xref: /freebsd/sys/netinet/ip_output.c (revision 3c4ba5f55438f7afd4f4b0b56f88f2bb505fd6a6)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1990, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_inet.h"
38 #include "opt_ipsec.h"
39 #include "opt_kern_tls.h"
40 #include "opt_mbuf_stress_test.h"
41 #include "opt_ratelimit.h"
42 #include "opt_route.h"
43 #include "opt_rss.h"
44 #include "opt_sctp.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
49 #include <sys/ktls.h>
50 #include <sys/lock.h>
51 #include <sys/malloc.h>
52 #include <sys/mbuf.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/protosw.h>
56 #include <sys/sdt.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/sysctl.h>
60 #include <sys/ucred.h>
61 
62 #include <net/if.h>
63 #include <net/if_var.h>
64 #include <net/if_private.h>
65 #include <net/if_vlan_var.h>
66 #include <net/if_llatbl.h>
67 #include <net/ethernet.h>
68 #include <net/netisr.h>
69 #include <net/pfil.h>
70 #include <net/route.h>
71 #include <net/route/nhop.h>
72 #include <net/rss_config.h>
73 #include <net/vnet.h>
74 
75 #include <netinet/in.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_kdtrace.h>
78 #include <netinet/in_systm.h>
79 #include <netinet/ip.h>
80 #include <netinet/in_fib.h>
81 #include <netinet/in_pcb.h>
82 #include <netinet/in_rss.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip_var.h>
85 #include <netinet/ip_options.h>
86 
87 #include <netinet/udp.h>
88 #include <netinet/udp_var.h>
89 
90 #if defined(SCTP) || defined(SCTP_SUPPORT)
91 #include <netinet/sctp.h>
92 #include <netinet/sctp_crc32.h>
93 #endif
94 
95 #include <netipsec/ipsec_support.h>
96 
97 #include <machine/in_cksum.h>
98 
99 #include <security/mac/mac_framework.h>
100 
101 #ifdef MBUF_STRESS_TEST
102 static int mbuf_frag_size = 0;
103 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
104 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
105 #endif
106 
107 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
108 
109 extern int in_mcast_loop;
110 
111 static inline int
112 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
113     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
114 {
115 	struct m_tag *fwd_tag = NULL;
116 	struct mbuf *m;
117 	struct in_addr odst;
118 	struct ip *ip;
119 
120 	m = *mp;
121 	ip = mtod(m, struct ip *);
122 
123 	/* Run through list of hooks for output packets. */
124 	odst.s_addr = ip->ip_dst.s_addr;
125 	switch (pfil_mbuf_out(V_inet_pfil_head, mp, ifp, inp)) {
126 	case PFIL_DROPPED:
127 		*error = EACCES;
128 		/* FALLTHROUGH */
129 	case PFIL_CONSUMED:
130 		return 1; /* Finished */
131 	case PFIL_PASS:
132 		*error = 0;
133 	}
134 	m = *mp;
135 	ip = mtod(m, struct ip *);
136 
137 	/* See if destination IP address was changed by packet filter. */
138 	if (odst.s_addr != ip->ip_dst.s_addr) {
139 		m->m_flags |= M_SKIP_FIREWALL;
140 		/* If destination is now ourself drop to ip_input(). */
141 		if (in_localip(ip->ip_dst)) {
142 			m->m_flags |= M_FASTFWD_OURS;
143 			if (m->m_pkthdr.rcvif == NULL)
144 				m->m_pkthdr.rcvif = V_loif;
145 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
146 				m->m_pkthdr.csum_flags |=
147 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
148 				m->m_pkthdr.csum_data = 0xffff;
149 			}
150 			m->m_pkthdr.csum_flags |=
151 				CSUM_IP_CHECKED | CSUM_IP_VALID;
152 #if defined(SCTP) || defined(SCTP_SUPPORT)
153 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
154 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
155 #endif
156 			*error = netisr_queue(NETISR_IP, m);
157 			return 1; /* Finished */
158 		}
159 
160 		bzero(dst, sizeof(*dst));
161 		dst->sin_family = AF_INET;
162 		dst->sin_len = sizeof(*dst);
163 		dst->sin_addr = ip->ip_dst;
164 
165 		return -1; /* Reloop */
166 	}
167 	/* See if fib was changed by packet filter. */
168 	if ((*fibnum) != M_GETFIB(m)) {
169 		m->m_flags |= M_SKIP_FIREWALL;
170 		*fibnum = M_GETFIB(m);
171 		return -1; /* Reloop for FIB change */
172 	}
173 
174 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
175 	if (m->m_flags & M_FASTFWD_OURS) {
176 		if (m->m_pkthdr.rcvif == NULL)
177 			m->m_pkthdr.rcvif = V_loif;
178 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
179 			m->m_pkthdr.csum_flags |=
180 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
181 			m->m_pkthdr.csum_data = 0xffff;
182 		}
183 #if defined(SCTP) || defined(SCTP_SUPPORT)
184 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
185 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
186 #endif
187 		m->m_pkthdr.csum_flags |=
188 			CSUM_IP_CHECKED | CSUM_IP_VALID;
189 
190 		*error = netisr_queue(NETISR_IP, m);
191 		return 1; /* Finished */
192 	}
193 	/* Or forward to some other address? */
194 	if ((m->m_flags & M_IP_NEXTHOP) &&
195 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
196 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
197 		m->m_flags |= M_SKIP_FIREWALL;
198 		m->m_flags &= ~M_IP_NEXTHOP;
199 		m_tag_delete(m, fwd_tag);
200 
201 		return -1; /* Reloop for CHANGE of dst */
202 	}
203 
204 	return 0;
205 }
206 
207 static int
208 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
209     const struct sockaddr *gw, struct route *ro, bool stamp_tag)
210 {
211 #ifdef KERN_TLS
212 	struct ktls_session *tls = NULL;
213 #endif
214 	struct m_snd_tag *mst;
215 	int error;
216 
217 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
218 	mst = NULL;
219 
220 #ifdef KERN_TLS
221 	/*
222 	 * If this is an unencrypted TLS record, save a reference to
223 	 * the record.  This local reference is used to call
224 	 * ktls_output_eagain after the mbuf has been freed (thus
225 	 * dropping the mbuf's reference) in if_output.
226 	 */
227 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
228 		tls = ktls_hold(m->m_next->m_epg_tls);
229 		mst = tls->snd_tag;
230 
231 		/*
232 		 * If a TLS session doesn't have a valid tag, it must
233 		 * have had an earlier ifp mismatch, so drop this
234 		 * packet.
