xref: /freebsd/sys/netinet/ip_output.c (revision b88cc53d4ddda4486683ee2121f131b10ed21c30)
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_mbuf_stress_test.h"
40 #include "opt_mpath.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/lock.h>
50 #include <sys/malloc.h>
51 #include <sys/mbuf.h>
52 #include <sys/priv.h>
53 #include <sys/proc.h>
54 #include <sys/protosw.h>
55 #include <sys/rmlock.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_llatbl.h>
65 #include <net/netisr.h>
66 #include <net/pfil.h>
67 #include <net/route.h>
68 #ifdef RADIX_MPATH
69 #include <net/radix_mpath.h>
70 #endif
71 #include <net/rss_config.h>
72 #include <net/vnet.h>
73 
74 #include <netinet/in.h>
75 #include <netinet/in_fib.h>
76 #include <netinet/in_kdtrace.h>
77 #include <netinet/in_systm.h>
78 #include <netinet/ip.h>
79 #include <netinet/in_pcb.h>
80 #include <netinet/in_rss.h>
81 #include <netinet/in_var.h>
82 #include <netinet/ip_var.h>
83 #include <netinet/ip_options.h>
84 
85 #include <netinet/udp.h>
86 #include <netinet/udp_var.h>
87 
88 #ifdef SCTP
89 #include <netinet/sctp.h>
90 #include <netinet/sctp_crc32.h>
91 #endif
92 
93 #include <netipsec/ipsec_support.h>
94 
95 #include <machine/in_cksum.h>
96 
97 #include <security/mac/mac_framework.h>
98 
99 #ifdef MBUF_STRESS_TEST
100 static int mbuf_frag_size = 0;
101 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
102 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
103 #endif
104 
105 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
106 
107 
108 extern int in_mcast_loop;
109 extern	struct protosw inetsw[];
110 
111 static inline int
112 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
113     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
114 {
115 	struct m_tag *fwd_tag = NULL;
116 	struct mbuf *m;
117 	struct in_addr odst;
118 	struct ip *ip;
119 	int pflags = PFIL_OUT;
120 
121 	if (flags & IP_FORWARDING)
122 		pflags |= PFIL_FWD;
123 
124 	m = *mp;
125 	ip = mtod(m, struct ip *);
126 
127 	/* Run through list of hooks for output packets. */
128 	odst.s_addr = ip->ip_dst.s_addr;
129 	switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
130 	case PFIL_DROPPED:
131 		*error = EPERM;
132 		/* FALLTHROUGH */
133 	case PFIL_CONSUMED:
134 		return 1; /* Finished */
135 	case PFIL_PASS:
136 		*error = 0;
137 	}
138 	m = *mp;
139 	ip = mtod(m, struct ip *);
140 
141 	/* See if destination IP address was changed by packet filter. */
142 	if (odst.s_addr != ip->ip_dst.s_addr) {
143 		m->m_flags |= M_SKIP_FIREWALL;
144 		/* If destination is now ourself drop to ip_input(). */
145 		if (in_localip(ip->ip_dst)) {
146 			m->m_flags |= M_FASTFWD_OURS;
147 			if (m->m_pkthdr.rcvif == NULL)
148 				m->m_pkthdr.rcvif = V_loif;
149 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
150 				m->m_pkthdr.csum_flags |=
151 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
152 				m->m_pkthdr.csum_data = 0xffff;
153 			}
154 			m->m_pkthdr.csum_flags |=
155 				CSUM_IP_CHECKED | CSUM_IP_VALID;
156 #ifdef SCTP
157 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
158 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
159 #endif
160 			*error = netisr_queue(NETISR_IP, m);
161 			return 1; /* Finished */
162 		}
163 
164 		bzero(dst, sizeof(*dst));
165 		dst->sin_family = AF_INET;
166 		dst->sin_len = sizeof(*dst);
167 		dst->sin_addr = ip->ip_dst;
168 
169 		return -1; /* Reloop */
170 	}
171 	/* See if fib was changed by packet filter. */
172 	if ((*fibnum) != M_GETFIB(m)) {
173 		m->m_flags |= M_SKIP_FIREWALL;
174 		*fibnum = M_GETFIB(m);
175 		return -1; /* Reloop for FIB change */
176 	}
177 
178 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
179 	if (m->m_flags & M_FASTFWD_OURS) {
180 		if (m->m_pkthdr.rcvif == NULL)
181 			m->m_pkthdr.rcvif = V_loif;
182 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
183 			m->m_pkthdr.csum_flags |=
184 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
185 			m->m_pkthdr.csum_data = 0xffff;
186 		}
187 #ifdef SCTP
188 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
189 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
190 #endif
