xref: /freebsd/sys/netinet/ip_output.c (revision 3cbb4cc200f8a0ad7ed08233425ea54524a21f1c)
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/rmlock.h>
57 #include <sys/sdt.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
60 #include <sys/sysctl.h>
61 #include <sys/ucred.h>
62 
63 #include <net/if.h>
64 #include <net/if_var.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 extern	struct protosw inetsw[];
111 
112 static inline int
113 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
114     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
115 {
116 	struct m_tag *fwd_tag = NULL;
117 	struct mbuf *m;
118 	struct in_addr odst;
119 	struct ip *ip;
120 	int pflags = PFIL_OUT;
121 
122 	if (flags & IP_FORWARDING)
123 		pflags |= PFIL_FWD;
124 
125 	m = *mp;
126 	ip = mtod(m, struct ip *);
127 
128 	/* Run through list of hooks for output packets. */
129 	odst.s_addr = ip->ip_dst.s_addr;
130 	switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
131 	case PFIL_DROPPED:
132 		*error = EACCES;
133 		/* FALLTHROUGH */
134 	case PFIL_CONSUMED:
135 		return 1; /* Finished */
136 	case PFIL_PASS:
137 		*error = 0;
138 	}
139 	m = *mp;
140 	ip = mtod(m, struct ip *);
141 
142 	/* See if destination IP address was changed by packet filter. */
143 	if (odst.s_addr != ip->ip_dst.s_addr) {
144 		m->m_flags |= M_SKIP_FIREWALL;
145 		/* If destination is now ourself drop to ip_input(). */
146 		if (in_localip(ip->ip_dst)) {
147 			m->m_flags |= M_FASTFWD_OURS;
148 			if (m->m_pkthdr.rcvif == NULL)
149 				m->m_pkthdr.rcvif = V_loif;
150 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
151 				m->m_pkthdr.csum_flags |=
152 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
153 				m->m_pkthdr.csum_data = 0xffff;
154 			}
155 			m->m_pkthdr.csum_flags |=
156 				CSUM_IP_CHECKED | CSUM_IP_VALID;
157 #if defined(SCTP) || defined(SCTP_SUPPORT)
158 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
159 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
160 #endif
161 			*error = netisr_queue(NETISR_IP, m);
162 			return 1; /* Finished */
163 		}
164 
165 		bzero(dst, sizeof(*dst));
166 		dst->sin_family = AF_INET;
167 		dst->sin_len = sizeof(*dst);
168 		dst->sin_addr = ip->ip_dst;
169 
170 		return -1; /* Reloop */
171 	}
172 	/* See if fib was changed by packet filter. */
173 	if ((*fibnum) != M_GETFIB(m)) {
174 		m->m_flags |= M_SKIP_FIREWALL;
175 		*fibnum = M_GETFIB(m);
176 		return -1; /* Reloop for FIB change */
177 	}
178 
179 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
180 	if (m->m_flags & M_FASTFWD_OURS) {
181 		if (m->m_pkthdr.rcvif == NULL)
182 			m->m_pkthdr.rcvif = V_loif;
183 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
184 			m->m_pkthdr.csum_flags |=
185 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
186 			m->m_pkthdr.csum_data = 0xffff;
187 		}
188 #if defined(SCTP) || defined(SCTP_SUPPORT)
189 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
190 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
191 #endif
192 		m->m_pkthdr.csum_flags |=
193 			CSUM_IP_CHECKED | CSUM_IP_VALID;
194 
195 		*error = netisr_queue(NETISR_IP, m);
196 		return 1; /* Finished */
197 	}
198 	/* Or forward to some other address? */
199 	if ((m->m_flags & M_IP_NEXTHOP) &&
200 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
201 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
202 		m->m_flags |= M_SKIP_FIREWALL;
203 		m->m_flags &= ~M_IP_NEXTHOP;
204 		m_tag_delete(m, fwd_tag);
205 
206 		return -1; /* Reloop for CHANGE of dst */
207 	}
208 
209 	return 0;
210 }
211 
212 static int
213 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
214     const struct sockaddr_in *gw, struct route *ro, bool stamp_tag)
215 {
216 #ifdef KERN_TLS
217 	struct ktls_session *tls = NULL;
218 #endif
219 	struct m_snd_tag *mst;
220 	int error;
221 
222 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
223 	mst = NULL;
224 
225 #ifdef KERN_TLS
226 	/*
227 	 * If this is an unencrypted TLS record, save a reference to
228 	 * the record.  This local reference is used to call
229 	 * ktls_output_eagain after the mbuf has been freed (thus
230 	 * dropping the mbuf's reference) in if_output.
231 	 */
232 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
233 		tls = ktls_hold(m->m_next->m_epg_tls);
234 		mst = tls->snd_tag;
235 
236 		/*
237 		 * If a TLS session doesn't have a valid tag, it must
238 		 * have had an earlier ifp mismatch, so drop this
239 		 * packet.
240 		 */
241 		if (mst == NULL) {
242 			error = EAGAIN;
243 			goto done;
244 		}
245 		/*
246 		 * Always stamp tags that include NIC ktls.
247 		 */
248 		stamp_tag = true;
249 	}
250 #endif
251 #ifdef RATELIMIT
252 	if (inp != NULL && mst == NULL) {
253 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
254 		    (inp->inp_snd_tag != NULL &&
255 		    inp->inp_snd_tag->ifp != ifp))
256 			in_pcboutput_txrtlmt(inp, ifp, m);
257 
258 		if (inp->inp_snd_tag != NULL)
259 			mst = inp->inp_snd_tag;
260 	}
261 #endif
262 	if (stamp_tag && mst != NULL) {
263 		KASSERT(m->m_pkthdr.rcvif == NULL,
264 		    ("trying to add a send tag to a forwarded packet"));
265 		if (mst->ifp != ifp) {
266 			error = EAGAIN;
267 			goto done;
268 		}
269 
270 		/* stamp send tag on mbuf */
271 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
272 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
273 	}
274 
275 	error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
276 
277 done:
278 	/* Check for route change invalidating send tags. */
279 #ifdef KERN_TLS
280 	if (tls != NULL) {
281 		if (error == EAGAIN)
282 			error = ktls_output_eagain(inp, tls);
283 		ktls_free(tls);
284 	}
285 #endif
286 #ifdef RATELIMIT
287 	if (error == EAGAIN)
288 		in_pcboutput_eagain(inp);
289 #endif
290 	return (error);
291 }
292 
293 /* rte<>ro_flags translation */
294 static inline void
295 rt_update_ro_flags(struct route *ro)
296 {
297 	int nh_flags = ro->ro_nh->nh_flags;
298 
299 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
300 
301 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
302 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
303 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
304 }
305 
306 /*
307  * IP output.  The packet in mbuf chain m contains a skeletal IP
308  * header (with len, off, ttl, proto, tos, src, dst).