235 		 */
236 		if (mst == NULL) {
237 			m_freem(m);
238 			error = EAGAIN;
239 			goto done;
240 		}
241 		/*
242 		 * Always stamp tags that include NIC ktls.
243 		 */
244 		stamp_tag = true;
245 	}
246 #endif
247 #ifdef RATELIMIT
248 	if (inp != NULL && mst == NULL) {
249 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
250 		    (inp->inp_snd_tag != NULL &&
251 		    inp->inp_snd_tag->ifp != ifp))
252 			in_pcboutput_txrtlmt(inp, ifp, m);
253 
254 		if (inp->inp_snd_tag != NULL)
255 			mst = inp->inp_snd_tag;
256 	}
257 #endif
258 	if (stamp_tag && mst != NULL) {
259 		KASSERT(m->m_pkthdr.rcvif == NULL,
260 		    ("trying to add a send tag to a forwarded packet"));
261 		if (mst->ifp != ifp) {
262 			m_freem(m);
263 			error = EAGAIN;
264 			goto done;
265 		}
266 
267 		/* stamp send tag on mbuf */
268 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
269 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
270 	}
271 
272 	error = (*ifp->if_output)(ifp, m, gw, ro);
273 
274 done:
275 	/* Check for route change invalidating send tags. */
276 #ifdef KERN_TLS
277 	if (tls != NULL) {
278 		if (error == EAGAIN)
279 			error = ktls_output_eagain(inp, tls);
280 		ktls_free(tls);
281 	}
282 #endif
283 #ifdef RATELIMIT
284 	if (error == EAGAIN)
285 		in_pcboutput_eagain(inp);
286 #endif
287 	return (error);
288 }
289 
290 /* rte<>ro_flags translation */
291 static inline void
292 rt_update_ro_flags(struct route *ro, const struct nhop_object *nh)
293 {
294 	int nh_flags = nh->nh_flags;
295 
296 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
297 
298 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
299 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
300 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
301 }
302 
303 /*
304  * IP output.  The packet in mbuf chain m contains a skeletal IP
305  * header (with len, off, ttl, proto, tos, src, dst).
306  * The mbuf chain containing the packet will be freed.
307  * The mbuf opt, if present, will not be freed.
308  * If route ro is present and has ro_rt initialized, route lookup would be
309  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
310  * then result of route lookup is stored in ro->ro_rt.
311  *
312  * In the IP forwarding case, the packet will arrive with options already
313  * inserted, so must have a NULL opt pointer.
314  */
315 int
316 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
317     struct ip_moptions *imo, struct inpcb *inp)
318 {
319 	struct ip *ip;
320 	struct ifnet *ifp = NULL;	/* keep compiler happy */
321 	struct mbuf *m0;
322 	int hlen = sizeof (struct ip);
323 	int mtu = 0;
324 	int error = 0;
325 	int vlan_pcp = -1;
326 	struct sockaddr_in *dst;
327 	const struct sockaddr *gw;
328 	struct in_ifaddr *ia = NULL;
329 	struct in_addr src;
330 	int isbroadcast;
331 	uint16_t ip_len, ip_off;
332 	struct route iproute;
333 	uint32_t fibnum;
334 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
335 	int no_route_but_check_spd = 0;
336 #endif
337 
338 	M_ASSERTPKTHDR(m);
339 	NET_EPOCH_ASSERT();
340 
341 	if (inp != NULL) {
342 		INP_LOCK_ASSERT(inp);
343 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
344 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
345 			m->m_pkthdr.flowid = inp->inp_flowid;
346 			M_HASHTYPE_SET(m, inp->inp_flowtype);
347 		}
348 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
349 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
350 			    INP_2PCP_SHIFT;
351 #ifdef NUMA
352 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
353 #endif
354 	}
355 
356 	if (opt) {
357 		int len = 0;
358 		m = ip_insertoptions(m, opt, &len);
359 		if (len != 0)
360 			hlen = len; /* ip->ip_hl is updated above */
361 	}
362 	ip = mtod(m, struct ip *);
363 	ip_len = ntohs(ip->ip_len);
364 	ip_off = ntohs(ip->ip_off);
365 
366 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
367 		ip->ip_v = IPVERSION;
368 		ip->ip_hl = hlen >> 2;
369 		ip_fillid(ip);
370 	} else {
371 		/* Header already set, fetch hlen from there */
372 		hlen = ip->ip_hl << 2;
373 	}
374 	if ((flags & IP_FORWARDING) == 0)
375 		IPSTAT_INC(ips_localout);
376 
377 	/*
378 	 * dst/gw handling:
379 	 *
380 	 * gw is readonly but can point either to dst OR rt_gateway,
381 	 * therefore we need restore gw if we're redoing lookup.
382 	 */
383 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
384 	if (ro == NULL) {
385 		ro = &iproute;
386 		bzero(ro, sizeof (*ro));
387 	}
388 	dst = (struct sockaddr_in *)&ro->ro_dst;
389 	if (ro->ro_nh == NULL) {
390 		dst->sin_family = AF_INET;
391 		dst->sin_len = sizeof(*dst);
392 		dst->sin_addr = ip->ip_dst;
393 	}
394 	gw = (const struct sockaddr *)dst;
395 again:
396 	/*
397 	 * Validate route against routing table additions;
398 	 * a better/more specific route might have been added.
399 	 */
400 	if (inp != NULL && ro->ro_nh != NULL)
401 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
402 	/*
403 	 * If there is a cached route,
404 	 * check that it is to the same destination
405 	 * and is still up.  If not, free it and try again.
406 	 * The address family should also be checked in case of sharing the
407 	 * cache with IPv6.
408 	 * Also check whether routing cache needs invalidation.
409 	 */
410 	if (ro->ro_nh != NULL &&
411 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
412 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
413 		RO_INVALIDATE_CACHE(ro);
414 	ia = NULL;
415 	/*
416 	 * If routing to interface only, short circuit routing lookup.
417 	 * The use of an all-ones broadcast address implies this; an
418 	 * interface is specified by the broadcast address of an interface,
419 	 * or the destination address of a ptp interface.