191 		m->m_pkthdr.csum_flags |=
192 			CSUM_IP_CHECKED | CSUM_IP_VALID;
193 
194 		*error = netisr_queue(NETISR_IP, m);
195 		return 1; /* Finished */
196 	}
197 	/* Or forward to some other address? */
198 	if ((m->m_flags & M_IP_NEXTHOP) &&
199 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
200 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
201 		m->m_flags |= M_SKIP_FIREWALL;
202 		m->m_flags &= ~M_IP_NEXTHOP;
203 		m_tag_delete(m, fwd_tag);
204 
205 		return -1; /* Reloop for CHANGE of dst */
206 	}
207 
208 	return 0;
209 }
210 
211 static int
212 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
213     const struct sockaddr_in *gw, struct route *ro)
214 {
215 	struct m_snd_tag *mst;
216 	int error;
217 
218 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
219 	mst = NULL;
220 
221 #ifdef RATELIMIT
222 	if (inp != NULL) {
223 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
224 		    (inp->inp_snd_tag != NULL &&
225 		    inp->inp_snd_tag->ifp != ifp))
226 			in_pcboutput_txrtlmt(inp, ifp, m);
227 
228 		if (inp->inp_snd_tag != NULL)
229 			mst = inp->inp_snd_tag;
230 	}
231 #endif
232 	if (mst != NULL) {
233 		KASSERT(m->m_pkthdr.rcvif == NULL,
234 		    ("trying to add a send tag to a forwarded packet"));
235 		if (mst->ifp != ifp) {
236 			error = EAGAIN;
237 			goto done;
238 		}
239 
240 		/* stamp send tag on mbuf */
241 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
242 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
243 	}
244 
245 	error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
246 
247 done:
248 	/* Check for route change invalidating send tags. */
249 #ifdef RATELIMIT
250 	if (error == EAGAIN)
251 		in_pcboutput_eagain(inp);
252 #endif
253 	return (error);
254 }
255 
256 /*
257  * IP output.  The packet in mbuf chain m contains a skeletal IP
258  * header (with len, off, ttl, proto, tos, src, dst).
259  * The mbuf chain containing the packet will be freed.
260  * The mbuf opt, if present, will not be freed.
261  * If route ro is present and has ro_rt initialized, route lookup would be
262  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
263  * then result of route lookup is stored in ro->ro_rt.
264  *
265  * In the IP forwarding case, the packet will arrive with options already
266  * inserted, so must have a NULL opt pointer.
267  */
268 int
269 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
270     struct ip_moptions *imo, struct inpcb *inp)
271 {
272 	struct rm_priotracker in_ifa_tracker;
273 	struct epoch_tracker et;
274 	struct ip *ip;
275 	struct ifnet *ifp = NULL;	/* keep compiler happy */
276 	struct mbuf *m0;
277 	int hlen = sizeof (struct ip);
278 	int mtu;
279 	int error = 0;
280 	struct sockaddr_in *dst, sin;
281 	const struct sockaddr_in *gw;
282 	struct in_ifaddr *ia;
283 	struct in_addr src;
284 	int isbroadcast;
285 	uint16_t ip_len, ip_off;
286 	uint32_t fibnum;
287 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
288 	int no_route_but_check_spd = 0;
289 #endif
290 
291 	M_ASSERTPKTHDR(m);
292 
293 	if (inp != NULL) {
294 		INP_LOCK_ASSERT(inp);
295 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
296 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
297 			m->m_pkthdr.flowid = inp->inp_flowid;
298 			M_HASHTYPE_SET(m, inp->inp_flowtype);
299 		}
300 #ifdef NUMA
301 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
302 #endif
303 	}
304 
305 	if (opt) {
306 		int len = 0;
307 		m = ip_insertoptions(m, opt, &len);
308 		if (len != 0)
309 			hlen = len; /* ip->ip_hl is updated above */
310 	}
311 	ip = mtod(m, struct ip *);
312 	ip_len = ntohs(ip->ip_len);
313 	ip_off = ntohs(ip->ip_off);
314 
315 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
316 		ip->ip_v = IPVERSION;
317 		ip->ip_hl = hlen >> 2;
318 		ip_fillid(ip);
319 	} else {
320 		/* Header already set, fetch hlen from there */
321 		hlen = ip->ip_hl << 2;
322 	}
323 	if ((flags & IP_FORWARDING) == 0)
324 		IPSTAT_INC(ips_localout);
325 
326 	/*
327 	 * dst/gw handling:
328 	 *
329 	 * gw is readonly but can point either to dst OR rt_gateway,
330 	 * therefore we need restore gw if we're redoing lookup.
