147 				break;
1148 			}
1149 			m->m_len = sopt->sopt_valsize;
1150 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1151 					    m->m_len);
1152 			if (error) {
1153 				m_free(m);
1154 				break;
1155 			}
1156 			INP_WLOCK(inp);
1157 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1158 			INP_WUNLOCK(inp);
1159 			return (error);
1160 		}
1161 
1162 		case IP_BINDANY:
1163 			if (sopt->sopt_td != NULL) {
1164 				error = priv_check(sopt->sopt_td,
1165 				    PRIV_NETINET_BINDANY);
1166 				if (error)
1167 					break;
1168 			}
1169 			/* FALLTHROUGH */
1170 		case IP_TOS:
1171 		case IP_TTL:
1172 		case IP_MINTTL:
1173 		case IP_RECVOPTS:
1174 		case IP_RECVRETOPTS:
1175 		case IP_ORIGDSTADDR:
1176 		case IP_RECVDSTADDR:
1177 		case IP_RECVTTL:
1178 		case IP_RECVIF:
1179 		case IP_ONESBCAST:
1180 		case IP_DONTFRAG:
1181 		case IP_RECVTOS:
1182 		case IP_RECVFLOWID:
1183 #ifdef	RSS
1184 		case IP_RECVRSSBUCKETID:
1185 #endif
1186 		case IP_VLAN_PCP:
1187 			error = sooptcopyin(sopt, &optval, sizeof optval,
1188 					    sizeof optval);
1189 			if (error)
1190 				break;
1191 
1192 			switch (sopt->sopt_name) {
1193 			case IP_TOS:
1194 				inp->inp_ip_tos = optval;
1195 				break;
1196 
1197 			case IP_TTL:
1198 				inp->inp_ip_ttl = optval;
1199 				break;
1200 
1201 			case IP_MINTTL:
1202 				if (optval >= 0 && optval <= MAXTTL)
1203 					inp->inp_ip_minttl = optval;
1204 				else
1205 					error = EINVAL;
1206 				break;
1207 
1208 #define	OPTSET(bit) do {						\
1209 	INP_WLOCK(inp);							\
1210 	if (optval)							\
1211 		inp->inp_flags |= bit;					\
1212 	else								\
1213 		inp->inp_flags &= ~bit;					\
1214 	INP_WUNLOCK(inp);						\
1215 } while (0)
1216 
1217 #define	OPTSET2(bit, val) do {						\
1218 	INP_WLOCK(inp);							\
1219 	if (val)							\
1220 		inp->inp_flags2 |= bit;					\
1221 	else								\
1222 		inp->inp_flags2 &= ~bit;				\
1223 	INP_WUNLOCK(inp);						\
1224 } while (0)
1225 
1226 			case IP_RECVOPTS:
1227 				OPTSET(INP_RECVOPTS);
1228 				break;
1229 
1230 			case IP_RECVRETOPTS:
1231 				OPTSET(INP_RECVRETOPTS);
1232 				break;
1233 
1234 			case IP_RECVDSTADDR:
1235 				OPTSET(INP_RECVDSTADDR);
1236 				break;
1237 
1238 			case IP_ORIGDSTADDR:
1239 				OPTSET2(INP_ORIGDSTADDR, optval);
1240 				break;
1241 
1242 			case IP_RECVTTL:
1243 				OPTSET(INP_RECVTTL);
1244 				break;
1245 
1246 			case IP_RECVIF:
1247 				OPTSET(INP_RECVIF);
1248 				break;
1249 
1250 			case IP_ONESBCAST:
1251 				OPTSET(INP_ONESBCAST);
1252 				break;
1253 			case IP_DONTFRAG:
1254 				OPTSET(INP_DONTFRAG);
1255 				break;
1256 			case IP_BINDANY:
1257 				OPTSET(INP_BINDANY);
1258 				break;
1259 			case IP_RECVTOS:
1260 				OPTSET(INP_RECVTOS);
1261 				break;
1262 			case IP_RECVFLOWID:
1263 				OPTSET2(INP_RECVFLOWID, optval);
1264 				break;
1265 #ifdef RSS
1266 			case IP_RECVRSSBUCKETID:
1267 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1268 				break;
1269 #endif
1270 			case IP_VLAN_PCP:
1271 				if ((optval >= -1) && (optval <=
1272 				    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1273 					if (optval == -1) {
1274 						INP_WLOCK(inp);
1275 						inp->inp_flags2 &=
1276 						    ~(INP_2PCP_SET |
1277 						      INP_2PCP_MASK);
1278 						INP_WUNLOCK(inp);
1279 					} else {
1280 						INP_WLOCK(inp);
1281 						inp->inp_flags2 |=
1282 						    INP_2PCP_SET;
1283 						inp->inp_flags2 &=
1284 						    ~INP_2PCP_MASK;
1285 						inp->inp_flags2 |=
1286 						    optval << INP_2PCP_SHIFT;
1287 						INP_WUNLOCK(inp);
1288 					}
1289 				} else
1290 					error = EINVAL;
1291 				break;
1292 			}
1293 			break;
1294 #undef OPTSET
1295 #undef OPTSET2
1296 
1297 		/*
1298 		 * Multicast socket options are processed by the in_mcast
1299 		 * module.