420 	 */
421 	if (flags & IP_SENDONES) {
422 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
423 						      M_GETFIB(m)))) == NULL &&
424 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
425 						    M_GETFIB(m)))) == NULL) {
426 			IPSTAT_INC(ips_noroute);
427 			error = ENETUNREACH;
428 			goto bad;
429 		}
430 		ip->ip_dst.s_addr = INADDR_BROADCAST;
431 		dst->sin_addr = ip->ip_dst;
432 		ifp = ia->ia_ifp;
433 		mtu = ifp->if_mtu;
434 		ip->ip_ttl = 1;
435 		isbroadcast = 1;
436 		src = IA_SIN(ia)->sin_addr;
437 	} else if (flags & IP_ROUTETOIF) {
438 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
439 						    M_GETFIB(m)))) == NULL &&
440 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
441 						M_GETFIB(m)))) == NULL) {
442 			IPSTAT_INC(ips_noroute);
443 			error = ENETUNREACH;
444 			goto bad;
445 		}
446 		ifp = ia->ia_ifp;
447 		mtu = ifp->if_mtu;
448 		ip->ip_ttl = 1;
449 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
450 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
451 		src = IA_SIN(ia)->sin_addr;
452 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
453 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
454 		/*
455 		 * Bypass the normal routing lookup for multicast
456 		 * packets if the interface is specified.
457 		 */
458 		ifp = imo->imo_multicast_ifp;
459 		mtu = ifp->if_mtu;
460 		IFP_TO_IA(ifp, ia);
461 		isbroadcast = 0;	/* fool gcc */
462 		/* Interface may have no addresses. */
463 		if (ia != NULL)
464 			src = IA_SIN(ia)->sin_addr;
465 		else
466 			src.s_addr = INADDR_ANY;
467 	} else if (ro != &iproute) {
468 		if (ro->ro_nh == NULL) {
469 			/*
470 			 * We want to do any cloning requested by the link
471 			 * layer, as this is probably required in all cases
472 			 * for correct operation (as it is for ARP).
473 			 */
474 			uint32_t flowid;
475 			flowid = m->m_pkthdr.flowid;
476 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
477 			    NHR_REF, flowid);
478 
479 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
480 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
481 				/*
482 				 * There is no route for this packet, but it is
483 				 * possible that a matching SPD entry exists.
484 				 */
485 				no_route_but_check_spd = 1;
486 				goto sendit;
487 #endif
488 				IPSTAT_INC(ips_noroute);
489 				error = EHOSTUNREACH;
490 				goto bad;
491 			}
492 		}
493 		struct nhop_object *nh = ro->ro_nh;
494 
495 		ia = ifatoia(nh->nh_ifa);
496 		ifp = nh->nh_ifp;
497 		counter_u64_add(nh->nh_pksent, 1);
498 		rt_update_ro_flags(ro, nh);
499 		if (nh->nh_flags & NHF_GATEWAY)
500 			gw = &nh->gw_sa;
501 		if (nh->nh_flags & NHF_HOST)
502 			isbroadcast = (nh->nh_flags & NHF_BROADCAST);
503 		else if ((ifp->if_flags & IFF_BROADCAST) && (gw->sa_family == AF_INET))
504 			isbroadcast = in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia);
505 		else
506 			isbroadcast = 0;
507 		mtu = nh->nh_mtu;
508 		src = IA_SIN(ia)->sin_addr;
509 	} else {
510 		struct nhop_object *nh;
511 
512 		nh = fib4_lookup(M_GETFIB(m), dst->sin_addr, 0, NHR_NONE,
513 		    m->m_pkthdr.flowid);
514 		if (nh == NULL) {
515 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
516 			/*
517 			 * There is no route for this packet, but it is
518 			 * possible that a matching SPD entry exists.
519 			 */
520 			no_route_but_check_spd = 1;
521 			goto sendit;
522 #endif
523 			IPSTAT_INC(ips_noroute);
524 			error = EHOSTUNREACH;
525 			goto bad;
526 		}
527 		ifp = nh->nh_ifp;
528 		mtu = nh->nh_mtu;
529 		rt_update_ro_flags(ro, nh);
530 		if (nh->nh_flags & NHF_GATEWAY)
531 			gw = &nh->gw_sa;
532 		ia = ifatoia(nh->nh_ifa);
533 		src = IA_SIN(ia)->sin_addr;
534 		isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
535 		    (NHF_HOST | NHF_BROADCAST)) ||
536 		    ((ifp->if_flags & IFF_BROADCAST) &&
537 		    (gw->sa_family == AF_INET) &&
538 		    in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia)));
539 	}
540 
541 	/* Catch a possible divide by zero later. */
542 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
543 	    __func__, mtu, ro,
544 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
545 
546 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
547 		m->m_flags |= M_MCAST;
548 		/*
549 		 * IP destination address is multicast.  Make sure "gw"
550 		 * still points to the address in "ro".  (It may have been
551 		 * changed to point to a gateway address, above.)
552 		 */
553 		gw = (const struct sockaddr *)dst;
554 		/*
555 		 * See if the caller provided any multicast options
556 		 */
557 		if (imo != NULL) {
558 			ip->ip_ttl = imo->imo_multicast_ttl;
559 			if (imo->imo_multicast_vif != -1)
560 				ip->ip_src.s_addr =
561 				    ip_mcast_src ?
562 				    ip_mcast_src(imo->imo_multicast_vif) :
563 				    INADDR_ANY;
564 		} else
565 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
566 		/*
567 		 * Confirm that the outgoing interface supports multicast.
568 		 */
569 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
570 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
571 				IPSTAT_INC(ips_noroute);
572 				error = ENETUNREACH;
573 				goto bad;
574 			}
575 		}
576 		/*
577 		 * If source address not specified yet, use address
578 		 * of outgoing interface.
579 		 */
580 		if (ip->ip_src.s_addr == INADDR_ANY)
581 			ip->ip_src = src;
582 
583 		if ((imo == NULL && in_mcast_loop) ||
584 		    (imo && imo->imo_multicast_loop)) {
585 			/*
586 			 * Loop back multicast datagram if not expressly
587 			 * forbidden to do so, even if we are not a member
588 			 * of the group; ip_input() will filter it later,
589 			 * thus deferring a hash lookup and mutex acquisition
590 			 * at the expense of a cheap copy using m_copym().
591 			 */
592 			ip_mloopback(ifp, m, hlen);
593 		} else {
594 			/*
595 			 * If we are acting as a multicast router, perform
596 			 * multicast forwarding as if the packet had just
597 			 * arrived on the interface to which we are about
598 			 * to send.  The multicast forwarding function
599 			 * recursively calls this function, using the
600 			 * IP_FORWARDING flag to prevent infinite recursion.