331 	 */
332 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
333 	if (ro != NULL)
334 		dst = (struct sockaddr_in *)&ro->ro_dst;
335 	else
336 		dst = &sin;
337 	if (ro == NULL || ro->ro_rt == NULL) {
338 		bzero(dst, sizeof(*dst));
339 		dst->sin_family = AF_INET;
340 		dst->sin_len = sizeof(*dst);
341 		dst->sin_addr = ip->ip_dst;
342 	}
343 	gw = dst;
344 	NET_EPOCH_ENTER(et);
345 again:
346 	/*
347 	 * Validate route against routing table additions;
348 	 * a better/more specific route might have been added.
349 	 */
350 	if (inp != NULL && ro != NULL && ro->ro_rt != NULL)
351 		RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
352 	/*
353 	 * If there is a cached route,
354 	 * check that it is to the same destination
355 	 * and is still up.  If not, free it and try again.
356 	 * The address family should also be checked in case of sharing the
357 	 * cache with IPv6.
358 	 * Also check whether routing cache needs invalidation.
359 	 */
360 	if (ro != NULL && ro->ro_rt != NULL &&
361 	    ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
362 	    ro->ro_rt->rt_ifp == NULL || !RT_LINK_IS_UP(ro->ro_rt->rt_ifp) ||
363 	    dst->sin_family != AF_INET ||
364 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
365 		RO_INVALIDATE_CACHE(ro);
366 	ia = NULL;
367 	/*
368 	 * If routing to interface only, short circuit routing lookup.
369 	 * The use of an all-ones broadcast address implies this; an
370 	 * interface is specified by the broadcast address of an interface,
371 	 * or the destination address of a ptp interface.
372 	 */
373 	if (flags & IP_SENDONES) {
374 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
375 						      M_GETFIB(m)))) == NULL &&
376 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
377 						    M_GETFIB(m)))) == NULL) {
378 			IPSTAT_INC(ips_noroute);
379 			error = ENETUNREACH;
380 			goto bad;
381 		}
382 		ip->ip_dst.s_addr = INADDR_BROADCAST;
383 		dst->sin_addr = ip->ip_dst;
384 		ifp = ia->ia_ifp;
385 		mtu = ifp->if_mtu;
386 		ip->ip_ttl = 1;
387 		isbroadcast = 1;
388 		src = IA_SIN(ia)->sin_addr;
389 	} else if (flags & IP_ROUTETOIF) {
390 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
391 						    M_GETFIB(m)))) == NULL &&
392 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
393 						M_GETFIB(m)))) == NULL) {
394 			IPSTAT_INC(ips_noroute);
395 			error = ENETUNREACH;
396 			goto bad;
397 		}
398 		ifp = ia->ia_ifp;
399 		mtu = ifp->if_mtu;
400 		ip->ip_ttl = 1;
401 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
402 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
403 		src = IA_SIN(ia)->sin_addr;
404 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
405 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
406 		/*
407 		 * Bypass the normal routing lookup for multicast
408 		 * packets if the interface is specified.