309  * The mbuf chain containing the packet will be freed.
310  * The mbuf opt, if present, will not be freed.
311  * If route ro is present and has ro_rt initialized, route lookup would be
312  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
313  * then result of route lookup is stored in ro->ro_rt.
314  *
315  * In the IP forwarding case, the packet will arrive with options already
316  * inserted, so must have a NULL opt pointer.
317  */
318 int
319 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
320     struct ip_moptions *imo, struct inpcb *inp)
321 {
322 	struct rm_priotracker in_ifa_tracker;
323 	struct ip *ip;
324 	struct ifnet *ifp = NULL;	/* keep compiler happy */
325 	struct mbuf *m0;
326 	int hlen = sizeof (struct ip);
327 	int mtu = 0;
328 	int error = 0;
329 	int vlan_pcp = -1;
330 	struct sockaddr_in *dst, sin;
331 	const struct sockaddr_in *gw;
332 	struct in_ifaddr *ia = NULL;
333 	struct in_addr src;
334 	int isbroadcast;
335 	uint16_t ip_len, ip_off;
336 	uint32_t fibnum;
337 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
338 	int no_route_but_check_spd = 0;
339 #endif
340 
341 	M_ASSERTPKTHDR(m);
342 	NET_EPOCH_ASSERT();
343 
344 	if (inp != NULL) {
345 		INP_LOCK_ASSERT(inp);
346 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
347 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
348 			m->m_pkthdr.flowid = inp->inp_flowid;
349 			M_HASHTYPE_SET(m, inp->inp_flowtype);
350 		}
351 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
352 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
353 			    INP_2PCP_SHIFT;
354 #ifdef NUMA
355 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
356 #endif
357 	}
358 
359 	if (opt) {
360 		int len = 0;
361 		m = ip_insertoptions(m, opt, &len);
362 		if (len != 0)
363 			hlen = len; /* ip->ip_hl is updated above */
364 	}
365 	ip = mtod(m, struct ip *);
366 	ip_len = ntohs(ip->ip_len);
367 	ip_off = ntohs(ip->ip_off);
368 
369 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
370 		ip->ip_v = IPVERSION;
371 		ip->ip_hl = hlen >> 2;
372 		ip_fillid(ip);
373 	} else {
374 		/* Header already set, fetch hlen from there */
375 		hlen = ip->ip_hl << 2;
376 	}
377 	if ((flags & IP_FORWARDING) == 0)
378 		IPSTAT_INC(ips_localout);
379 
380 	/*
381 	 * dst/gw handling:
382 	 *
383 	 * gw is readonly but can point either to dst OR rt_gateway,
384 	 * therefore we need restore gw if we're redoing lookup.
385 	 */
386 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
387 	if (ro != NULL)
388 		dst = (struct sockaddr_in *)&ro->ro_dst;
389 	else
390 		dst = &sin;
391 	if (ro == NULL || ro->ro_nh == NULL) {
392 		bzero(dst, sizeof(*dst));
393 		dst->sin_family = AF_INET;
394 		dst->sin_len = sizeof(*dst);
395 		dst->sin_addr = ip->ip_dst;
396 	}
397 	gw = dst;
398 again:
399 	/*
400 	 * Validate route against routing table additions;
401 	 * a better/more specific route might have been added.
402 	 */
403 	if (inp != NULL && ro != NULL && ro->ro_nh != NULL)
404 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
405 	/*
406 	 * If there is a cached route,
407 	 * check that it is to the same destination
408 	 * and is still up.  If not, free it and try again.
409 	 * The address family should also be checked in case of sharing the
410 	 * cache with IPv6.
411 	 * Also check whether routing cache needs invalidation.
412 	 */
413 	if (ro != NULL && ro->ro_nh != NULL &&
414 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
415 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
416 		RO_INVALIDATE_CACHE(ro);
417 	ia = NULL;
418 	/*
419 	 * If routing to interface only, short circuit routing lookup.
420 	 * The use of an all-ones broadcast address implies this; an
421 	 * interface is specified by the broadcast address of an interface,
422 	 * or the destination address of a ptp interface.
423 	 */
424 	if (flags & IP_SENDONES) {
425 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
426 						      M_GETFIB(m)))) == NULL &&
427 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
428 						    M_GETFIB(m)))) == NULL) {
429 			IPSTAT_INC(ips_noroute);
430 			error = ENETUNREACH;
431 			goto bad;
432 		}
433 		ip->ip_dst.s_addr = INADDR_BROADCAST;
434 		dst->sin_addr = ip->ip_dst;
435 		ifp = ia->ia_ifp;
436 		mtu = ifp->if_mtu;
437 		ip->ip_ttl = 1;
438 		isbroadcast = 1;
439 		src = IA_SIN(ia)->sin_addr;
440 	} else if (flags & IP_ROUTETOIF) {
441 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
442 						    M_GETFIB(m)))) == NULL &&
443 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
444 						M_GETFIB(m)))) == NULL) {
445 			IPSTAT_INC(ips_noroute);
446 			error = ENETUNREACH;
447 			goto bad;
448 		}
449 		ifp = ia->ia_ifp;
450 		mtu = ifp->if_mtu;
451 		ip->ip_ttl = 1;
452 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
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, &in_ifa_tracker);
464 		isbroadcast = 0;	/* fool gcc */
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 != NULL) {
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 		ia = ifatoia(ro->ro_nh->nh_ifa);
497 		ifp = ro->ro_nh->nh_ifp;
498 		counter_u64_add(ro->ro_nh->nh_pksent, 1);
499 		rt_update_ro_flags(ro);
500 		if (ro->ro_nh->nh_flags & NHF_GATEWAY)
501 			gw = &ro->ro_nh->gw4_sa;
502 		if (ro->ro_nh->nh_flags & NHF_HOST)
503 			isbroadcast = (ro->ro_nh->nh_flags & NHF_BROADCAST);
504 		else if (ifp->if_flags & IFF_BROADCAST)
505 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
506 		else
507 			isbroadcast = 0;
508 		if (ro->ro_nh->nh_flags & NHF_HOST)
509 			mtu = ro->ro_nh->nh_mtu;
510 		else
511 			mtu = ifp->if_mtu;
512 		src = IA_SIN(ia)->sin_addr;
513 	} else {
514 		struct nhop_object *nh;
515 
516 		nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE,
517 		    m->m_pkthdr.flowid);
518 		if (nh == NULL) {
519 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
520 			/*
521 			 * There is no route for this packet, but it is
522 			 * possible that a matching SPD entry exists.