601 			 *
602 			 * Multicasts that are looped back by ip_mloopback(),
603 			 * above, will be forwarded by the ip_input() routine,
604 			 * if necessary.
605 			 */
606 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
607 				/*
608 				 * If rsvp daemon is not running, do not
609 				 * set ip_moptions. This ensures that the packet
610 				 * is multicast and not just sent down one link
611 				 * as prescribed by rsvpd.
612 				 */
613 				if (!V_rsvp_on)
614 					imo = NULL;
615 				if (ip_mforward &&
616 				    ip_mforward(ip, ifp, m, imo) != 0) {
617 					m_freem(m);
618 					goto done;
619 				}
620 			}
621 		}
622 
623 		/*
624 		 * Multicasts with a time-to-live of zero may be looped-
625 		 * back, above, but must not be transmitted on a network.
626 		 * Also, multicasts addressed to the loopback interface
627 		 * are not sent -- the above call to ip_mloopback() will
628 		 * loop back a copy. ip_input() will drop the copy if
629 		 * this host does not belong to the destination group on
630 		 * the loopback interface.
631 		 */
632 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
633 			m_freem(m);
634 			goto done;
635 		}
636 
637 		goto sendit;
638 	}
639 
640 	/*
641 	 * If the source address is not specified yet, use the address
642 	 * of the outoing interface.
643 	 */
644 	if (ip->ip_src.s_addr == INADDR_ANY)
645 		ip->ip_src = src;
646 
647 	/*
648 	 * Look for broadcast address and
649 	 * verify user is allowed to send
650 	 * such a packet.
651 	 */
652 	if (isbroadcast) {
653 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
654 			error = EADDRNOTAVAIL;
655 			goto bad;
656 		}
657 		if ((flags & IP_ALLOWBROADCAST) == 0) {
658 			error = EACCES;
659 			goto bad;
660 		}
661 		/* don't allow broadcast messages to be fragmented */
662 		if (ip_len > mtu) {
663 			error = EMSGSIZE;
664 			goto bad;
665 		}
666 		m->m_flags |= M_BCAST;
667 	} else {
668 		m->m_flags &= ~M_BCAST;
669 	}
670 
671 sendit:
672 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
673 	if (IPSEC_ENABLED(ipv4)) {
674 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
675 			if (error == EINPROGRESS)
676 				error = 0;
677 			goto done;
678 		}
679 	}
680 	/*
681 	 * Check if there was a route for this packet; return error if not.
682 	 */
683 	if (no_route_but_check_spd) {
684 		IPSTAT_INC(ips_noroute);
685 		error = EHOSTUNREACH;
686 		goto bad;
687 	}
688 	/* Update variables that are affected by ipsec4_output(). */
689 	ip = mtod(m, struct ip *);
690 	hlen = ip->ip_hl << 2;
691 #endif /* IPSEC */
692 
693 	/* Jump over all PFIL processing if hooks are not active. */
694 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
695 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
696 		    &error)) {
697 		case 1: /* Finished */
698 			goto done;
699 
700 		case 0: /* Continue normally */
701 			ip = mtod(m, struct ip *);
702 			break;
703 
704 		case -1: /* Need to try again */
705 			/* Reset everything for a new round */
706 			if (ro != NULL) {
707 				RO_NHFREE(ro);
708 				ro->ro_prepend = NULL;
709 			}
710 			gw = (const struct sockaddr *)dst;
711 			ip = mtod(m, struct ip *);
712 			goto again;
713 		}
714 	}
715 
716 	if (vlan_pcp > -1)
717 		EVL_APPLY_PRI(m, vlan_pcp);
718 
719 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
720 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
721 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
722 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
723 			IPSTAT_INC(ips_badaddr);
724 			error = EADDRNOTAVAIL;
725 			goto bad;
726 		}
727 	}
728 
729 	/* Ensure the packet data is mapped if the interface requires it. */
730 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
731 		m = mb_unmapped_to_ext(m);
732 		if (m == NULL) {
733 			IPSTAT_INC(ips_odropped);
734 			error = ENOBUFS;
735 			goto bad;
736 		}
737 	}
738 
739 	m->m_pkthdr.csum_flags |= CSUM_IP;
740 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
741 		in_delayed_cksum(m);
742 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
743 	}
744 #if defined(SCTP) || defined(SCTP_SUPPORT)
745 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
746 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
747 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
748 	}
749 #endif
750 
751 	/*
752 	 * If small enough for interface, or the interface will take
753 	 * care of the fragmentation for us, we can just send directly.
754 	 * Note that if_vxlan could have requested TSO even though the outer
755 	 * frame is UDP.  It is correct to not fragment such datagrams and
756 	 * instead just pass them on to the driver.
757 	 */
758 	if (ip_len <= mtu ||
759 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
760 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
761 		ip->ip_sum = 0;
762 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
763 			ip->ip_sum = in_cksum(m, hlen);
764 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
765 		}
766 
767 		/*
768 		 * Record statistics for this interface address.
769 		 * With CSUM_TSO the byte/packet count will be slightly
770 		 * incorrect because we count the IP+TCP headers only
771 		 * once instead of for every generated packet.
772 		 */
773 		if (!(flags & IP_FORWARDING) && ia) {
774 			if (m->m_pkthdr.csum_flags &
775 			    (CSUM_TSO | CSUM_INNER_TSO))
776 				counter_u64_add(ia->ia_ifa.ifa_opackets,
777 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
778 			else
779 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
780 
781 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
782 		}
783 #ifdef MBUF_STRESS_TEST
784 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
785 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
786 #endif
787 		/*
788 		 * Reset layer specific mbuf flags
789 		 * to avoid confusing lower layers.
790 		 */
791 		m_clrprotoflags(m);
792 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
793 		error = ip_output_send(inp, ifp, m, gw, ro,
794 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
795 		goto done;
796 	}
797 
798 	/* Balk when DF bit is set or the interface didn't support TSO. */
799 	if ((ip_off & IP_DF) ||
800 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
801 		error = EMSGSIZE;
802 		IPSTAT_INC(ips_cantfrag);
803 		goto bad;
804 	}
805 
806 	/*
807 	 * Too large for interface; fragment if possible. If successful,
808 	 * on return, m will point to a list of packets to be sent.