409 		 */
410 		ifp = imo->imo_multicast_ifp;
411 		mtu = ifp->if_mtu;
412 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
413 		isbroadcast = 0;	/* fool gcc */
414 		/* Interface may have no addresses. */
415 		if (ia != NULL)
416 			src = IA_SIN(ia)->sin_addr;
417 		else
418 			src.s_addr = INADDR_ANY;
419 	} else if (ro != NULL) {
420 		if (ro->ro_rt == NULL) {
421 			/*
422 			 * We want to do any cloning requested by the link
423 			 * layer, as this is probably required in all cases
424 			 * for correct operation (as it is for ARP).
425 			 */
426 #ifdef RADIX_MPATH
427 			rtalloc_mpath_fib(ro,
428 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
429 			    fibnum);
430 #else
431 			in_rtalloc_ign(ro, 0, fibnum);
432 #endif
433 			if (ro->ro_rt == NULL ||
434 			    (ro->ro_rt->rt_flags & RTF_UP) == 0 ||
435 			    ro->ro_rt->rt_ifp == NULL ||
436 			    !RT_LINK_IS_UP(ro->ro_rt->rt_ifp)) {
437 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
438 				/*
439 				 * There is no route for this packet, but it is
440 				 * possible that a matching SPD entry exists.
441 				 */
442 				no_route_but_check_spd = 1;
443 				mtu = 0; /* Silence GCC warning. */
444 				goto sendit;
445 #endif
446 				IPSTAT_INC(ips_noroute);
447 				error = EHOSTUNREACH;
448 				goto bad;
449 			}
450 		}
451 		ia = ifatoia(ro->ro_rt->rt_ifa);
452 		ifp = ro->ro_rt->rt_ifp;
453 		counter_u64_add(ro->ro_rt->rt_pksent, 1);
454 		rt_update_ro_flags(ro);
455 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
456 			gw = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
457 		if (ro->ro_rt->rt_flags & RTF_HOST)
458 			isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
459 		else if (ifp->if_flags & IFF_BROADCAST)
460 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
461 		else
462 			isbroadcast = 0;
463 		if (ro->ro_rt->rt_flags & RTF_HOST)
464 			mtu = ro->ro_rt->rt_mtu;
465 		else
466 			mtu = ifp->if_mtu;
467 		src = IA_SIN(ia)->sin_addr;
468 	} else {
469 		struct nhop4_extended nh;
470 
471 		bzero(&nh, sizeof(nh));
472 		if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh) !=
473 		    0) {
474 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
475 			/*
476 			 * There is no route for this packet, but it is
477 			 * possible that a matching SPD entry exists.
478 			 */
479 			no_route_but_check_spd = 1;
480 			mtu = 0; /* Silence GCC warning. */
481 			goto sendit;
482 #endif
483 			IPSTAT_INC(ips_noroute);
484 			error = EHOSTUNREACH;
485 			goto bad;
486 		}
487 		ifp = nh.nh_ifp;
488 		mtu = nh.nh_mtu;
489 		/*
490 		 * We are rewriting here dst to be gw actually, contradicting
491 		 * comment at the beginning of the function. However, in this
492 		 * case we are always dealing with on stack dst.
493 		 * In case if pfil(9) sends us back to beginning of the
494 		 * function, the dst would be rewritten by ip_output_pfil().
495 		 */
496 		MPASS(dst == &sin);
497 		dst->sin_addr = nh.nh_addr;
498 		ia = nh.nh_ia;
499 		src = nh.nh_src;
500 		isbroadcast = (((nh.nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
501 		    (NHF_HOST | NHF_BROADCAST)) ||
502 		    ((ifp->if_flags & IFF_BROADCAST) &&
503 		    in_ifaddr_broadcast(dst->sin_addr, ia)));
504 	}
505 
506 	/* Catch a possible divide by zero later. */
507 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (rt_flags=0x%08x) ifp=%p",
508 	    __func__, mtu, ro,
509 	    (ro != NULL && ro->ro_rt != NULL) ? ro->ro_rt->rt_flags : 0, ifp));
510 
511 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
512 		m->m_flags |= M_MCAST;
513 		/*
514 		 * IP destination address is multicast.  Make sure "gw"
515 		 * still points to the address in "ro".  (It may have been
516 		 * changed to point to a gateway address, above.)