523 			 */
524 			no_route_but_check_spd = 1;
525 			goto sendit;
526 #endif
527 			IPSTAT_INC(ips_noroute);
528 			error = EHOSTUNREACH;
529 			goto bad;
530 		}
531 		ifp = nh->nh_ifp;
532 		mtu = nh->nh_mtu;
533 		/*
534 		 * We are rewriting here dst to be gw actually, contradicting
535 		 * comment at the beginning of the function. However, in this
536 		 * case we are always dealing with on stack dst.
537 		 * In case if pfil(9) sends us back to beginning of the
538 		 * function, the dst would be rewritten by ip_output_pfil().
539 		 */
540 		MPASS(dst == &sin);
541 		if (nh->nh_flags & NHF_GATEWAY)
542 			dst->sin_addr = nh->gw4_sa.sin_addr;
543 		ia = ifatoia(nh->nh_ifa);
544 		src = IA_SIN(ia)->sin_addr;
545 		isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
546 		    (NHF_HOST | NHF_BROADCAST)) ||
547 		    ((ifp->if_flags & IFF_BROADCAST) &&
548 		    in_ifaddr_broadcast(dst->sin_addr, ia)));
549 	}
550 
551 	/* Catch a possible divide by zero later. */
552 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
553 	    __func__, mtu, ro,
554 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
555 
556 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
557 		m->m_flags |= M_MCAST;
558 		/*
559 		 * IP destination address is multicast.  Make sure "gw"
560 		 * still points to the address in "ro".  (It may have been
561 		 * changed to point to a gateway address, above.)
562 		 */
563 		gw = dst;
564 		/*
565 		 * See if the caller provided any multicast options
566 		 */
567 		if (imo != NULL) {
568 			ip->ip_ttl = imo->imo_multicast_ttl;
569 			if (imo->imo_multicast_vif != -1)
570 				ip->ip_src.s_addr =
571 				    ip_mcast_src ?
572 				    ip_mcast_src(imo->imo_multicast_vif) :
573 				    INADDR_ANY;
574 		} else
575 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
576 		/*
577 		 * Confirm that the outgoing interface supports multicast.
578 		 */
579 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
580 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
581 				IPSTAT_INC(ips_noroute);
582 				error = ENETUNREACH;
583 				goto bad;
584 			}
585 		}
586 		/*
587 		 * If source address not specified yet, use address
588 		 * of outgoing interface.
589 		 */
590 		if (ip->ip_src.s_addr == INADDR_ANY)
591 			ip->ip_src = src;
592 
593 		if ((imo == NULL && in_mcast_loop) ||
594 		    (imo && imo->imo_multicast_loop)) {
595 			/*
596 			 * Loop back multicast datagram if not expressly
597 			 * forbidden to do so, even if we are not a member
598 			 * of the group; ip_input() will filter it later,
599 			 * thus deferring a hash lookup and mutex acquisition
600 			 * at the expense of a cheap copy using m_copym().
601 			 */
602 			ip_mloopback(ifp, m, hlen);
603 		} else {
604 			/*
605 			 * If we are acting as a multicast router, perform
606 			 * multicast forwarding as if the packet had just
607 			 * arrived on the interface to which we are about
608 			 * to send.  The multicast forwarding function
609 			 * recursively calls this function, using the
610 			 * IP_FORWARDING flag to prevent infinite recursion.
611 			 *
612 			 * Multicasts that are looped back by ip_mloopback(),
613 			 * above, will be forwarded by the ip_input() routine,
614 			 * if necessary.
615 			 */
616 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
617 				/*
618 				 * If rsvp daemon is not running, do not
619 				 * set ip_moptions. This ensures that the packet
620 				 * is multicast and not just sent down one link
621 				 * as prescribed by rsvpd.
622 				 */
623 				if (!V_rsvp_on)
624 					imo = NULL;
625 				if (ip_mforward &&
626 				    ip_mforward(ip, ifp, m, imo) != 0) {
627 					m_freem(m);
628 					goto done;
629 				}
630 			}
631 		}
632 
633 		/*
634 		 * Multicasts with a time-to-live of zero may be looped-
635 		 * back, above, but must not be transmitted on a network.
636 		 * Also, multicasts addressed to the loopback interface
637 		 * are not sent -- the above call to ip_mloopback() will
638 		 * loop back a copy. ip_input() will drop the copy if
639 		 * this host does not belong to the destination group on
640 		 * the loopback interface.
641 		 */
642 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
643 			m_freem(m);
644 			goto done;
645 		}
646 
647 		goto sendit;
648 	}
649 
650 	/*
651 	 * If the source address is not specified yet, use the address
652 	 * of the outoing interface.
653 	 */
654 	if (ip->ip_src.s_addr == INADDR_ANY)
655 		ip->ip_src = src;
656 
657 	/*
658 	 * Look for broadcast address and
659 	 * verify user is allowed to send
660 	 * such a packet.
661 	 */
662 	if (isbroadcast) {
663 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
664 			error = EADDRNOTAVAIL;
665 			goto bad;
666 		}
667 		if ((flags & IP_ALLOWBROADCAST) == 0) {
668 			error = EACCES;
669 			goto bad;
670 		}
671 		/* don't allow broadcast messages to be fragmented */
672 		if (ip_len > mtu) {
673 			error = EMSGSIZE;
674 			goto bad;
675 		}
676 		m->m_flags |= M_BCAST;
677 	} else {
678 		m->m_flags &= ~M_BCAST;
679 	}
680 
681 sendit:
682 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
683 	if (IPSEC_ENABLED(ipv4)) {
684 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
685 			if (error == EINPROGRESS)
686 				error = 0;
687 			goto done;
688 		}
689 	}
690 	/*
691 	 * Check if there was a route for this packet; return error if not.