809 	 */
810 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
811 	if (error)
812 		goto bad;
813 	for (; m; m = m0) {
814 		m0 = m->m_nextpkt;
815 		m->m_nextpkt = 0;
816 		if (error == 0) {
817 			/* Record statistics for this interface address. */
818 			if (ia != NULL) {
819 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
820 				counter_u64_add(ia->ia_ifa.ifa_obytes,
821 				    m->m_pkthdr.len);
822 			}
823 			/*
824 			 * Reset layer specific mbuf flags
825 			 * to avoid confusing upper layers.
826 			 */
827 			m_clrprotoflags(m);
828 
829 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
830 			    mtod(m, struct ip *), NULL);
831 			error = ip_output_send(inp, ifp, m, gw, ro, true);
832 		} else
833 			m_freem(m);
834 	}
835 
836 	if (error == 0)
837 		IPSTAT_INC(ips_fragmented);
838 
839 done:
840 	return (error);
841  bad:
842 	m_freem(m);
843 	goto done;
844 }
845 
846 /*
847  * Create a chain of fragments which fit the given mtu. m_frag points to the
848  * mbuf to be fragmented; on return it points to the chain with the fragments.
849  * Return 0 if no error. If error, m_frag may contain a partially built
850  * chain of fragments that should be freed by the caller.
851  *
852  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
853  */
854 int
855 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
856     u_long if_hwassist_flags)
857 {
858 	int error = 0;
859 	int hlen = ip->ip_hl << 2;
860 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
861 	int off;
862 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
863 	int firstlen;
864 	struct mbuf **mnext;
865 	int nfrags;
866 	uint16_t ip_len, ip_off;
867 
868 	ip_len = ntohs(ip->ip_len);
869 	ip_off = ntohs(ip->ip_off);
870 
871 	/*
872 	 * Packet shall not have "Don't Fragment" flag and have at least 8
873 	 * bytes of payload.
874 	 */
875 	if (__predict_false((ip_off & IP_DF) || len < 8)) {
876 		IPSTAT_INC(ips_cantfrag);
877 		return (EMSGSIZE);
878 	}
879 
880 	/*
881 	 * If the interface will not calculate checksums on
882 	 * fragmented packets, then do it here.
883 	 */
884 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
885 		in_delayed_cksum(m0);
886 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
887 	}
888 #if defined(SCTP) || defined(SCTP_SUPPORT)
889 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
890 		sctp_delayed_cksum(m0, hlen);
891 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
892 	}
893 #endif
894 	if (len > PAGE_SIZE) {
895 		/*
896 		 * Fragment large datagrams such that each segment
897 		 * contains a multiple of PAGE_SIZE amount of data,
898 		 * plus headers. This enables a receiver to perform
899 		 * page-flipping zero-copy optimizations.
900 		 *
901 		 * XXX When does this help given that sender and receiver
902 		 * could have different page sizes, and also mtu could
903 		 * be less than the receiver's page size ?
904 		 */
905 		int newlen;
906 
907 		off = MIN(mtu, m0->m_pkthdr.len);
908 
909 		/*
910 		 * firstlen (off - hlen) must be aligned on an
911 		 * 8-byte boundary
912 		 */
913 		if (off < hlen)
914 			goto smart_frag_failure;
915 		off = ((off - hlen) & ~7) + hlen;
916 		newlen = (~PAGE_MASK) & mtu;
917 		if ((newlen + sizeof (struct ip)) > mtu) {
918 			/* we failed, go back the default */
919 smart_frag_failure:
920 			newlen = len;
921 			off = hlen + len;
922 		}
923 		len = newlen;
924 
925 	} else {
926 		off = hlen + len;
927 	}
928 
929 	firstlen = off - hlen;
930 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
931 
932 	/*
933 	 * Loop through length of segment after first fragment,
934 	 * make new header and copy data of each part and link onto chain.
935 	 * Here, m0 is the original packet, m is the fragment being created.
936 	 * The fragments are linked off the m_nextpkt of the original
937 	 * packet, which after processing serves as the first fragment.
938 	 */
939 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
940 		struct ip *mhip;	/* ip header on the fragment */
941 		struct mbuf *m;
942 		int mhlen = sizeof (struct ip);
943 
944 		m = m_gethdr(M_NOWAIT, MT_DATA);
945 		if (m == NULL) {
946 			error = ENOBUFS;
947 			IPSTAT_INC(ips_odropped);
948 			goto done;
949 		}
950 		/*
951 		 * Make sure the complete packet header gets copied
952 		 * from the originating mbuf to the newly created
953 		 * mbuf. This also ensures that existing firewall
954 		 * classification(s), VLAN tags and so on get copied
955 		 * to the resulting fragmented packet(s):
956 		 */
957 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
958 			m_free(m);
959 			error = ENOBUFS;
960 			IPSTAT_INC(ips_odropped);
961 			goto done;
962 		}
963 		/*
964 		 * In the first mbuf, leave room for the link header, then
965 		 * copy the original IP header including options. The payload
966 		 * goes into an additional mbuf chain returned by m_copym().
967 		 */
968 		m->m_data += max_linkhdr;
969 		mhip = mtod(m, struct ip *);
970 		*mhip = *ip;
971 		if (hlen > sizeof (struct ip)) {
972 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
973 			mhip->ip_v = IPVERSION;
974 			mhip->ip_hl = mhlen >> 2;
975 		}
976 		m->m_len = mhlen;
977 		/* XXX do we need to add ip_off below ? */
978 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
979 		if (off + len >= ip_len)
980 			len = ip_len - off;
981 		else
982 			mhip->ip_off |= IP_MF;
983 		mhip->ip_len = htons((u_short)(len + mhlen));
984 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
985 		if (m->m_next == NULL) {	/* copy failed */
986 			m_free(m);
987 			error = ENOBUFS;	/* ??? */
988 			IPSTAT_INC(ips_odropped);
989 			goto done;
990 		}
991 		m->m_pkthdr.len = mhlen + len;
992 #ifdef MAC
993 		mac_netinet_fragment(m0, m);
994 #endif
995 		mhip->ip_off = htons(mhip->ip_off);
996 		mhip->ip_sum = 0;
997 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
998 			mhip->ip_sum = in_cksum(m, mhlen);
999 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1000 		}
1001 		*mnext = m;
1002 		mnext = &m->m_nextpkt;
1003 	}
1004 	IPSTAT_ADD(ips_ofragments, nfrags);
1005 
1006 	/*
1007 	 * Update first fragment by trimming what's been copied out
1008 	 * and updating header.