1300 		 */
1301 		case IP_MULTICAST_IF:
1302 		case IP_MULTICAST_VIF:
1303 		case IP_MULTICAST_TTL:
1304 		case IP_MULTICAST_LOOP:
1305 		case IP_ADD_MEMBERSHIP:
1306 		case IP_DROP_MEMBERSHIP:
1307 		case IP_ADD_SOURCE_MEMBERSHIP:
1308 		case IP_DROP_SOURCE_MEMBERSHIP:
1309 		case IP_BLOCK_SOURCE:
1310 		case IP_UNBLOCK_SOURCE:
1311 		case IP_MSFILTER:
1312 		case MCAST_JOIN_GROUP:
1313 		case MCAST_LEAVE_GROUP:
1314 		case MCAST_JOIN_SOURCE_GROUP:
1315 		case MCAST_LEAVE_SOURCE_GROUP:
1316 		case MCAST_BLOCK_SOURCE:
1317 		case MCAST_UNBLOCK_SOURCE:
1318 			error = inp_setmoptions(inp, sopt);
1319 			break;
1320 
1321 		case IP_PORTRANGE:
1322 			error = sooptcopyin(sopt, &optval, sizeof optval,
1323 					    sizeof optval);
1324 			if (error)
1325 				break;
1326 
1327 			INP_WLOCK(inp);
1328 			switch (optval) {
1329 			case IP_PORTRANGE_DEFAULT:
1330 				inp->inp_flags &= ~(INP_LOWPORT);
1331 				inp->inp_flags &= ~(INP_HIGHPORT);
1332 				break;
1333 
1334 			case IP_PORTRANGE_HIGH:
1335 				inp->inp_flags &= ~(INP_LOWPORT);
1336 				inp->inp_flags |= INP_HIGHPORT;
1337 				break;
1338 
1339 			case IP_PORTRANGE_LOW:
1340 				inp->inp_flags &= ~(INP_HIGHPORT);
1341 				inp->inp_flags |= INP_LOWPORT;
1342 				break;
1343 
1344 			default:
1345 				error = EINVAL;
1346 				break;
1347 			}
1348 			INP_WUNLOCK(inp);
1349 			break;
1350 
1351 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1352 		case IP_IPSEC_POLICY:
1353 			if (IPSEC_ENABLED(ipv4)) {
1354 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1355 				break;
1356 			}
1357 			/* FALLTHROUGH */
1358 #endif /* IPSEC */
1359 
1360 		default:
1361 			error = ENOPROTOOPT;
1362 			break;
1363 		}
1364 		break;
1365 
1366 	case SOPT_GET:
1367 		switch (sopt->sopt_name) {
1368 		case IP_OPTIONS:
1369 		case IP_RETOPTS:
1370 			INP_RLOCK(inp);
1371 			if (inp->inp_options) {
1372 				struct mbuf *options;
1373 
1374 				options = m_copym(inp->inp_options, 0,
1375 				    M_COPYALL, M_NOWAIT);
1376 				INP_RUNLOCK(inp);
1377 				if (options != NULL) {
1378 					error = sooptcopyout(sopt,
1379 							     mtod(options, char *),
1380 							     options->m_len);
1381 					m_freem(options);
1382 				} else
1383 					error = ENOMEM;
1384 			} else {
1385 				INP_RUNLOCK(inp);
1386 				sopt->sopt_valsize = 0;
1387 			}
1388 			break;
1389 
1390 		case IP_TOS:
1391 		case IP_TTL:
1392 		case IP_MINTTL:
1393 		case IP_RECVOPTS:
1394 		case IP_RECVRETOPTS:
1395 		case IP_ORIGDSTADDR:
1396 		case IP_RECVDSTADDR:
1397 		case IP_RECVTTL:
1398 		case IP_RECVIF:
1399 		case IP_PORTRANGE:
1400 		case IP_ONESBCAST:
1401 		case IP_DONTFRAG:
1402 		case IP_BINDANY:
1403 		case IP_RECVTOS:
1404 		case IP_FLOWID:
1405 		case IP_FLOWTYPE:
1406 		case IP_RECVFLOWID:
1407 #ifdef	RSS
1408 		case IP_RSSBUCKETID:
1409 		case IP_RECVRSSBUCKETID:
1410 #endif
1411 		case IP_VLAN_PCP:
1412 			switch (sopt->sopt_name) {
1413 			case IP_TOS:
1414 				optval = inp->inp_ip_tos;
1415 				break;
1416 
1417 			case IP_TTL:
1418 				optval = inp->inp_ip_ttl;
1419 				break;
1420 
1421 			case IP_MINTTL:
1422 				optval = inp->inp_ip_minttl;
1423 				break;
1424 
1425 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1426 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1427 
1428 			case IP_RECVOPTS:
1429 				optval = OPTBIT(INP_RECVOPTS);
1430 				break;
1431 
1432 			case IP_RECVRETOPTS:
1433 				optval = OPTBIT(INP_RECVRETOPTS);
1434 				break;
1435 
1436 			case IP_RECVDSTADDR:
1437 				optval = OPTBIT(INP_RECVDSTADDR);
1438 				break;
1439 
1440 			case IP_ORIGDSTADDR:
1441 				optval = OPTBIT2(INP_ORIGDSTADDR);
1442 				break;
1443 
1444 			case IP_RECVTTL:
1445 				optval = OPTBIT(INP_RECVTTL);
1446 				break;
1447 
1448 			case IP_RECVIF:
1449 				optval = OPTBIT(INP_RECVIF);
1450 				break;
1451 
1452 			case IP_PORTRANGE:
1453 				if (inp->inp_flags & INP_HIGHPORT)
1454 					optval = IP_PORTRANGE_HIGH;
1455 				else if (inp->inp_flags & INP_LOWPORT)
1456 					optval = IP_PORTRANGE_LOW;
1457 				else
1458 					optval = 0;
1459 				break;
1460 
1461 			case IP_ONESBCAST:
1462 				optval = OPTBIT(INP_ONESBCAST);
1463 				break;
1464 			case IP_DONTFRAG:
1465 				optval = OPTBIT(INP_DONTFRAG);
1466 				break;
1467 			case IP_BINDANY:
1468 				optval = OPTBIT(INP_BINDANY);
1469 				break;
1470 			case IP_RECVTOS:
1471 				optval = OPTBIT(INP_RECVTOS);
1472 				break;
1473 			case IP_FLOWID:
1474 				optval = inp->inp_flowid;
1475 				break;
1476 			case IP_FLOWTYPE:
1477 				optval = inp->inp_flowtype;
1478 				break;
1479 			case IP_RECVFLOWID:
1480 				optval = OPTBIT2(INP_RECVFLOWID);
1481 				break;
1482 #ifdef	RSS
1483 			case IP_RSSBUCKETID:
1484 				retval = rss_hash2bucket(inp->inp_flowid,
1485 				    inp->inp_flowtype,
1486 				    &rss_bucket);
1487 				if (retval == 0)
1488 					optval = rss_bucket;
1489 				else
1490 					error = EINVAL;
1491 				break;
1492 			case IP_RECVRSSBUCKETID:
1493 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1494 				break;
1495 #endif
1496 			case IP_VLAN_PCP:
1497 				if (OPTBIT2(INP_2PCP_SET)) {
1498 					optval = (inp->inp_flags2 &
1499 					    INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1500 				} else {
1501 					optval = -1;
1502 				}
1503 				break;
1504 			}
1505 			error = sooptcopyout(sopt, &optval, sizeof optval);
1506 			break;
1507 
1508 		/*
1509 		 * Multicast socket options are processed by the in_mcast
1510 		 * module.
1511 		 */
1512 		case IP_MULTICAST_IF:
1513 		case IP_MULTICAST_VIF:
1514 		case IP_MULTICAST_TTL:
1515 		case IP_MULTICAST_LOOP:
1516 		case IP_MSFILTER:
1517 			error = inp_getmoptions(inp, sopt);
1518 			break;
1519 
1520 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1521 		case IP_IPSEC_POLICY:
1522 			if (IPSEC_ENABLED(ipv4)) {
1523 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1524 				break;
1525 			}
1526 			/* FALLTHROUGH */
1527 #endif /* IPSEC */
1528 
1529 		default:
1530 			error = ENOPROTOOPT;
1531 			break;
1532 		}
1533 		break;
1534 	}
1535 	return (error);
1536 }
1537 
1538 /*
1539  * Routine called from ip_output() to loop back a copy of an IP multicast
1540  * packet to the input queue of a specified interface.  Note that this
1541  * calls the output routine of the loopback "driver", but with an interface
1542  * pointer that might NOT be a loopback interface -- evil, but easier than
1543  * replicating that code here.
1544  */
1545 static void
1546 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1547 {
1548 	struct ip *ip;
1549 	struct mbuf *copym;
1550 
1551 	/*
1552 	 * Make a deep copy of the packet because we're going to
1553 	 * modify the pack in order to generate checksums.
1554 	 */
1555 	copym = m_dup(m, M_NOWAIT);
1556 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1557 		copym = m_pullup(copym, hlen);
1558 	if (copym != NULL) {
1559 		/* If needed, compute the checksum and mark it as valid. */
1560 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1561 			in_delayed_cksum(copym);
1562 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1563 			copym->m_pkthdr.csum_flags |=
1564 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1565 			copym->m_pkthdr.csum_data = 0xffff;
1566 		}
1567 		/*
1568 		 * We don't bother to fragment if the IP length is greater
1569 		 * than the interface's MTU.  Can this possibly matter?
1570 		 */
1571 		ip = mtod(copym, struct ip *);
1572 		ip->ip_sum = 0;
1573 		ip->ip_sum = in_cksum(copym, hlen);
1574 		if_simloop(ifp, copym, AF_INET, 0);
1575 	}
1576 }
1577