xref: /freebsd/sys/netinet/ip_output.c (revision 32cd3ee5901ea33d41ff550e5f40ce743c8d4165)
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 "opt_inet.h"
33 #include "opt_ipsec.h"
34 #include "opt_kern_tls.h"
35 #include "opt_mbuf_stress_test.h"
36 #include "opt_ratelimit.h"
37 #include "opt_route.h"
38 #include "opt_rss.h"
39 #include "opt_sctp.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/ktls.h>
45 #include <sys/lock.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/priv.h>
49 #include <sys/proc.h>
50 #include <sys/protosw.h>
51 #include <sys/sdt.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/sysctl.h>
55 #include <sys/ucred.h>
56 
57 #include <net/if.h>
58 #include <net/if_var.h>
59 #include <net/if_private.h>
60 #include <net/if_vlan_var.h>
61 #include <net/if_llatbl.h>
62 #include <net/ethernet.h>
63 #include <net/netisr.h>
64 #include <net/pfil.h>
65 #include <net/route.h>
66 #include <net/route/nhop.h>
67 #include <net/rss_config.h>
68 #include <net/vnet.h>
69 
70 #include <netinet/in.h>
71 #include <netinet/in_fib.h>
72 #include <netinet/in_kdtrace.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/ip.h>
75 #include <netinet/in_fib.h>
76 #include <netinet/in_pcb.h>
77 #include <netinet/in_rss.h>
78 #include <netinet/in_var.h>
79 #include <netinet/ip_var.h>
80 #include <netinet/ip_options.h>
81 #include <netinet/ip_mroute.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, V_ip_random_id);
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_broadcast(ip->ip_dst) ||
453 		    in_ifaddr_broadcast(dst->sin_addr, ia)) : 0;
454 		src = IA_SIN(ia)->sin_addr;
455 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
456 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
457 		/*
458 		 * Bypass the normal routing lookup for multicast
459 		 * packets if the interface is specified.
460 		 */
461 		ifp = imo->imo_multicast_ifp;
462 		mtu = ifp->if_mtu;
463 		IFP_TO_IA(ifp, ia);
464 		isbroadcast = false;
465 		/* Interface may have no addresses. */
466 		if (ia != NULL)
467 			src = IA_SIN(ia)->sin_addr;
468 		else
469 			src.s_addr = INADDR_ANY;
470 	} else if (ro != &iproute) {
471 		if (ro->ro_nh == NULL) {
472 			/*
473 			 * We want to do any cloning requested by the link
474 			 * layer, as this is probably required in all cases
475 			 * for correct operation (as it is for ARP).
476 			 */
477 			uint32_t flowid;
478 			flowid = m->m_pkthdr.flowid;
479 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
480 			    NHR_REF, flowid);
481 
482 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
483 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
484 				/*
485 				 * There is no route for this packet, but it is
486 				 * possible that a matching SPD entry exists.
487 				 */
488 				no_route_but_check_spd = 1;
489 				goto sendit;
490 #endif
491 				IPSTAT_INC(ips_noroute);
492 				error = EHOSTUNREACH;
493 				goto bad;
494 			}
495 		}
496 		struct nhop_object *nh = ro->ro_nh;
497 
498 		ia = ifatoia(nh->nh_ifa);
499 		ifp = nh->nh_ifp;
500 		counter_u64_add(nh->nh_pksent, 1);
501 		rt_update_ro_flags(ro, nh);
502 		if (nh->nh_flags & NHF_GATEWAY)
503 			gw = &nh->gw_sa;
504 		if (nh->nh_flags & NHF_HOST)
505 			isbroadcast = (nh->nh_flags & NHF_BROADCAST);
506 		else if ((ifp->if_flags & IFF_BROADCAST) &&
507 		    (gw->sa_family == AF_INET))
508 			isbroadcast = in_broadcast(ip->ip_dst) ||
509 			    in_ifaddr_broadcast(
510 			    ((const struct sockaddr_in *)gw)->sin_addr, ia);
511 		else
512 			isbroadcast = false;
513 		mtu = nh->nh_mtu;
514 		src = IA_SIN(ia)->sin_addr;
515 	} else {
516 		struct nhop_object *nh;
517 
518 		nh = fib4_lookup(M_GETFIB(m), dst->sin_addr, 0, NHR_NONE,
519 		    m->m_pkthdr.flowid);
520 		if (nh == NULL) {
521 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
522 			/*
523 			 * There is no route for this packet, but it is
524 			 * possible that a matching SPD entry exists.