1009 	 */
1010 	m_adj(m0, hlen + firstlen - ip_len);
1011 	m0->m_pkthdr.len = hlen + firstlen;
1012 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1013 	ip->ip_off = htons(ip_off | IP_MF);
1014 	ip->ip_sum = 0;
1015 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1016 		ip->ip_sum = in_cksum(m0, hlen);
1017 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1018 	}
1019 
1020 done:
1021 	*m_frag = m0;
1022 	return error;
1023 }
1024 
1025 void
1026 in_delayed_cksum(struct mbuf *m)
1027 {
1028 	struct ip *ip;
1029 	struct udphdr *uh;
1030 	uint16_t cklen, csum, offset;
1031 
1032 	ip = mtod(m, struct ip *);
1033 	offset = ip->ip_hl << 2 ;
1034 
1035 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1036 		/* if udp header is not in the first mbuf copy udplen */
1037 		if (offset + sizeof(struct udphdr) > m->m_len) {
1038 			m_copydata(m, offset + offsetof(struct udphdr,
1039 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1040 			cklen = ntohs(cklen);
1041 		} else {
1042 			uh = (struct udphdr *)mtodo(m, offset);
1043 			cklen = ntohs(uh->uh_ulen);
1044 		}
1045 		csum = in_cksum_skip(m, cklen + offset, offset);
1046 		if (csum == 0)
1047 			csum = 0xffff;
1048 	} else {
1049 		cklen = ntohs(ip->ip_len);
1050 		csum = in_cksum_skip(m, cklen, offset);
1051 	}
1052 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1053 
1054 	if (offset + sizeof(csum) > m->m_len)
1055 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1056 	else
1057 		*(u_short *)mtodo(m, offset) = csum;
1058 }
1059 
1060 /*
1061  * IP socket option processing.
1062  */
1063 int
1064 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1065 {
1066 	struct inpcb *inp = sotoinpcb(so);
1067 	int	error, optval;
1068 #ifdef	RSS
1069 	uint32_t rss_bucket;
1070 	int retval;
1071 #endif
1072 
1073 	error = optval = 0;
1074 	if (sopt->sopt_level != IPPROTO_IP) {
1075 		error = EINVAL;
1076 
1077 		if (sopt->sopt_level == SOL_SOCKET &&
1078 		    sopt->sopt_dir == SOPT_SET) {
1079 			switch (sopt->sopt_name) {
1080 			case SO_REUSEADDR:
1081 				INP_WLOCK(inp);
1082 				if ((so->so_options & SO_REUSEADDR) != 0)
1083 					inp->inp_flags2 |= INP_REUSEADDR;
1084 				else
1085 					inp->inp_flags2 &= ~INP_REUSEADDR;
1086 				INP_WUNLOCK(inp);
1087 				error = 0;
1088 				break;
1089 			case SO_REUSEPORT:
1090 				INP_WLOCK(inp);
1091 				if ((so->so_options & SO_REUSEPORT) != 0)
1092 					inp->inp_flags2 |= INP_REUSEPORT;
1093 				else
1094 					inp->inp_flags2 &= ~INP_REUSEPORT;
1095 				INP_WUNLOCK(inp);
1096 				error = 0;
1097 				break;
1098 			case SO_REUSEPORT_LB:
1099 				INP_WLOCK(inp);
1100 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1101 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1102 				else
1103 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1104 				INP_WUNLOCK(inp);
1105 				error = 0;
1106 				break;
1107 			case SO_SETFIB:
1108 				INP_WLOCK(inp);
1109 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1110 				INP_WUNLOCK(inp);
1111 				error = 0;
1112 				break;
1113 			case SO_MAX_PACING_RATE:
1114 #ifdef RATELIMIT
1115 				INP_WLOCK(inp);
1116 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1117 				INP_WUNLOCK(inp);
1118 				error = 0;
1119 #else
1120 				error = EOPNOTSUPP;
1121 #endif
1122 				break;
1123 			default:
1124 				break;
1125 			}
1126 		}
1127 		return (error);
1128 	}
1129 
1130 	switch (sopt->sopt_dir) {
1131 	case SOPT_SET:
1132 		switch (sopt->sopt_name) {
1133 		case IP_OPTIONS:
1134 #ifdef notyet
1135 		case IP_RETOPTS:
1136 #endif
1137 		{
1138 			struct mbuf *m;
1139 			if (sopt->sopt_valsize > MLEN) {
1140 				error = EMSGSIZE;
1141 				break;
1142 			}
1143 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1144 			if (m == NULL) {
1145 				error = ENOBUFS;
1146 				break;
1147 			}
1148 			m->m_len = sopt->sopt_valsize;
1149 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1150 					    m->m_len);
1151 			if (error) {
1152 				m_free(m);
1153 				break;
1154 			}
1155 			INP_WLOCK(inp);
1156 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1157 			INP_WUNLOCK(inp);
1158 			return (error);
1159 		}
1160 
1161 		case IP_BINDANY:
1162 			if (sopt->sopt_td != NULL) {
1163 				error = priv_check(sopt->sopt_td,
1164 				    PRIV_NETINET_BINDANY);
1165 				if (error)
1166 					break;
1167 			}
1168 			/* FALLTHROUGH */
1169 		case IP_TOS:
1170 		case IP_TTL:
1171 		case IP_MINTTL:
1172 		case IP_RECVOPTS:
1173 		case IP_RECVRETOPTS:
1174 		case IP_ORIGDSTADDR:
1175 		case IP_RECVDSTADDR:
1176 		case IP_RECVTTL:
1177 		case IP_RECVIF:
1178 		case IP_ONESBCAST:
1179 		case IP_DONTFRAG:
1180 		case IP_RECVTOS:
1181 		case IP_RECVFLOWID:
1182 #ifdef	RSS
1183 		case IP_RECVRSSBUCKETID:
1184 #endif
1185 		case IP_VLAN_PCP:
1186 			error = sooptcopyin(sopt, &optval, sizeof optval,
1187 					    sizeof optval);
1188 			if (error)
1189 				break;
1190 
1191 			switch (sopt->sopt_name) {
1192 			case IP_TOS:
1193 				inp->inp_ip_tos = optval;
1194 				break;
1195 
1196 			case IP_TTL:
1197 				inp->inp_ip_ttl = optval;
1198 				break;
1199 
1200 			case IP_MINTTL:
1201 				if (optval >= 0 && optval <= MAXTTL)
1202 					inp->inp_ip_minttl = optval;
1203 				else
1204 					error = EINVAL;
1205 				break;
1206 
1207 #define	OPTSET(bit) do {						\
1208 	INP_WLOCK(inp);							\
1209 	if (optval)							\
1210 		inp->inp_flags |= bit;					\
1211 	else								\
1212 		inp->inp_flags &= ~bit;					\
1213 	INP_WUNLOCK(inp);						\
1214 } while (0)
1215 
1216 #define	OPTSET2(bit, val) do {						\
1217 	INP_WLOCK(inp);							\
1218 	if (val)							\
1219 		inp->inp_flags2 |= bit;					\
1220 	else								\
1221 		inp->inp_flags2 &= ~bit;				\
1222 	INP_WUNLOCK(inp);						\
1223 } while (0)
1224 
1225 			case IP_RECVOPTS:
1226 				OPTSET(INP_RECVOPTS);
1227 				break;
1228 
1229 			case IP_RECVRETOPTS:
1230 				OPTSET(INP_RECVRETOPTS);
1231 				break;
1232 
1233 			case IP_RECVDSTADDR:
1234 				OPTSET(INP_RECVDSTADDR);
1235 				break;
1236 
1237 			case IP_ORIGDSTADDR:
1238 				OPTSET2(INP_ORIGDSTADDR, optval);
1239 				break;
1240 
1241 			case IP_RECVTTL:
1242 				OPTSET(INP_RECVTTL);
1243 				break;
1244 
1245 			case IP_RECVIF:
1246 				OPTSET(INP_RECVIF);
1247 				break;
1248 
1249 			case IP_ONESBCAST:
1250 				OPTSET(INP_ONESBCAST);
1251 				break;
1252 			case IP_DONTFRAG:
1253 				OPTSET(INP_DONTFRAG);
1254 				break;
1255 			case IP_BINDANY:
1256 				OPTSET(INP_BINDANY);
1257 				break;
1258 			case IP_RECVTOS:
1259 				OPTSET(INP_RECVTOS);
1260 				break;
1261 			case IP_RECVFLOWID:
1262 				OPTSET2(INP_RECVFLOWID, optval);
1263 				break;
1264 #ifdef RSS
1265 			case IP_RECVRSSBUCKETID:
1266 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1267 				break;
1268 #endif
1269 			case IP_VLAN_PCP:
1270 				if ((optval >= -1) && (optval <=
1271 				    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1272 					if (optval == -1) {
1273 						INP_WLOCK(inp);
1274 						inp->inp_flags2 &=
1275 						    ~(INP_2PCP_SET |
1276 						      INP_2PCP_MASK);
1277 						INP_WUNLOCK(inp);
1278 					} else {
1279 						INP_WLOCK(inp);
1280 						inp->inp_flags2 |=
1281 						    INP_2PCP_SET;
1282 						inp->inp_flags2 &=
1283 						    ~INP_2PCP_MASK;
1284 						inp->inp_flags2 |=
1285 						    optval << INP_2PCP_SHIFT;
1286 						INP_WUNLOCK(inp);
1287 					}
1288 				} else
1289 					error = EINVAL;
1290 				break;
1291 			}
1292 			break;
1293 #undef OPTSET
1294 #undef OPTSET2
1295 
1296 		/*
1297 		 * Multicast socket options are processed by the in_mcast
1298 		 * module.
1299 		 */
1300 		case IP_MULTICAST_IF:
1301 		case IP_MULTICAST_VIF:
1302 		case IP_MULTICAST_TTL:
1303 		case IP_MULTICAST_LOOP:
1304 		case IP_ADD_MEMBERSHIP:
1305 		case IP_DROP_MEMBERSHIP:
1306 		case IP_ADD_SOURCE_MEMBERSHIP:
1307 		case IP_DROP_SOURCE_MEMBERSHIP:
1308 		case IP_BLOCK_SOURCE:
1309 		case IP_UNBLOCK_SOURCE:
1310 		case IP_MSFILTER:
1311 		case MCAST_JOIN_GROUP:
1312 		case MCAST_LEAVE_GROUP:
1313 		case MCAST_JOIN_SOURCE_GROUP:
1314 		case MCAST_LEAVE_SOURCE_GROUP:
1315 		case MCAST_BLOCK_SOURCE:
1316 		case MCAST_UNBLOCK_SOURCE:
1317 			error = inp_setmoptions(inp, sopt);
1318 			break;
1319 
1320 		case IP_PORTRANGE:
1321 			error = sooptcopyin(sopt, &optval, sizeof optval,
1322 					    sizeof optval);
1323 			if (error)
1324 				break;
1325 
1326 			INP_WLOCK(inp);
1327 			switch (optval) {
1328 			case IP_PORTRANGE_DEFAULT:
1329 				inp->inp_flags &= ~(INP_LOWPORT);
1330 				inp->inp_flags &= ~(INP_HIGHPORT);
1331 				break;
1332 
1333 			case IP_PORTRANGE_HIGH:
1334 				inp->inp_flags &= ~(INP_LOWPORT);
1335 				inp->inp_flags |= INP_HIGHPORT;
1336 				break;
1337 
1338 			case IP_PORTRANGE_LOW:
1339 				inp->inp_flags &= ~(INP_HIGHPORT);
1340 				inp->inp_flags |= INP_LOWPORT;
1341 				break;
1342 
1343 			default:
1344 				error = EINVAL;
1345 				break;
1346 			}
1347 			INP_WUNLOCK(inp);
1348 			break;
1349 
1350 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1351 		case IP_IPSEC_POLICY:
1352 			if (IPSEC_ENABLED(ipv4)) {
1353 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1354 				break;
1355 			}
1356 			/* FALLTHROUGH */
1357 #endif /* IPSEC */
1358 
1359 		default:
1360 			error = ENOPROTOOPT;
1361 			break;
1362 		}
1363 		break;
1364 
1365 	case SOPT_GET:
1366 		switch (sopt->sopt_name) {
1367 		case IP_OPTIONS:
1368 		case IP_RETOPTS:
1369 			INP_RLOCK(inp);
1370 			if (inp->inp_options) {
1371 				struct mbuf *options;
1372 
1373 				options = m_copym(inp->inp_options, 0,
1374 				    M_COPYALL, M_NOWAIT);
1375 				INP_RUNLOCK(inp);
1376 				if (options != NULL) {
1377 					error = sooptcopyout(sopt,
1378 							     mtod(options, char *),
1379 							     options->m_len);
1380 					m_freem(options);
1381 				} else
1382 					error = ENOMEM;
1383 			} else {
1384 				INP_RUNLOCK(inp);
1385 				sopt->sopt_valsize = 0;
1386 			}
1387 			break;
1388 
1389 		case IP_TOS:
1390 		case IP_TTL:
1391 		case IP_MINTTL:
1392 		case IP_RECVOPTS:
1393 		case IP_RECVRETOPTS:
1394 		case IP_ORIGDSTADDR:
1395 		case IP_RECVDSTADDR:
1396 		case IP_RECVTTL:
1397 		case IP_RECVIF:
1398 		case IP_PORTRANGE:
1399 		case IP_ONESBCAST:
1400 		case IP_DONTFRAG:
1401 		case IP_BINDANY:
1402 		case IP_RECVTOS:
1403 		case IP_FLOWID:
1404 		case IP_FLOWTYPE:
1405 		case IP_RECVFLOWID:
1406 #ifdef	RSS
1407 		case IP_RSSBUCKETID:
1408 		case IP_RECVRSSBUCKETID:
1409 #endif
1410 		case IP_VLAN_PCP:
1411 			switch (sopt->sopt_name) {
1412 			case IP_TOS:
1413 				optval = inp->inp_ip_tos;
1414 				break;
1415 
1416 			case IP_TTL:
1417 				optval = inp->inp_ip_ttl;
1418 				break;
1419 
1420 			case IP_MINTTL:
1421 				optval = inp->inp_ip_minttl;
1422 				break;
1423 
1424 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1425 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1426 
1427 			case IP_RECVOPTS:
1428 				optval = OPTBIT(INP_RECVOPTS);
1429 				break;
1430 
1431 			case IP_RECVRETOPTS:
1432 				optval = OPTBIT(INP_RECVRETOPTS);
1433 				break;
1434 
1435 			case IP_RECVDSTADDR:
1436 				optval = OPTBIT(INP_RECVDSTADDR);
1437 				break;
1438 
1439 			case IP_ORIGDSTADDR:
1440 				optval = OPTBIT2(INP_ORIGDSTADDR);
1441 				break;
1442 
1443 			case IP_RECVTTL:
1444 				optval = OPTBIT(INP_RECVTTL);
1445 				break;
1446 
1447 			case IP_RECVIF:
1448 				optval = OPTBIT(INP_RECVIF);
1449 				break;
1450 
1451 			case IP_PORTRANGE:
1452 				if (inp->inp_flags & INP_HIGHPORT)
1453 					optval = IP_PORTRANGE_HIGH;
1454 				else if (inp->inp_flags & INP_LOWPORT)
1455 					optval = IP_PORTRANGE_LOW;
1456 				else
1457 					optval = 0;
1458 				break;
1459 
1460 			case IP_ONESBCAST:
1461 				optval = OPTBIT(INP_ONESBCAST);
1462 				break;
1463 			case IP_DONTFRAG:
1464 				optval = OPTBIT(INP_DONTFRAG);
1465 				break;
1466 			case IP_BINDANY:
1467 				optval = OPTBIT(INP_BINDANY);
1468 				break;
1469 			case IP_RECVTOS:
1470 				optval = OPTBIT(INP_RECVTOS);
1471 				break;
1472 			case IP_FLOWID:
1473 				optval = inp->inp_flowid;
1474 				break;
1475 			case IP_FLOWTYPE:
1476 				optval = inp->inp_flowtype;
1477 				break;
1478 			case IP_RECVFLOWID:
1479 				optval = OPTBIT2(INP_RECVFLOWID);
1480 				break;
1481 #ifdef	RSS
1482 			case IP_RSSBUCKETID:
1483 				retval = rss_hash2bucket(inp->inp_flowid,
1484 				    inp->inp_flowtype,
1485 				    &rss_bucket);
1486 				if (retval == 0)
1487 					optval = rss_bucket;
1488 				else
1489 					error = EINVAL;
1490 				break;
1491 			case IP_RECVRSSBUCKETID:
1492 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1493 				break;
1494 #endif
1495 			case IP_VLAN_PCP:
1496 				if (OPTBIT2(INP_2PCP_SET)) {
1497 					optval = (inp->inp_flags2 &
1498 					    INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1499 				} else {
1500 					optval = -1;
1501 				}
1502 				break;
1503 			}
1504 			error = sooptcopyout(sopt, &optval, sizeof optval);
1505 			break;
1506 
1507 		/*
1508 		 * Multicast socket options are processed by the in_mcast
1509 		 * module.
1510 		 */
1511 		case IP_MULTICAST_IF:
1512 		case IP_MULTICAST_VIF:
1513 		case IP_MULTICAST_TTL:
1514 		case IP_MULTICAST_LOOP:
1515 		case IP_MSFILTER:
1516 			error = inp_getmoptions(inp, sopt);
1517 			break;
1518 
1519 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1520 		case IP_IPSEC_POLICY:
1521 			if (IPSEC_ENABLED(ipv4)) {
1522 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1523 				break;
1524 			}
1525 			/* FALLTHROUGH */
1526 #endif /* IPSEC */
1527 
1528 		default:
1529 			error = ENOPROTOOPT;
1530 			break;
1531 		}
1532 		break;
1533 	}
1534 	return (error);
1535 }
1536 
1537 /*
1538  * Routine called from ip_output() to loop back a copy of an IP multicast
1539  * packet to the input queue of a specified interface.  Note that this
1540  * calls the output routine of the loopback "driver", but with an interface
1541  * pointer that might NOT be a loopback interface -- evil, but easier than
1542  * replicating that code here.
1543  */
1544 static void
1545 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1546 {
1547 	struct ip *ip;
1548 	struct mbuf *copym;
1549 
1550 	/*
1551 	 * Make a deep copy of the packet because we're going to
1552 	 * modify the pack in order to generate checksums.
1553 	 */
1554 	copym = m_dup(m, M_NOWAIT);
1555 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1556 		copym = m_pullup(copym, hlen);
1557 	if (copym != NULL) {
1558 		/* If needed, compute the checksum and mark it as valid. */
1559 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1560 			in_delayed_cksum(copym);
1561 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1562 			copym->m_pkthdr.csum_flags |=
1563 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1564 			copym->m_pkthdr.csum_data = 0xffff;
1565 		}
1566 		/*
1567 		 * We don't bother to fragment if the IP length is greater
1568 		 * than the interface's MTU.  Can this possibly matter?
1569 		 */
1570 		ip = mtod(copym, struct ip *);
1571 		ip->ip_sum = 0;
1572 		ip->ip_sum = in_cksum(copym, hlen);
1573 		if_simloop(ifp, copym, AF_INET, 0);
1574 	}
1575 }
1576