692 	 */
693 	if (no_route_but_check_spd) {
694 		IPSTAT_INC(ips_noroute);
695 		error = EHOSTUNREACH;
696 		goto bad;
697 	}
698 	/* Update variables that are affected by ipsec4_output(). */
699 	ip = mtod(m, struct ip *);
700 	hlen = ip->ip_hl << 2;
701 #endif /* IPSEC */
702 
703 	/* Jump over all PFIL processing if hooks are not active. */
704 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
705 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
706 		    &error)) {
707 		case 1: /* Finished */
708 			goto done;
709 
710 		case 0: /* Continue normally */
711 			ip = mtod(m, struct ip *);
712 			break;
713 
714 		case -1: /* Need to try again */
715 			/* Reset everything for a new round */
716 			if (ro != NULL) {
717 				RO_NHFREE(ro);
718 				ro->ro_prepend = NULL;
719 			}
720 			gw = dst;
721 			ip = mtod(m, struct ip *);
722 			goto again;
723 		}
724 	}
725 
726 	if (vlan_pcp > -1)
727 		EVL_APPLY_PRI(m, vlan_pcp);
728 
729 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
730 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
731 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
732 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
733 			IPSTAT_INC(ips_badaddr);
734 			error = EADDRNOTAVAIL;
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 		m = mb_unmapped_to_ext(m);
742 		if (m == NULL) {
743 			IPSTAT_INC(ips_odropped);
744 			error = ENOBUFS;
745 			goto bad;
746 		}
747 		in_delayed_cksum(m);
748 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
749 	} else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
750 		m = mb_unmapped_to_ext(m);
751 		if (m == NULL) {
752 			IPSTAT_INC(ips_odropped);
753 			error = ENOBUFS;
754 			goto bad;
755 		}
756 	}
757 #if defined(SCTP) || defined(SCTP_SUPPORT)
758 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
759 		m = mb_unmapped_to_ext(m);
760 		if (m == NULL) {
761 			IPSTAT_INC(ips_odropped);
762 			error = ENOBUFS;
763 			goto bad;
764 		}
765 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
766 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
767 	}
768 #endif
769 
770 	/*
771 	 * If small enough for interface, or the interface will take
772 	 * care of the fragmentation for us, we can just send directly.
773 	 * Note that if_vxlan could have requested TSO even though the outer
774 	 * frame is UDP.  It is correct to not fragment such datagrams and
775 	 * instead just pass them on to the driver.
776 	 */
777 	if (ip_len <= mtu ||
778 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
779 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
780 		ip->ip_sum = 0;
781 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
782 			ip->ip_sum = in_cksum(m, hlen);
783 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
784 		}
785 
786 		/*
787 		 * Record statistics for this interface address.
788 		 * With CSUM_TSO the byte/packet count will be slightly
789 		 * incorrect because we count the IP+TCP headers only
790 		 * once instead of for every generated packet.
791 		 */
792 		if (!(flags & IP_FORWARDING) && ia) {
793 			if (m->m_pkthdr.csum_flags &
794 			    (CSUM_TSO | CSUM_INNER_TSO))
795 				counter_u64_add(ia->ia_ifa.ifa_opackets,
796 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
797 			else
798 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
799 
800 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
801 		}
802 #ifdef MBUF_STRESS_TEST
803 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
804 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
805 #endif
806 		/*
807 		 * Reset layer specific mbuf flags
808 		 * to avoid confusing lower layers.
809 		 */
810 		m_clrprotoflags(m);
811 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
812 		error = ip_output_send(inp, ifp, m, gw, ro,
813 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
814 		goto done;
815 	}
816 
817 	/* Balk when DF bit is set or the interface didn't support TSO. */
818 	if ((ip_off & IP_DF) ||
819 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
820 		error = EMSGSIZE;
821 		IPSTAT_INC(ips_cantfrag);
822 		goto bad;
823 	}
824 
825 	/*
826 	 * Too large for interface; fragment if possible. If successful,
827 	 * on return, m will point to a list of packets to be sent.
828 	 */
829 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
830 	if (error)
831 		goto bad;
832 	for (; m; m = m0) {
833 		m0 = m->m_nextpkt;
834 		m->m_nextpkt = 0;
835 		if (error == 0) {
836 			/* Record statistics for this interface address. */
837 			if (ia != NULL) {
838 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
839 				counter_u64_add(ia->ia_ifa.ifa_obytes,
840 				    m->m_pkthdr.len);
841 			}
842 			/*
843 			 * Reset layer specific mbuf flags
844 			 * to avoid confusing upper layers.
845 			 */
846 			m_clrprotoflags(m);
847 
848 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
849 			    mtod(m, struct ip *), NULL);
850 			error = ip_output_send(inp, ifp, m, gw, ro, true);
851 		} else
852 			m_freem(m);
853 	}
854 
855 	if (error == 0)
856 		IPSTAT_INC(ips_fragmented);
857 
858 done:
859 	return (error);
860  bad:
861 	m_freem(m);
862 	goto done;
863 }
864 
865 /*
866  * Create a chain of fragments which fit the given mtu. m_frag points to the
867  * mbuf to be fragmented; on return it points to the chain with the fragments.
868  * Return 0 if no error. If error, m_frag may contain a partially built
869  * chain of fragments that should be freed by the caller.
870  *
871  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
872  */
873 int
874 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
875     u_long if_hwassist_flags)
876 {
877 	int error = 0;
878 	int hlen = ip->ip_hl << 2;
879 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
880 	int off;
881 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
882 	int firstlen;
883 	struct mbuf **mnext;
884 	int nfrags;
885 	uint16_t ip_len, ip_off;
886 
887 	ip_len = ntohs(ip->ip_len);
888 	ip_off = ntohs(ip->ip_off);
889 
890 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
891 		IPSTAT_INC(ips_cantfrag);
892 		return EMSGSIZE;
893 	}
894 
895 	/*
896 	 * Must be able to put at least 8 bytes per fragment.
897 	 */
898 	if (len < 8)
899 		return EMSGSIZE;
900 
901 	/*
902 	 * If the interface will not calculate checksums on
903 	 * fragmented packets, then do it here.