525 			 */
526 			no_route_but_check_spd = 1;
527 			goto sendit;
528 #endif
529 			IPSTAT_INC(ips_noroute);
530 			error = EHOSTUNREACH;
531 			goto bad;
532 		}
533 		ifp = nh->nh_ifp;
534 		mtu = nh->nh_mtu;
535 		rt_update_ro_flags(ro, nh);
536 		if (nh->nh_flags & NHF_GATEWAY)
537 			gw = &nh->gw_sa;
538 		ia = ifatoia(nh->nh_ifa);
539 		src = IA_SIN(ia)->sin_addr;
540 		isbroadcast = ((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
541 		    (NHF_HOST | NHF_BROADCAST)) ||
542 		    ((ifp->if_flags & IFF_BROADCAST) &&
543 		    (gw->sa_family == AF_INET) &&
544 		    (in_broadcast(ip->ip_dst) || in_ifaddr_broadcast(
545 		    ((const struct sockaddr_in *)gw)->sin_addr, ia)));
546 	}
547 
548 	/* Catch a possible divide by zero later. */
549 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
550 	    __func__, mtu, ro,
551 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
552 
553 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
554 		m->m_flags |= M_MCAST;
555 		/*
556 		 * IP destination address is multicast.  Make sure "gw"
557 		 * still points to the address in "ro".  (It may have been
558 		 * changed to point to a gateway address, above.)
559 		 */
560 		gw = (const struct sockaddr *)dst;
561 		/*
562 		 * See if the caller provided any multicast options
563 		 */
564 		if (imo != NULL) {
565 			ip->ip_ttl = imo->imo_multicast_ttl;
566 			if (imo->imo_multicast_vif != -1)
567 				ip->ip_src.s_addr =
568 				    ip_mcast_src ?
569 				    ip_mcast_src(imo->imo_multicast_vif) :
570 				    INADDR_ANY;
571 		} else
572 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
573 		/*
574 		 * Confirm that the outgoing interface supports multicast.
575 		 */
576 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
577 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
578 				IPSTAT_INC(ips_noroute);
579 				error = ENETUNREACH;
580 				goto bad;
581 			}
582 		}
583 		/*
584 		 * If source address not specified yet, use address
585 		 * of outgoing interface.
586 		 */
587 		if (ip->ip_src.s_addr == INADDR_ANY)
588 			ip->ip_src = src;
589 
590 		if ((imo == NULL && in_mcast_loop) ||
591 		    (imo && imo->imo_multicast_loop)) {
592 			/*
593 			 * Loop back multicast datagram if not expressly
594 			 * forbidden to do so, even if we are not a member
595 			 * of the group; ip_input() will filter it later,
596 			 * thus deferring a hash lookup and mutex acquisition
597 			 * at the expense of a cheap copy using m_copym().
598 			 */
599 			ip_mloopback(ifp, m, hlen);
600 		} else {
601 			/*
602 			 * If we are acting as a multicast router, perform
603 			 * multicast forwarding as if the packet had just
604 			 * arrived on the interface to which we are about
605 			 * to send.  The multicast forwarding function
606 			 * recursively calls this function, using the
607 			 * IP_FORWARDING flag to prevent infinite recursion.
608 			 *
609 			 * Multicasts that are looped back by ip_mloopback(),
610 			 * above, will be forwarded by the ip_input() routine,
611 			 * if necessary.
612 			 */
613 			if (V_ip_mrouting_enabled &&
614 			    (flags & IP_FORWARDING) == 0) {
615 				/*
616 				 * If rsvp daemon is not running, do not
617 				 * set ip_moptions. This ensures that the packet
618 				 * is multicast and not just sent down one link
619 				 * as prescribed by rsvpd.
620 				 */
621 				if (!V_rsvp_on)
622 					imo = NULL;
623 				if (ip_mforward &&
624 				    ip_mforward(ip, ifp, m, imo) != 0) {
625 					m_freem(m);
626 					goto done;
627 				}
628 			}
629 		}
630 
631 		/*
632 		 * Multicasts with a time-to-live of zero may be looped-
633 		 * back, above, but must not be transmitted on a network.
634 		 * Also, multicasts addressed to the loopback interface
635 		 * are not sent -- the above call to ip_mloopback() will
636 		 * loop back a copy. ip_input() will drop the copy if
637 		 * this host does not belong to the destination group on
638 		 * the loopback interface.
639 		 */
640 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
641 			m_freem(m);
642 			goto done;
643 		}
644 
645 		goto sendit;
646 	}
647 
648 	/*
649 	 * If the source address is not specified yet, use the address
650 	 * of the outoing interface.
651 	 */
652 	if (ip->ip_src.s_addr == INADDR_ANY)
653 		ip->ip_src = src;
654 
655 	/*
656 	 * Look for broadcast address and
657 	 * verify user is allowed to send
658 	 * such a packet.
659 	 */
660 	if (isbroadcast) {
661 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
662 			error = EADDRNOTAVAIL;
663 			goto bad;
664 		}
665 		if ((flags & IP_ALLOWBROADCAST) == 0) {
666 			error = EACCES;
667 			goto bad;
668 		}
669 		/* don't allow broadcast messages to be fragmented */
670 		if (ip_len > mtu) {
671 			error = EMSGSIZE;
672 			goto bad;
673 		}
674 		m->m_flags |= M_BCAST;
675 	} else {
676 		m->m_flags &= ~M_BCAST;
677 	}
678 
679 sendit:
680 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
681 	if (IPSEC_ENABLED(ipv4)) {
682 		struct ip ip_hdr;
683 
684 		if ((error = IPSEC_OUTPUT(ipv4, ifp, m, inp, mtu)) != 0) {
685 			if (error == EINPROGRESS)
686 				error = 0;
687 			goto done;
688 		}
689 
690 		/* Update variables that are affected by ipsec4_output(). */
691 		m_copydata(m, 0, sizeof(ip_hdr), (char *)&ip_hdr);
692 		hlen = ip_hdr.ip_hl << 2;
693 	}
694 
695 	/*
696 	 * Check if there was a route for this packet; return error if not.