517 		 */
518 		gw = dst;
519 		/*
520 		 * See if the caller provided any multicast options
521 		 */
522 		if (imo != NULL) {
523 			ip->ip_ttl = imo->imo_multicast_ttl;
524 			if (imo->imo_multicast_vif != -1)
525 				ip->ip_src.s_addr =
526 				    ip_mcast_src ?
527 				    ip_mcast_src(imo->imo_multicast_vif) :
528 				    INADDR_ANY;
529 		} else
530 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
531 		/*
532 		 * Confirm that the outgoing interface supports multicast.
533 		 */
534 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
535 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
536 				IPSTAT_INC(ips_noroute);
537 				error = ENETUNREACH;
538 				goto bad;
539 			}
540 		}
541 		/*
542 		 * If source address not specified yet, use address
543 		 * of outgoing interface.
544 		 */
545 		if (ip->ip_src.s_addr == INADDR_ANY)
546 			ip->ip_src = src;
547 
548 		if ((imo == NULL && in_mcast_loop) ||
549 		    (imo && imo->imo_multicast_loop)) {
550 			/*
551 			 * Loop back multicast datagram if not expressly
552 			 * forbidden to do so, even if we are not a member
553 			 * of the group; ip_input() will filter it later,
554 			 * thus deferring a hash lookup and mutex acquisition
555 			 * at the expense of a cheap copy using m_copym().
556 			 */
557 			ip_mloopback(ifp, m, hlen);
558 		} else {
559 			/*
560 			 * If we are acting as a multicast router, perform
561 			 * multicast forwarding as if the packet had just
562 			 * arrived on the interface to which we are about
563 			 * to send.  The multicast forwarding function
564 			 * recursively calls this function, using the
565 			 * IP_FORWARDING flag to prevent infinite recursion.
566 			 *
567 			 * Multicasts that are looped back by ip_mloopback(),
568 			 * above, will be forwarded by the ip_input() routine,
569 			 * if necessary.
570 			 */
571 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
572 				/*
573 				 * If rsvp daemon is not running, do not
574 				 * set ip_moptions. This ensures that the packet
575 				 * is multicast and not just sent down one link
576 				 * as prescribed by rsvpd.
577 				 */
578 				if (!V_rsvp_on)
579 					imo = NULL;
580 				if (ip_mforward &&
581 				    ip_mforward(ip, ifp, m, imo) != 0) {
582 					m_freem(m);
583 					goto done;
584 				}
585 			}
586 		}
587 
588 		/*
589 		 * Multicasts with a time-to-live of zero may be looped-
590 		 * back, above, but must not be transmitted on a network.
591 		 * Also, multicasts addressed to the loopback interface
592 		 * are not sent -- the above call to ip_mloopback() will
593 		 * loop back a copy. ip_input() will drop the copy if
594 		 * this host does not belong to the destination group on
595 		 * the loopback interface.
596 		 */
597 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
598 			m_freem(m);
599 			goto done;
600 		}
601 
602 		goto sendit;
603 	}
604 
605 	/*
606 	 * If the source address is not specified yet, use the address
607 	 * of the outoing interface.
608 	 */
609 	if (ip->ip_src.s_addr == INADDR_ANY)
610 		ip->ip_src = src;
611 
612 	/*
613 	 * Look for broadcast address and
614 	 * verify user is allowed to send
615 	 * such a packet.
616 	 */
617 	if (isbroadcast) {
618 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
619 			error = EADDRNOTAVAIL;
620 			goto bad;
621 		}
622 		if ((flags & IP_ALLOWBROADCAST) == 0) {
623 			error = EACCES;
624 			goto bad;
625 		}
626 		/* don't allow broadcast messages to be fragmented */
627 		if (ip_len > mtu) {
628 			error = EMSGSIZE;
629 			goto bad;
630 		}
631 		m->m_flags |= M_BCAST;
632 	} else {
633 		m->m_flags &= ~M_BCAST;
634 	}
635 
636 sendit:
637 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
638 	if (IPSEC_ENABLED(ipv4)) {
639 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
640 			if (error == EINPROGRESS)
641 				error = 0;
642 			goto done;
643 		}
644 	}
645 	/*
646 	 * Check if there was a route for this packet; return error if not.