904 	 */
905 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
906 		m0 = mb_unmapped_to_ext(m0);
907 		if (m0 == NULL) {
908 			error = ENOBUFS;
909 			IPSTAT_INC(ips_odropped);
910 			goto done;
911 		}
912 		in_delayed_cksum(m0);
913 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
914 	}
915 #if defined(SCTP) || defined(SCTP_SUPPORT)
916 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
917 		m0 = mb_unmapped_to_ext(m0);
918 		if (m0 == NULL) {
919 			error = ENOBUFS;
920 			IPSTAT_INC(ips_odropped);
921 			goto done;
922 		}
923 		sctp_delayed_cksum(m0, hlen);
924 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
925 	}
926 #endif
927 	if (len > PAGE_SIZE) {
928 		/*
929 		 * Fragment large datagrams such that each segment
930 		 * contains a multiple of PAGE_SIZE amount of data,
931 		 * plus headers. This enables a receiver to perform
932 		 * page-flipping zero-copy optimizations.
933 		 *
934 		 * XXX When does this help given that sender and receiver
935 		 * could have different page sizes, and also mtu could
936 		 * be less than the receiver's page size ?
937 		 */
938 		int newlen;
939 
940 		off = MIN(mtu, m0->m_pkthdr.len);
941 
942 		/*
943 		 * firstlen (off - hlen) must be aligned on an
944 		 * 8-byte boundary
945 		 */
946 		if (off < hlen)
947 			goto smart_frag_failure;
948 		off = ((off - hlen) & ~7) + hlen;
949 		newlen = (~PAGE_MASK) & mtu;
950 		if ((newlen + sizeof (struct ip)) > mtu) {
951 			/* we failed, go back the default */
952 smart_frag_failure:
953 			newlen = len;
954 			off = hlen + len;
955 		}
956 		len = newlen;
957 
958 	} else {
959 		off = hlen + len;
960 	}
961 
962 	firstlen = off - hlen;
963 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
964 
965 	/*
966 	 * Loop through length of segment after first fragment,
967 	 * make new header and copy data of each part and link onto chain.
968 	 * Here, m0 is the original packet, m is the fragment being created.
969 	 * The fragments are linked off the m_nextpkt of the original
970 	 * packet, which after processing serves as the first fragment.
971 	 */
972 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
973 		struct ip *mhip;	/* ip header on the fragment */
974 		struct mbuf *m;
975 		int mhlen = sizeof (struct ip);
976 
977 		m = m_gethdr(M_NOWAIT, MT_DATA);
978 		if (m == NULL) {
979 			error = ENOBUFS;
980 			IPSTAT_INC(ips_odropped);
981 			goto done;
982 		}
983 		/*
984 		 * Make sure the complete packet header gets copied
985 		 * from the originating mbuf to the newly created
986 		 * mbuf. This also ensures that existing firewall
987 		 * classification(s), VLAN tags and so on get copied
988 		 * to the resulting fragmented packet(s):
989 		 */
990 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
991 			m_free(m);
992 			error = ENOBUFS;
993 			IPSTAT_INC(ips_odropped);
994 			goto done;
995 		}
996 		/*
997 		 * In the first mbuf, leave room for the link header, then
998 		 * copy the original IP header including options. The payload
999 		 * goes into an additional mbuf chain returned by m_copym().
1000 		 */
1001 		m->m_data += max_linkhdr;
1002 		mhip = mtod(m, struct ip *);
1003 		*mhip = *ip;
1004 		if (hlen > sizeof (struct ip)) {
1005 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1006 			mhip->ip_v = IPVERSION;
1007 			mhip->ip_hl = mhlen >> 2;
1008 		}
1009 		m->m_len = mhlen;
1010 		/* XXX do we need to add ip_off below ? */
1011 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1012 		if (off + len >= ip_len)
1013 			len = ip_len - off;
1014 		else
1015 			mhip->ip_off |= IP_MF;
1016 		mhip->ip_len = htons((u_short)(len + mhlen));
1017 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1018 		if (m->m_next == NULL) {	/* copy failed */
1019 			m_free(m);
1020 			error = ENOBUFS;	/* ??? */
1021 			IPSTAT_INC(ips_odropped);
1022 			goto done;
1023 		}
1024 		m->m_pkthdr.len = mhlen + len;
1025 #ifdef MAC
1026 		mac_netinet_fragment(m0, m);
1027 #endif
1028 		mhip->ip_off = htons(mhip->ip_off);
1029 		mhip->ip_sum = 0;
1030 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1031 			mhip->ip_sum = in_cksum(m, mhlen);
1032 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1033 		}
1034 		*mnext = m;
1035 		mnext = &m->m_nextpkt;
1036 	}
1037 	IPSTAT_ADD(ips_ofragments, nfrags);
1038 
1039 	/*
1040 	 * Update first fragment by trimming what's been copied out
1041 	 * and updating header.
1042 	 */
1043 	m_adj(m0, hlen + firstlen - ip_len);
1044 	m0->m_pkthdr.len = hlen + firstlen;
1045 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1046 	ip->ip_off = htons(ip_off | IP_MF);
1047 	ip->ip_sum = 0;
1048 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1049 		ip->ip_sum = in_cksum(m0, hlen);
1050 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1051 	}
1052 
1053 done:
1054 	*m_frag = m0;
1055 	return error;
1056 }
1057 
1058 void
1059 in_delayed_cksum(struct mbuf *m)
1060 {
1061 	struct ip *ip;
1062 	struct udphdr *uh;
1063 	uint16_t cklen, csum, offset;
1064 
1065 	ip = mtod(m, struct ip *);
1066 	offset = ip->ip_hl << 2 ;
1067 
1068 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1069 		/* if udp header is not in the first mbuf copy udplen */
1070 		if (offset + sizeof(struct udphdr) > m->m_len) {
1071 			m_copydata(m, offset + offsetof(struct udphdr,
1072 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1073 			cklen = ntohs(cklen);
1074 		} else {
1075 			uh = (struct udphdr *)mtodo(m, offset);
1076 			cklen = ntohs(uh->uh_ulen);
1077 		}
1078 		csum = in_cksum_skip(m, cklen + offset, offset);
1079 		if (csum == 0)
1080 			csum = 0xffff;
1081 	} else {
1082 		cklen = ntohs(ip->ip_len);
1083 		csum = in_cksum_skip(m, cklen, offset);
1084 	}
1085 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1086 
1087 	if (offset + sizeof(csum) > m->m_len)
1088 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1089 	else
1090 		*(u_short *)mtodo(m, offset) = csum;
1091 }
1092 
1093 /*
1094  * IP socket option processing.