697 	 */
698 	if (no_route_but_check_spd) {
699 		IPSTAT_INC(ips_noroute);
700 		error = EHOSTUNREACH;
701 		goto bad;
702 	}
703 #endif /* IPSEC */
704 
705 	/* Jump over all PFIL processing if hooks are not active. */
706 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
707 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
708 		    &error)) {
709 		case 1: /* Finished */
710 			goto done;
711 
712 		case 0: /* Continue normally */
713 			ip = mtod(m, struct ip *);
714 			ip_len = ntohs(ip->ip_len);
715 			break;
716 
717 		case -1: /* Need to try again */
718 			/* Reset everything for a new round */
719 			if (ro != NULL) {
720 				RO_NHFREE(ro);
721 				ro->ro_prepend = NULL;
722 			}
723 			gw = (const struct sockaddr *)dst;
724 			ip = mtod(m, struct ip *);
725 			goto again;
726 		}
727 	}
728 
729 	if (vlan_pcp > -1)
730 		EVL_APPLY_PRI(m, vlan_pcp);
731 
732 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
733 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
734 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
735 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
736 			IPSTAT_INC(ips_badaddr);
737 			error = EADDRNOTAVAIL;
738 			goto bad;
739 		}
740 	}
741 
742 	/* Ensure the packet data is mapped if the interface requires it. */
743 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
744 		struct mbuf *m1;
745 
746 		error = mb_unmapped_to_ext(m, &m1);
747 		if (error != 0) {
748 			if (error == EINVAL) {
749 				if_printf(ifp, "TLS packet\n");
750 				/* XXXKIB */
751 			} else if (error == ENOMEM) {
752 				error = ENOBUFS;
753 			}
754 			IPSTAT_INC(ips_odropped);
755 			goto done;
756 		} else {
757 			m = m1;
758 		}
759 	}
760 
761 	m->m_pkthdr.csum_flags |= CSUM_IP;
762 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
763 		in_delayed_cksum(m);
764 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
765 	}
766 #if defined(SCTP) || defined(SCTP_SUPPORT)
767 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
768 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
769 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
770 	}
771 #endif
772 
773 	/*
774 	 * If small enough for interface, or the interface will take
775 	 * care of the fragmentation for us, we can just send directly.
776 	 * Note that if_vxlan could have requested TSO even though the outer
777 	 * frame is UDP.  It is correct to not fragment such datagrams and
778 	 * instead just pass them on to the driver.
779 	 */
780 	if (ip_len <= mtu ||
781 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
782 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
783 		ip->ip_sum = 0;
784 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
785 			ip->ip_sum = in_cksum(m, hlen);
786 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
787 		}
788 
789 		/*
790 		 * Record statistics for this interface address.
791 		 * With CSUM_TSO the byte/packet count will be slightly
792 		 * incorrect because we count the IP+TCP headers only
793 		 * once instead of for every generated packet.
794 		 */
795 		if (!(flags & IP_FORWARDING) && ia) {
796 			if (m->m_pkthdr.csum_flags &
797 			    (CSUM_TSO | CSUM_INNER_TSO))
798 				counter_u64_add(ia->ia_ifa.ifa_opackets,
799 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
800 			else
801 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
802 
803 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
804 		}
805 #ifdef MBUF_STRESS_TEST
806 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
807 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
808 #endif
809 		/*
810 		 * Reset layer specific mbuf flags
811 		 * to avoid confusing lower layers.
812 		 */
813 		m_clrprotoflags(m);
814 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
815 		error = ip_output_send(inp, ifp, m, gw, ro,
816 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
817 		goto done;
818 	}
819 
820 	/* Balk when DF bit is set or the interface didn't support TSO. */
821 	if ((ip_off & IP_DF) ||
822 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
823 		error = EMSGSIZE;
824 		IPSTAT_INC(ips_cantfrag);
825 		goto bad;
826 	}
827 
828 	/*
829 	 * Too large for interface; fragment if possible. If successful,
830 	 * on return, m will point to a list of packets to be sent.
831 	 */
832 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
833 	if (error)
834 		goto bad;
835 	for (; m; m = m0) {
836 		m0 = m->m_nextpkt;
837 		m->m_nextpkt = 0;
838 		if (error == 0) {
839 			/* Record statistics for this interface address. */
840 			if (ia != NULL) {
841 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
842 				counter_u64_add(ia->ia_ifa.ifa_obytes,
843 				    m->m_pkthdr.len);
844 			}
845 			/*
846 			 * Reset layer specific mbuf flags
847 			 * to avoid confusing upper layers.