647 	 */
648 	if (no_route_but_check_spd) {
649 		IPSTAT_INC(ips_noroute);
650 		error = EHOSTUNREACH;
651 		goto bad;
652 	}
653 	/* Update variables that are affected by ipsec4_output(). */
654 	ip = mtod(m, struct ip *);
655 	hlen = ip->ip_hl << 2;
656 #endif /* IPSEC */
657 
658 	/* Jump over all PFIL processing if hooks are not active. */
659 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
660 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
661 		    &error)) {
662 		case 1: /* Finished */
663 			goto done;
664 
665 		case 0: /* Continue normally */
666 			ip = mtod(m, struct ip *);
667 			break;
668 
669 		case -1: /* Need to try again */
670 			/* Reset everything for a new round */
671 			if (ro != NULL) {
672 				RO_RTFREE(ro);
673 				ro->ro_prepend = NULL;
674 			}
675 			gw = dst;
676 			ip = mtod(m, struct ip *);
677 			goto again;
678 
679 		}
680 	}
681 
682 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
683 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
684 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
685 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
686 			IPSTAT_INC(ips_badaddr);
687 			error = EADDRNOTAVAIL;
688 			goto bad;
689 		}
690 	}
691 
692 	m->m_pkthdr.csum_flags |= CSUM_IP;
693 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
694 		in_delayed_cksum(m);
695 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
696 	}
697 #ifdef SCTP
698 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
699 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
700 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
701 	}
702 #endif
703 
704 	/*
705 	 * If small enough for interface, or the interface will take
706 	 * care of the fragmentation for us, we can just send directly.
707 	 */
708 	if (ip_len <= mtu ||
709 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
710 		ip->ip_sum = 0;
711 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
712 			ip->ip_sum = in_cksum(m, hlen);
713 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
714 		}
715 
716 		/*
717 		 * Record statistics for this interface address.
718 		 * With CSUM_TSO the byte/packet count will be slightly
719 		 * incorrect because we count the IP+TCP headers only
720 		 * once instead of for every generated packet.
721 		 */
722 		if (!(flags & IP_FORWARDING) && ia) {
723 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
724 				counter_u64_add(ia->ia_ifa.ifa_opackets,
725 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
726 			else
727 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
728 
729 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
730 		}
731 #ifdef MBUF_STRESS_TEST
732 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
733 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
734 #endif
735 		/*
736 		 * Reset layer specific mbuf flags
737 		 * to avoid confusing lower layers.
738 		 */
739 		m_clrprotoflags(m);
740 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
741 		error = ip_output_send(inp, ifp, m, gw, ro);
742 		goto done;
743 	}
744 
745 	/* Balk when DF bit is set or the interface didn't support TSO. */
746 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
747 		error = EMSGSIZE;
748 		IPSTAT_INC(ips_cantfrag);
749 		goto bad;
750 	}
751 
752 	/*
753 	 * Too large for interface; fragment if possible. If successful,
754 	 * on return, m will point to a list of packets to be sent.
755 	 */
756 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
757 	if (error)
758 		goto bad;
759 	for (; m; m = m0) {
760 		m0 = m->m_nextpkt;
761 		m->m_nextpkt = 0;
762 		if (error == 0) {
763 			/* Record statistics for this interface address. */
764 			if (ia != NULL) {
765 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
766 				counter_u64_add(ia->ia_ifa.ifa_obytes,
767 				    m->m_pkthdr.len);
768 			}
769 			/*
770 			 * Reset layer specific mbuf flags
771 			 * to avoid confusing upper layers.
772 			 */
773 			m_clrprotoflags(m);
774 
775 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
776 			    mtod(m, struct ip *), NULL);
777 			error = ip_output_send(inp, ifp, m, gw, ro);
778 		} else
779 			m_freem(m);
780 	}
781 
782 	if (error == 0)
783 		IPSTAT_INC(ips_fragmented);
784 
785 done:
786 	NET_EPOCH_EXIT(et);
787 	return (error);
788  bad:
789 	m_freem(m);
790 	goto done;
791 }
792 
793 /*
794  * Create a chain of fragments which fit the given mtu. m_frag points to the
795  * mbuf to be fragmented; on return it points to the chain with the fragments.