1095  */
1096 int
1097 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1098 {
1099 	struct inpcb *inp = sotoinpcb(so);
1100 	int	error, optval;
1101 #ifdef	RSS
1102 	uint32_t rss_bucket;
1103 	int retval;
1104 #endif
1105 
1106 	error = optval = 0;
1107 	if (sopt->sopt_level != IPPROTO_IP) {
1108 		error = EINVAL;
1109 
1110 		if (sopt->sopt_level == SOL_SOCKET &&
1111 		    sopt->sopt_dir == SOPT_SET) {
1112 			switch (sopt->sopt_name) {
1113 			case SO_REUSEADDR:
1114 				INP_WLOCK(inp);
1115 				if ((so->so_options & SO_REUSEADDR) != 0)
1116 					inp->inp_flags2 |= INP_REUSEADDR;
1117 				else
1118 					inp->inp_flags2 &= ~INP_REUSEADDR;
1119 				INP_WUNLOCK(inp);
1120 				error = 0;
1121 				break;
1122 			case SO_REUSEPORT:
1123 				INP_WLOCK(inp);
1124 				if ((so->so_options & SO_REUSEPORT) != 0)
1125 					inp->inp_flags2 |= INP_REUSEPORT;
1126 				else
1127 					inp->inp_flags2 &= ~INP_REUSEPORT;
1128 				INP_WUNLOCK(inp);
1129 				error = 0;
1130 				break;
1131 			case SO_REUSEPORT_LB:
1132 				INP_WLOCK(inp);
1133 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1134 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1135 				else
1136 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1137 				INP_WUNLOCK(inp);
1138 				error = 0;
1139 				break;
1140 			case SO_SETFIB:
1141 				INP_WLOCK(inp);
1142 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1143 				INP_WUNLOCK(inp);
1144 				error = 0;
1145 				break;
1146 			case SO_MAX_PACING_RATE:
1147 #ifdef RATELIMIT
1148 				INP_WLOCK(inp);
1149 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1150 				INP_WUNLOCK(inp);
1151 				error = 0;
1152 #else
1153 				error = EOPNOTSUPP;
1154 #endif
1155 				break;
1156 			default:
1157 				break;
1158 			}
1159 		}
1160 		return (error);
1161 	}
1162 
1163 	switch (sopt->sopt_dir) {
1164 	case SOPT_SET:
1165 		switch (sopt->sopt_name) {
1166 		case IP_OPTIONS:
1167 #ifdef notyet
1168 		case IP_RETOPTS:
1169 #endif
1170 		{
1171 			struct mbuf *m;
1172 			if (sopt->sopt_valsize > MLEN) {
1173 				error = EMSGSIZE;
1174 				break;
1175 			}
1176 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1177 			if (m == NULL) {
1178 				error = ENOBUFS;
1179 				break;
1180 			}
1181 			m->m_len = sopt->sopt_valsize;
1182 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1183 					    m->m_len);
1184 			if (error) {
1185 				m_free(m);
1186 				break;
1187 			}
1188 			INP_WLOCK(inp);
1189 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1190 			INP_WUNLOCK(inp);
1191 			return (error);
1192 		}
1193 
1194 		case IP_BINDANY:
1195 			if (sopt->sopt_td != NULL) {
1196 				error = priv_check(sopt->sopt_td,
1197 				    PRIV_NETINET_BINDANY);
1198 				if (error)
1199 					break;
1200 			}
1201 			/* FALLTHROUGH */
1202 		case IP_BINDMULTI:
1203 #ifdef	RSS
1204 		case IP_RSS_LISTEN_BUCKET:
1205 #endif
1206 		case IP_TOS:
1207 		case IP_TTL:
1208 		case IP_MINTTL:
1209 		case IP_RECVOPTS:
1210 		case IP_RECVRETOPTS:
1211 		case IP_ORIGDSTADDR:
1212 		case IP_RECVDSTADDR:
1213 		case IP_RECVTTL:
1214 		case IP_RECVIF:
1215 		case IP_ONESBCAST:
1216 		case IP_DONTFRAG:
1217 		case IP_RECVTOS:
1218 		case IP_RECVFLOWID:
1219 #ifdef	RSS
1220 		case IP_RECVRSSBUCKETID:
1221 #endif
1222 		case IP_VLAN_PCP:
1223 			error = sooptcopyin(sopt, &optval, sizeof optval,
1224 					    sizeof optval);
1225 			if (error)
1226 				break;
1227 
1228 			switch (sopt->sopt_name) {
1229 			case IP_TOS:
1230 				inp->inp_ip_tos = optval;
1231 				break;
1232 
1233 			case IP_TTL:
1234 				inp->inp_ip_ttl = optval;
1235 				break;
1236 
1237 			case IP_MINTTL:
1238 				if (optval >= 0 && optval <= MAXTTL)
1239 					inp->inp_ip_minttl = optval;
1240 				else
1241 					error = EINVAL;
1242 				break;
1243 
1244 #define	OPTSET(bit) do {						\
1245 	INP_WLOCK(inp);							\
1246 	if (optval)							\
1247 		inp->inp_flags |= bit;					\
1248 	else								\
1249 		inp->inp_flags &= ~bit;					\
1250 	INP_WUNLOCK(inp);						\
1251 } while (0)
1252 
1253 #define	OPTSET2(bit, val) do {						\
1254 	INP_WLOCK(inp);							\
1255 	if (val)							\
1256 		inp->inp_flags2 |= bit;					\
1257 	else								\
1258 		inp->inp_flags2 &= ~bit;				\
1259 	INP_WUNLOCK(inp);						\
1260 } while (0)
1261 
1262 			case IP_RECVOPTS:
1263 				OPTSET(INP_RECVOPTS);
1264 				break;
1265 
1266 			case IP_RECVRETOPTS:
1267 				OPTSET(INP_RECVRETOPTS);
1268 				break;
1269 
1270 			case IP_RECVDSTADDR:
1271 				OPTSET(INP_RECVDSTADDR);
1272 				break;
1273 
1274 			case IP_ORIGDSTADDR:
1275 				OPTSET2(INP_ORIGDSTADDR, optval);
1276 				break;
1277 
1278 			case IP_RECVTTL:
1279 				OPTSET(INP_RECVTTL);
1280 				break;
1281 
1282 			case IP_RECVIF:
1283 				OPTSET(INP_RECVIF);
1284 				break;
1285 
1286 			case IP_ONESBCAST:
1287 				OPTSET(INP_ONESBCAST);
1288 				break;
1289 			case IP_DONTFRAG:
1290 				OPTSET(INP_DONTFRAG);
1291 				break;
1292 			case IP_BINDANY:
1293 				OPTSET(INP_BINDANY);
1294 				break;
1295 			case IP_RECVTOS:
1296 				OPTSET(INP_RECVTOS);
1297 				break;
1298 			case IP_BINDMULTI:
1299 				OPTSET2(INP_BINDMULTI, optval);
1300 				break;
1301 			case IP_RECVFLOWID:
1302 				OPTSET2(INP_RECVFLOWID, optval);
1303 				break;
1304 #ifdef	RSS
1305 			case IP_RSS_LISTEN_BUCKET:
1306 				if ((optval >= 0) &&
1307 				    (optval < rss_getnumbuckets())) {
1308 					inp->inp_rss_listen_bucket = optval;
1309 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1310 				} else {
1311 					error = EINVAL;
1312 				}
1313 				break;
1314 			case IP_RECVRSSBUCKETID:
1315 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1316 				break;
1317 #endif
1318 			case IP_VLAN_PCP:
1319 				if ((optval >= -1) && (optval <=
1320 				    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1321 					if (optval == -1) {
1322 						INP_WLOCK(inp);
1323 						inp->inp_flags2 &=
1324 						    ~(INP_2PCP_SET |
1325 						      INP_2PCP_MASK);
1326 						INP_WUNLOCK(inp);
1327 					} else {
1328 						INP_WLOCK(inp);
1329 						inp->inp_flags2 |=
1330 						    INP_2PCP_SET;
1331 						inp->inp_flags2 &=
1332 						    ~INP_2PCP_MASK;
1333 						inp->inp_flags2 |=
1334 						    optval << INP_2PCP_SHIFT;
1335 						INP_WUNLOCK(inp);
1336 					}
1337 				} else
1338 					error = EINVAL;
1339 				break;
1340 			}
1341 			break;
1342 #undef OPTSET
1343 #undef OPTSET2
1344 
1345 		/*
1346 		 * Multicast socket options are processed by the in_mcast
1347 		 * module.
1348 		 */
1349 		case IP_MULTICAST_IF:
1350 		case IP_MULTICAST_VIF:
1351 		case IP_MULTICAST_TTL:
1352 		case IP_MULTICAST_LOOP:
1353 		case IP_ADD_MEMBERSHIP:
1354 		case IP_DROP_MEMBERSHIP:
1355 		case IP_ADD_SOURCE_MEMBERSHIP:
1356 		case IP_DROP_SOURCE_MEMBERSHIP:
1357 		case IP_BLOCK_SOURCE:
1358 		case IP_UNBLOCK_SOURCE:
1359 		case IP_MSFILTER:
1360 		case MCAST_JOIN_GROUP:
1361 		case MCAST_LEAVE_GROUP:
1362 		case MCAST_JOIN_SOURCE_GROUP:
1363 		case MCAST_LEAVE_SOURCE_GROUP:
1364 		case MCAST_BLOCK_SOURCE:
1365 		case MCAST_UNBLOCK_SOURCE:
1366 			error = inp_setmoptions(inp, sopt);
1367 			break;
1368 
1369 		case IP_PORTRANGE:
1370 			error = sooptcopyin(sopt, &optval, sizeof optval,
1371 					    sizeof optval);
1372 			if (error)
1373 				break;
1374 
1375 			INP_WLOCK(inp);
1376 			switch (optval) {
1377 			case IP_PORTRANGE_DEFAULT:
1378 				inp->inp_flags &= ~(INP_LOWPORT);
1379 				inp->inp_flags &= ~(INP_HIGHPORT);
1380 				break;
1381 
1382 			case IP_PORTRANGE_HIGH:
1383 				inp->inp_flags &= ~(INP_LOWPORT);
1384 				inp->inp_flags |= INP_HIGHPORT;
1385 				break;
1386 
1387 			case IP_PORTRANGE_LOW:
1388 				inp->inp_flags &= ~(INP_HIGHPORT);
1389 				inp->inp_flags |= INP_LOWPORT;
1390 				break;
1391 
1392 			default:
1393 				error = EINVAL;
1394 				break;
1395 			}
1396 			INP_WUNLOCK(inp);
1397 			break;
1398 
1399 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1400 		case IP_IPSEC_POLICY:
1401 			if (IPSEC_ENABLED(ipv4)) {
1402 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1403 				break;
1404 			}
1405 			/* FALLTHROUGH */
1406 #endif /* IPSEC */
1407 
1408 		default:
1409 			error = ENOPROTOOPT;
1410 			break;
1411 		}
1412 		break;
1413 
1414 	case SOPT_GET:
1415 		switch (sopt->sopt_name) {
1416 		case IP_OPTIONS:
1417 		case IP_RETOPTS:
1418 			INP_RLOCK(inp);
1419 			if (inp->inp_options) {
1420 				struct mbuf *options;
1421 
1422 				options = m_copym(inp->inp_options, 0,
1423 				    M_COPYALL, M_NOWAIT);
1424 				INP_RUNLOCK(inp);
1425 				if (options != NULL) {
1426 					error = sooptcopyout(sopt,
1427 							     mtod(options, char *),
1428 							     options->m_len);
1429 					m_freem(options);
1430 				} else
1431 					error = ENOMEM;
1432 			} else {
1433 				INP_RUNLOCK(inp);
1434 				sopt->sopt_valsize = 0;
1435 			}
1436 			break;
1437 
1438 		case IP_TOS:
1439 		case IP_TTL:
1440 		case IP_MINTTL:
1441 		case IP_RECVOPTS:
1442 		case IP_RECVRETOPTS:
1443 		case IP_ORIGDSTADDR:
1444 		case IP_RECVDSTADDR:
1445 		case IP_RECVTTL:
1446 		case IP_RECVIF:
1447 		case IP_PORTRANGE:
1448 		case IP_ONESBCAST:
1449 		case IP_DONTFRAG:
1450 		case IP_BINDANY:
1451 		case IP_RECVTOS:
1452 		case IP_BINDMULTI:
1453 		case IP_FLOWID:
1454 		case IP_FLOWTYPE:
1455 		case IP_RECVFLOWID:
1456 #ifdef	RSS
1457 		case IP_RSSBUCKETID:
1458 		case IP_RECVRSSBUCKETID:
1459 #endif
1460 		case IP_VLAN_PCP:
1461 			switch (sopt->sopt_name) {
1462 			case IP_TOS:
1463 				optval = inp->inp_ip_tos;
1464 				break;
1465 
1466 			case IP_TTL:
1467 				optval = inp->inp_ip_ttl;
1468 				break;
1469 
1470 			case IP_MINTTL:
1471 				optval = inp->inp_ip_minttl;
1472 				break;
1473 
1474 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1475 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1476 
1477 			case IP_RECVOPTS:
1478 				optval = OPTBIT(INP_RECVOPTS);
1479 				break;
1480 
1481 			case IP_RECVRETOPTS:
1482 				optval = OPTBIT(INP_RECVRETOPTS);
1483 				break;
1484 
1485 			case IP_RECVDSTADDR:
1486 				optval = OPTBIT(INP_RECVDSTADDR);
1487 				break;
1488 
1489 			case IP_ORIGDSTADDR:
1490 				optval = OPTBIT2(INP_ORIGDSTADDR);
1491 				break;
1492 
1493 			case IP_RECVTTL:
1494 				optval = OPTBIT(INP_RECVTTL);
1495 				break;
1496 
1497 			case IP_RECVIF:
1498 				optval = OPTBIT(INP_RECVIF);
1499 				break;
1500 
1501 			case IP_PORTRANGE:
1502 				if (inp->inp_flags & INP_HIGHPORT)
1503 					optval = IP_PORTRANGE_HIGH;
1504 				else if (inp->inp_flags & INP_LOWPORT)
1505 					optval = IP_PORTRANGE_LOW;
1506 				else
1507 					optval = 0;
1508 				break;
1509 
1510 			case IP_ONESBCAST:
1511 				optval = OPTBIT(INP_ONESBCAST);
1512 				break;
1513 			case IP_DONTFRAG:
1514 				optval = OPTBIT(INP_DONTFRAG);
1515 				break;
1516 			case IP_BINDANY:
1517 				optval = OPTBIT(INP_BINDANY);
1518 				break;
1519 			case IP_RECVTOS:
1520 				optval = OPTBIT(INP_RECVTOS);
1521 				break;
1522 			case IP_FLOWID:
1523 				optval = inp->inp_flowid;
1524 				break;
1525 			case IP_FLOWTYPE:
1526 				optval = inp->inp_flowtype;
1527 				break;
1528 			case IP_RECVFLOWID:
1529 				optval = OPTBIT2(INP_RECVFLOWID);
1530 				break;
1531 #ifdef	RSS
1532 			case IP_RSSBUCKETID:
1533 				retval = rss_hash2bucket(inp->inp_flowid,
1534 				    inp->inp_flowtype,
1535 				    &rss_bucket);
1536 				if (retval == 0)
1537 					optval = rss_bucket;
1538 				else
1539 					error = EINVAL;
1540 				break;
1541 			case IP_RECVRSSBUCKETID:
1542 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1543 				break;
1544 #endif
1545 			case IP_BINDMULTI:
1546 				optval = OPTBIT2(INP_BINDMULTI);
1547 				break;
1548 			case IP_VLAN_PCP:
1549 				if (OPTBIT2(INP_2PCP_SET)) {
1550 					optval = (inp->inp_flags2 &
1551 					    INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1552 				} else {
1553 					optval = -1;
1554 				}
1555 				break;
1556 			}
1557 			error = sooptcopyout(sopt, &optval, sizeof optval);
1558 			break;
1559 
1560 		/*
1561 		 * Multicast socket options are processed by the in_mcast
1562 		 * module.
1563 		 */
1564 		case IP_MULTICAST_IF:
1565 		case IP_MULTICAST_VIF:
1566 		case IP_MULTICAST_TTL:
1567 		case IP_MULTICAST_LOOP:
1568 		case IP_MSFILTER:
1569 			error = inp_getmoptions(inp, sopt);
1570 			break;
1571 
1572 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1573 		case IP_IPSEC_POLICY:
1574 			if (IPSEC_ENABLED(ipv4)) {
1575 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1576 				break;
1577 			}
1578 			/* FALLTHROUGH */
1579 #endif /* IPSEC */
1580 
1581 		default:
1582 			error = ENOPROTOOPT;
1583 			break;
1584 		}
1585 		break;
1586 	}
1587 	return (error);
1588 }
1589 
1590 /*
1591  * Routine called from ip_output() to loop back a copy of an IP multicast
1592  * packet to the input queue of a specified interface.  Note that this
1593  * calls the output routine of the loopback "driver", but with an interface
1594  * pointer that might NOT be a loopback interface -- evil, but easier than
1595  * replicating that code here.
1596  */
1597 static void
1598 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1599 {
1600 	struct ip *ip;
1601 	struct mbuf *copym;
1602 
1603 	/*
1604 	 * Make a deep copy of the packet because we're going to
1605 	 * modify the pack in order to generate checksums.
1606 	 */
1607 	copym = m_dup(m, M_NOWAIT);
1608 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1609 		copym = m_pullup(copym, hlen);
1610 	if (copym != NULL) {
1611 		/* If needed, compute the checksum and mark it as valid. */
1612 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1613 			in_delayed_cksum(copym);
1614 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1615 			copym->m_pkthdr.csum_flags |=
1616 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1617 			copym->m_pkthdr.csum_data = 0xffff;
1618 		}
1619 		/*
1620 		 * We don't bother to fragment if the IP length is greater
1621 		 * than the interface's MTU.  Can this possibly matter?
1622 		 */
1623 		ip = mtod(copym, struct ip *);
1624 		ip->ip_sum = 0;
1625 		ip->ip_sum = in_cksum(copym, hlen);
1626 		if_simloop(ifp, copym, AF_INET, 0);
1627 	}
1628 }
1629