848 			 */
849 			m_clrprotoflags(m);
850 
851 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
852 			    mtod(m, struct ip *), NULL);
853 			error = ip_output_send(inp, ifp, m, gw, ro, true);
854 		} else
855 			m_freem(m);
856 	}
857 
858 	if (error == 0)
859 		IPSTAT_INC(ips_fragmented);
860 
861 done:
862 	return (error);
863 bad:
864 	m_freem(m);
865 	goto done;
866 }
867 
868 /*
869  * Create a chain of fragments which fit the given mtu. m_frag points to the
870  * mbuf to be fragmented; on return it points to the chain with the fragments.
871  * Return 0 if no error. If error, m_frag may contain a partially built
872  * chain of fragments that should be freed by the caller.
873  *
874  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
875  */
876 int
877 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
878     u_long if_hwassist_flags)
879 {
880 	int error = 0;
881 	int hlen = ip->ip_hl << 2;
882 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
883 	int off;
884 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
885 	int firstlen;
886 	struct mbuf **mnext;
887 	int nfrags;
888 	uint16_t ip_len, ip_off;
889 
890 	ip_len = ntohs(ip->ip_len);
891 	ip_off = ntohs(ip->ip_off);
892 
893 	/*
894 	 * Packet shall not have "Don't Fragment" flag and have at least 8
895 	 * bytes of payload.
896 	 */
897 	if (__predict_false((ip_off & IP_DF) || len < 8)) {
898 		IPSTAT_INC(ips_cantfrag);
899 		return (EMSGSIZE);
900 	}
901 
902 	/*
903 	 * If the interface will not calculate checksums on
904 	 * fragmented packets, then do it here.
905 	 */
906 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
907 		in_delayed_cksum(m0);
908 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
909 	}
910 #if defined(SCTP) || defined(SCTP_SUPPORT)
911 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
912 		sctp_delayed_cksum(m0, hlen);
913 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
914 	}
915 #endif
916 	if (len > PAGE_SIZE) {
917 		/*
918 		 * Fragment large datagrams such that each segment
919 		 * contains a multiple of PAGE_SIZE amount of data,
920 		 * plus headers. This enables a receiver to perform
921 		 * page-flipping zero-copy optimizations.
922 		 *
923 		 * XXX When does this help given that sender and receiver
924 		 * could have different page sizes, and also mtu could
925 		 * be less than the receiver's page size ?
926 		 */
927 		int newlen;
928 
929 		off = MIN(mtu, m0->m_pkthdr.len);
930 
931 		/*
932 		 * firstlen (off - hlen) must be aligned on an
933 		 * 8-byte boundary
934 		 */
935 		if (off < hlen)
936 			goto smart_frag_failure;
937 		off = ((off - hlen) & ~7) + hlen;
938 		newlen = (~PAGE_MASK) & mtu;
939 		if ((newlen + sizeof (struct ip)) > mtu) {
940 			/* we failed, go back the default */
941 smart_frag_failure:
942 			newlen = len;
943 			off = hlen + len;
944 		}
945 		len = newlen;
946 
947 	} else {
948 		off = hlen + len;
949 	}
950 
951 	firstlen = off - hlen;
952 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
953 
954 	/*
955 	 * Loop through length of segment after first fragment,
956 	 * make new header and copy data of each part and link onto chain.
957 	 * Here, m0 is the original packet, m is the fragment being created.
958 	 * The fragments are linked off the m_nextpkt of the original
959 	 * packet, which after processing serves as the first fragment.
960 	 */
961 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
962 		struct ip *mhip;	/* ip header on the fragment */
963 		struct mbuf *m;
964 		int mhlen = sizeof (struct ip);
965 
966 		m = m_gethdr(M_NOWAIT, MT_DATA);
967 		if (m == NULL) {
968 			error = ENOBUFS;
969 			IPSTAT_INC(ips_odropped);
970 			goto done;
971 		}
972 		/*
973 		 * Make sure the complete packet header gets copied
974 		 * from the originating mbuf to the newly created
975 		 * mbuf. This also ensures that existing firewall
976 		 * classification(s), VLAN tags and so on get copied
977 		 * to the resulting fragmented packet(s):
978 		 */
979 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
980 			m_free(m);
981 			error = ENOBUFS;
982 			IPSTAT_INC(ips_odropped);
983 			goto done;
984 		}
985 		/*
986 		 * In the first mbuf, leave room for the link header, then
987 		 * copy the original IP header including options. The payload
988 		 * goes into an additional mbuf chain returned by m_copym().
989 		 */
990 		m->m_data += max_linkhdr;
991 		mhip = mtod(m, struct ip *);
992 		*mhip = *ip;
993 		if (hlen > sizeof (struct ip)) {
994 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
995 			mhip->ip_v = IPVERSION;
996 			mhip->ip_hl = mhlen >> 2;
997 		}
998 		m->m_len = mhlen;
999 		/* XXX do we need to add ip_off below ? */
1000 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1001 		if (off + len >= ip_len)
1002 			len = ip_len - off;
1003 		else
1004 			mhip->ip_off |= IP_MF;
1005 		mhip->ip_len = htons((u_short)(len + mhlen));
1006 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1007 		if (m->m_next == NULL) {	/* copy failed */
1008 			m_free(m);
1009 			error = ENOBUFS;	/* ??? */
1010 			IPSTAT_INC(ips_odropped);
1011 			goto done;
1012 		}
1013 		m->m_pkthdr.len = mhlen + len;
1014 #ifdef MAC
1015 		mac_netinet_fragment(m0, m);
1016 #endif
1017 		mhip->ip_off = htons(mhip->ip_off);
1018 		mhip->ip_sum = 0;
1019 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1020 			mhip->ip_sum = in_cksum(m, mhlen);
1021 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1022 		}
1023 		*mnext = m;
1024 		mnext = &m->m_nextpkt;
1025 	}
1026 	IPSTAT_ADD(ips_ofragments, nfrags);
1027 
1028 	/*
1029 	 * Update first fragment by trimming what's been copied out
1030 	 * and updating header.
1031 	 */
1032 	m_adj(m0, hlen + firstlen - ip_len);
1033 	m0->m_pkthdr.len = hlen + firstlen;
1034 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1035 	ip->ip_off = htons(ip_off | IP_MF);
1036 	ip->ip_sum = 0;
1037 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1038 		ip->ip_sum = in_cksum(m0, hlen);
1039 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1040 	}
1041 
1042 done:
1043 	*m_frag = m0;
1044 	return error;
1045 }
1046 
1047 void
1048 in_delayed_cksum_o(struct mbuf *m, uint16_t iph_offset)
1049 {
1050 	struct ip *ip;
1051 	struct udphdr *uh;
1052 	uint16_t cklen, csum, offset;
1053 
1054 	ip = (struct ip *)mtodo(m, iph_offset);
1055 	offset = iph_offset + (ip->ip_hl << 2);
1056 
1057 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1058 		/* if udp header is not in the first mbuf copy udplen */
1059 		if (offset + sizeof(struct udphdr) > m->m_len) {
1060 			m_copydata(m, offset + offsetof(struct udphdr,
1061 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1062 			cklen = ntohs(cklen);
1063 		} else {
1064 			uh = (struct udphdr *)mtodo(m, offset);
1065 			cklen = ntohs(uh->uh_ulen);
1066 		}
1067 		csum = in_cksum_skip(m, cklen + offset, offset);
1068 		if (csum == 0)
1069 			csum = 0xffff;
1070 	} else {
1071 		cklen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1072 		csum = in_cksum_skip(m, cklen + offset, offset);
1073 	}
1074 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1075 
1076 	if (offset + sizeof(csum) > m->m_len)
1077 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1078 	else
1079 		*(u_short *)mtodo(m, offset) = csum;
1080 }
1081 
1082 void
1083 in_delayed_cksum(struct mbuf *m)
1084 {
1085 
1086 	in_delayed_cksum_o(m, 0);
1087 }
1088 
1089 /*
1090  * IP socket option processing.
1091  */
1092 int
1093 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1094 {
1095 	struct inpcb *inp = sotoinpcb(so);
1096 	int	error, optval;
1097 #ifdef	RSS
1098 	uint32_t rss_bucket;
1099 	int retval;
1100 #endif
1101 
1102 	error = optval = 0;
1103 	if (sopt->sopt_level != IPPROTO_IP) {
1104 		error = EINVAL;
1105 
1106 		if (sopt->sopt_level == SOL_SOCKET &&
1107 		    sopt->sopt_dir == SOPT_SET) {
1108 			switch (sopt->sopt_name) {
1109 			case SO_SETFIB:
1110 				error = sooptcopyin(sopt, &optval,
1111 				    sizeof(optval), sizeof(optval));
1112 				if (error != 0)
1113 					break;
1114 
1115 				INP_WLOCK(inp);
1116 				if ((inp->inp_flags & INP_BOUNDFIB) != 0 &&
1117 				    optval != so->so_fibnum) {
1118 					INP_WUNLOCK(inp);
1119 					error = EISCONN;
1120 					break;
1121 				}
1122 				error = sosetfib(inp->inp_socket, optval);
1123 				if (error == 0)
1124 					inp->inp_inc.inc_fibnum = optval;
1125 				INP_WUNLOCK(inp);
1126 				break;
1127 			case SO_MAX_PACING_RATE:
1128 #ifdef RATELIMIT
1129 				INP_WLOCK(inp);
1130 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1131 				INP_WUNLOCK(inp);
1132 				error = 0;
1133 #else
1134 				error = EOPNOTSUPP;
1135 #endif
1136 				break;
1137 			default:
1138 				break;
1139 			}
1140 		}
1141 		return (error);
1142 	}
1143 
1144 	switch (sopt->sopt_dir) {
1145 	case SOPT_SET:
1146 		switch (sopt->sopt_name) {
1147 		case IP_OPTIONS:
1148 #ifdef notyet
1149 		case IP_RETOPTS:
1150 #endif
1151 		{
1152 			struct mbuf *m;
1153 			if (sopt->sopt_valsize > MLEN) {
1154 				error = EMSGSIZE;
1155 				break;
1156 			}
1157 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1158 			if (m == NULL) {
1159 				error = ENOBUFS;
1160 				break;
1161 			}
1162 			m->m_len = sopt->sopt_valsize;
1163 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1164 					    m->m_len);
1165 			if (error) {
1166 				m_free(m);
1167 				break;
1168 			}
1169 			INP_WLOCK(inp);
1170 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1171 			INP_WUNLOCK(inp);
1172 			return (error);
1173 		}
1174 
1175 		case IP_BINDANY:
1176 			if (sopt->sopt_td != NULL) {
1177 				error = priv_check(sopt->sopt_td,
1178 				    PRIV_NETINET_BINDANY);
1179 				if (error)
1180 					break;
1181 			}
1182 			/* FALLTHROUGH */
1183 		case IP_TOS:
1184 		case IP_TTL:
1185 		case IP_MINTTL:
1186 		case IP_RECVOPTS:
1187 		case IP_RECVRETOPTS:
1188 		case IP_ORIGDSTADDR:
1189 		case IP_RECVDSTADDR:
1190 		case IP_RECVTTL:
1191 		case IP_RECVIF:
1192 		case IP_ONESBCAST:
1193 		case IP_DONTFRAG:
1194 		case IP_RECVTOS:
1195 		case IP_RECVFLOWID:
1196 #ifdef	RSS
1197 		case IP_RECVRSSBUCKETID:
1198 #endif
1199 		case IP_VLAN_PCP:
1200 			error = sooptcopyin(sopt, &optval, sizeof optval,
1201 					    sizeof optval);
1202 			if (error)
1203 				break;
1204 
1205 			switch (sopt->sopt_name) {
1206 			case IP_TOS:
1207 				inp->inp_ip_tos = optval;
1208 				break;
1209 
1210 			case IP_TTL:
1211 				inp->inp_ip_ttl = optval;
1212 				break;
1213 
1214 			case IP_MINTTL:
1215 				if (optval >= 0 && optval <= MAXTTL)
1216 					inp->inp_ip_minttl = optval;
1217 				else
1218 					error = EINVAL;
1219 				break;
1220 
1221 #define	OPTSET(bit) do {						\
1222 	INP_WLOCK(inp);							\
1223 	if (optval)							\
1224 		inp->inp_flags |= bit;					\
1225 	else								\
1226 		inp->inp_flags &= ~bit;					\
1227 	INP_WUNLOCK(inp);						\
1228 } while (0)
1229 
1230 #define	OPTSET2(bit, val) do {						\
1231 	INP_WLOCK(inp);							\
1232 	if (val)							\
1233 		inp->inp_flags2 |= bit;					\
1234 	else								\
1235 		inp->inp_flags2 &= ~bit;				\
1236 	INP_WUNLOCK(inp);						\
1237 } while (0)
1238 
1239 			case IP_RECVOPTS:
1240 				OPTSET(INP_RECVOPTS);
1241 				break;
1242 
1243 			case IP_RECVRETOPTS:
1244 				OPTSET(INP_RECVRETOPTS);
1245 				break;
1246 
1247 			case IP_RECVDSTADDR:
1248 				OPTSET(INP_RECVDSTADDR);
1249 				break;
1250 
1251 			case IP_ORIGDSTADDR:
1252 				OPTSET2(INP_ORIGDSTADDR, optval);
1253 				break;
1254 
1255 			case IP_RECVTTL:
1256 				OPTSET(INP_RECVTTL);
1257 				break;
1258 
1259 			case IP_RECVIF:
1260 				OPTSET(INP_RECVIF);
1261 				break;
1262 
1263 			case IP_ONESBCAST:
1264 				OPTSET(INP_ONESBCAST);
1265 				break;
1266 			case IP_DONTFRAG:
1267 				OPTSET(INP_DONTFRAG);
1268 				break;
1269 			case IP_BINDANY:
1270 				OPTSET(INP_BINDANY);
1271 				break;
1272 			case IP_RECVTOS:
1273 				OPTSET(INP_RECVTOS);
1274 				break;
1275 			case IP_RECVFLOWID:
1276 				OPTSET2(INP_RECVFLOWID, optval);
1277 				break;
1278 #ifdef RSS
1279 			case IP_RECVRSSBUCKETID:
1280 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1281 				break;
1282 #endif
1283 			case IP_VLAN_PCP:
1284 				if ((optval >= -1) && (optval <=
1285 				    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1286 					if (optval == -1) {
1287 						INP_WLOCK(inp);
1288 						inp->inp_flags2 &=
1289 						    ~(INP_2PCP_SET |
1290 						      INP_2PCP_MASK);
1291 						INP_WUNLOCK(inp);
1292 					} else {
1293 						INP_WLOCK(inp);
1294 						inp->inp_flags2 |=
1295 						    INP_2PCP_SET;
1296 						inp->inp_flags2 &=
1297 						    ~INP_2PCP_MASK;
1298 						inp->inp_flags2 |=
1299 						    optval << INP_2PCP_SHIFT;
1300 						INP_WUNLOCK(inp);
1301 					}
1302 				} else
1303 					error = EINVAL;
1304 				break;
1305 			}
1306 			break;
1307 #undef OPTSET
1308 #undef OPTSET2
1309 
1310 		/*
1311 		 * Multicast socket options are processed by the in_mcast
1312 		 * module.
1313 		 */
1314 		case IP_MULTICAST_IF:
1315 		case IP_MULTICAST_VIF:
1316 		case IP_MULTICAST_TTL:
1317 		case IP_MULTICAST_LOOP:
1318 		case IP_ADD_MEMBERSHIP:
1319 		case IP_DROP_MEMBERSHIP:
1320 		case IP_ADD_SOURCE_MEMBERSHIP:
1321 		case IP_DROP_SOURCE_MEMBERSHIP:
1322 		case IP_BLOCK_SOURCE:
1323 		case IP_UNBLOCK_SOURCE:
1324 		case IP_MSFILTER:
1325 		case MCAST_JOIN_GROUP:
1326 		case MCAST_LEAVE_GROUP:
1327 		case MCAST_JOIN_SOURCE_GROUP:
1328 		case MCAST_LEAVE_SOURCE_GROUP:
1329 		case MCAST_BLOCK_SOURCE:
1330 		case MCAST_UNBLOCK_SOURCE:
1331 			error = inp_setmoptions(inp, sopt);
1332 			break;
1333 
1334 		case IP_PORTRANGE:
1335 			error = sooptcopyin(sopt, &optval, sizeof optval,
1336 					    sizeof optval);
1337 			if (error)
1338 				break;
1339 
1340 			INP_WLOCK(inp);
1341 			switch (optval) {
1342 			case IP_PORTRANGE_DEFAULT:
1343 				inp->inp_flags &= ~(INP_LOWPORT);
1344 				inp->inp_flags &= ~(INP_HIGHPORT);
1345 				break;
1346 
1347 			case IP_PORTRANGE_HIGH:
1348 				inp->inp_flags &= ~(INP_LOWPORT);
1349 				inp->inp_flags |= INP_HIGHPORT;
1350 				break;
1351 
1352 			case IP_PORTRANGE_LOW:
1353 				inp->inp_flags &= ~(INP_HIGHPORT);
1354 				inp->inp_flags |= INP_LOWPORT;
1355 				break;
1356 
1357 			default:
1358 				error = EINVAL;
1359 				break;
1360 			}
1361 			INP_WUNLOCK(inp);
1362 			break;
1363 
1364 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1365 		case IP_IPSEC_POLICY:
1366 			if (IPSEC_ENABLED(ipv4)) {
1367 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1368 				break;
1369 			}
1370 			/* FALLTHROUGH */
1371 #endif /* IPSEC */
1372 
1373 		default:
1374 			error = ENOPROTOOPT;
1375 			break;
1376 		}
1377 		break;
1378 
1379 	case SOPT_GET:
1380 		switch (sopt->sopt_name) {
1381 		case IP_OPTIONS:
1382 		case IP_RETOPTS:
1383 			INP_RLOCK(inp);
1384 			if (inp->inp_options) {
1385 				struct mbuf *options;
1386 
1387 				options = m_copym(inp->inp_options, 0,
1388 				    M_COPYALL, M_NOWAIT);
1389 				INP_RUNLOCK(inp);
1390 				if (options != NULL) {
1391 					error = sooptcopyout(sopt,
1392 							     mtod(options, char *),
1393 							     options->m_len);
1394 					m_freem(options);
1395 				} else
1396 					error = ENOMEM;
1397 			} else {
1398 				INP_RUNLOCK(inp);
1399 				sopt->sopt_valsize = 0;
1400 			}
1401 			break;
1402 
1403 		case IP_TOS:
1404 		case IP_TTL:
1405 		case IP_MINTTL:
1406 		case IP_RECVOPTS:
1407 		case IP_RECVRETOPTS:
1408 		case IP_ORIGDSTADDR:
1409 		case IP_RECVDSTADDR:
1410 		case IP_RECVTTL:
1411 		case IP_RECVIF:
1412 		case IP_PORTRANGE:
1413 		case IP_ONESBCAST:
1414 		case IP_DONTFRAG:
1415 		case IP_BINDANY:
1416 		case IP_RECVTOS:
1417 		case IP_FLOWID:
1418 		case IP_FLOWTYPE:
1419 		case IP_RECVFLOWID:
1420 #ifdef	RSS
1421 		case IP_RSSBUCKETID:
1422 		case IP_RECVRSSBUCKETID:
1423 #endif
1424 		case IP_VLAN_PCP:
1425 			switch (sopt->sopt_name) {
1426 			case IP_TOS:
1427 				optval = inp->inp_ip_tos;
1428 				break;
1429 
1430 			case IP_TTL:
1431 				optval = inp->inp_ip_ttl;
1432 				break;
1433 
1434 			case IP_MINTTL:
1435 				optval = inp->inp_ip_minttl;
1436 				break;
1437 
1438 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1439 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1440 
1441 			case IP_RECVOPTS:
1442 				optval = OPTBIT(INP_RECVOPTS);
1443 				break;
1444 
1445 			case IP_RECVRETOPTS:
1446 				optval = OPTBIT(INP_RECVRETOPTS);
1447 				break;
1448 
1449 			case IP_RECVDSTADDR:
1450 				optval = OPTBIT(INP_RECVDSTADDR);
1451 				break;
1452 
1453 			case IP_ORIGDSTADDR:
1454 				optval = OPTBIT2(INP_ORIGDSTADDR);
1455 				break;
1456 
1457 			case IP_RECVTTL:
1458 				optval = OPTBIT(INP_RECVTTL);
1459 				break;
1460 
1461 			case IP_RECVIF:
1462 				optval = OPTBIT(INP_RECVIF);
1463 				break;
1464 
1465 			case IP_PORTRANGE:
1466 				if (inp->inp_flags & INP_HIGHPORT)
1467 					optval = IP_PORTRANGE_HIGH;
1468 				else if (inp->inp_flags & INP_LOWPORT)
1469 					optval = IP_PORTRANGE_LOW;
1470 				else
1471 					optval = 0;
1472 				break;
1473 
1474 			case IP_ONESBCAST:
1475 				optval = OPTBIT(INP_ONESBCAST);
1476 				break;
1477 			case IP_DONTFRAG:
1478 				optval = OPTBIT(INP_DONTFRAG);
1479 				break;
1480 			case IP_BINDANY:
1481 				optval = OPTBIT(INP_BINDANY);
1482 				break;
1483 			case IP_RECVTOS:
1484 				optval = OPTBIT(INP_RECVTOS);
1485 				break;
1486 			case IP_FLOWID:
1487 				optval = inp->inp_flowid;
1488 				break;
1489 			case IP_FLOWTYPE:
1490 				optval = inp->inp_flowtype;
1491 				break;
1492 			case IP_RECVFLOWID:
1493 				optval = OPTBIT2(INP_RECVFLOWID);
1494 				break;
1495 #ifdef	RSS
1496 			case IP_RSSBUCKETID:
1497 				retval = rss_hash2bucket(inp->inp_flowid,
1498 				    inp->inp_flowtype,
1499 				    &rss_bucket);
1500 				if (retval == 0)
1501 					optval = rss_bucket;
1502 				else
1503 					error = EINVAL;
1504 				break;
1505 			case IP_RECVRSSBUCKETID:
1506 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1507 				break;
1508 #endif
1509 			case IP_VLAN_PCP:
1510 				if (OPTBIT2(INP_2PCP_SET)) {
1511 					optval = (inp->inp_flags2 &
1512 					    INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1513 				} else {
1514 					optval = -1;
1515 				}
1516 				break;
1517 			}
1518 			error = sooptcopyout(sopt, &optval, sizeof optval);
1519 			break;
1520 
1521 		/*
1522 		 * Multicast socket options are processed by the in_mcast
1523 		 * module.
1524 		 */
1525 		case IP_MULTICAST_IF:
1526 		case IP_MULTICAST_VIF:
1527 		case IP_MULTICAST_TTL:
1528 		case IP_MULTICAST_LOOP:
1529 		case IP_MSFILTER:
1530 			error = inp_getmoptions(inp, sopt);
1531 			break;
1532 
1533 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1534 		case IP_IPSEC_POLICY:
1535 			if (IPSEC_ENABLED(ipv4)) {
1536 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1537 				break;
1538 			}
1539 			/* FALLTHROUGH */
1540 #endif /* IPSEC */
1541 
1542 		default:
1543 			error = ENOPROTOOPT;
1544 			break;
1545 		}
1546 		break;
1547 	}
1548 	return (error);
1549 }
1550 
1551 /*
1552  * Routine called from ip_output() to loop back a copy of an IP multicast
1553  * packet to the input queue of a specified interface.  Note that this
1554  * calls the output routine of the loopback "driver", but with an interface
1555  * pointer that might NOT be a loopback interface -- evil, but easier than
1556  * replicating that code here.
1557  */
1558 static void
1559 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1560 {
1561 	struct ip *ip;
1562 	struct mbuf *copym;
1563 
1564 	/*
1565 	 * Make a deep copy of the packet because we're going to
1566 	 * modify the pack in order to generate checksums.
1567 	 */
1568 	copym = m_dup(m, M_NOWAIT);
1569 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1570 		copym = m_pullup(copym, hlen);
1571 	if (copym != NULL) {
1572 		/* If needed, compute the checksum and mark it as valid. */
1573 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1574 			in_delayed_cksum(copym);
1575 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1576 			copym->m_pkthdr.csum_flags |=
1577 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1578 			copym->m_pkthdr.csum_data = 0xffff;
1579 		}
1580 		/*
1581 		 * We don't bother to fragment if the IP length is greater
1582 		 * than the interface's MTU.  Can this possibly matter?
1583 		 */
1584 		ip = mtod(copym, struct ip *);
1585 		ip->ip_sum = 0;
1586 		ip->ip_sum = in_cksum(copym, hlen);
1587 		if_simloop(ifp, copym, AF_INET, 0);
1588 	}
1589 }
1590