796  * Return 0 if no error. If error, m_frag may contain a partially built
797  * chain of fragments that should be freed by the caller.
798  *
799  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
800  */
801 int
802 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
803     u_long if_hwassist_flags)
804 {
805 	int error = 0;
806 	int hlen = ip->ip_hl << 2;
807 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
808 	int off;
809 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
810 	int firstlen;
811 	struct mbuf **mnext;
812 	int nfrags;
813 	uint16_t ip_len, ip_off;
814 
815 	ip_len = ntohs(ip->ip_len);
816 	ip_off = ntohs(ip->ip_off);
817 
818 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
819 		IPSTAT_INC(ips_cantfrag);
820 		return EMSGSIZE;
821 	}
822 
823 	/*
824 	 * Must be able to put at least 8 bytes per fragment.
825 	 */
826 	if (len < 8)
827 		return EMSGSIZE;
828 
829 	/*
830 	 * If the interface will not calculate checksums on
831 	 * fragmented packets, then do it here.
832 	 */
833 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
834 		in_delayed_cksum(m0);
835 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
836 	}
837 #ifdef SCTP
838 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
839 		sctp_delayed_cksum(m0, hlen);
840 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
841 	}
842 #endif
843 	if (len > PAGE_SIZE) {
844 		/*
845 		 * Fragment large datagrams such that each segment
846 		 * contains a multiple of PAGE_SIZE amount of data,
847 		 * plus headers. This enables a receiver to perform
848 		 * page-flipping zero-copy optimizations.
849 		 *
850 		 * XXX When does this help given that sender and receiver
851 		 * could have different page sizes, and also mtu could
852 		 * be less than the receiver's page size ?
853 		 */
854 		int newlen;
855 
856 		off = MIN(mtu, m0->m_pkthdr.len);
857 
858 		/*
859 		 * firstlen (off - hlen) must be aligned on an
860 		 * 8-byte boundary
861 		 */
862 		if (off < hlen)
863 			goto smart_frag_failure;
864 		off = ((off - hlen) & ~7) + hlen;
865 		newlen = (~PAGE_MASK) & mtu;
866 		if ((newlen + sizeof (struct ip)) > mtu) {
867 			/* we failed, go back the default */
868 smart_frag_failure:
869 			newlen = len;
870 			off = hlen + len;
871 		}
872 		len = newlen;
873 
874 	} else {
875 		off = hlen + len;
876 	}
877 
878 	firstlen = off - hlen;
879 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
880 
881 	/*
882 	 * Loop through length of segment after first fragment,
883 	 * make new header and copy data of each part and link onto chain.
884 	 * Here, m0 is the original packet, m is the fragment being created.
885 	 * The fragments are linked off the m_nextpkt of the original
886 	 * packet, which after processing serves as the first fragment.
887 	 */
888 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
889 		struct ip *mhip;	/* ip header on the fragment */
890 		struct mbuf *m;
891 		int mhlen = sizeof (struct ip);
892 
893 		m = m_gethdr(M_NOWAIT, MT_DATA);
894 		if (m == NULL) {
895 			error = ENOBUFS;
896 			IPSTAT_INC(ips_odropped);
897 			goto done;
898 		}
899 		/*
900 		 * Make sure the complete packet header gets copied
901 		 * from the originating mbuf to the newly created
902 		 * mbuf. This also ensures that existing firewall
903 		 * classification(s), VLAN tags and so on get copied
904 		 * to the resulting fragmented packet(s):
905 		 */
906 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
907 			m_free(m);
908 			error = ENOBUFS;
909 			IPSTAT_INC(ips_odropped);
910 			goto done;
911 		}
912 		/*
913 		 * In the first mbuf, leave room for the link header, then
914 		 * copy the original IP header including options. The payload
915 		 * goes into an additional mbuf chain returned by m_copym().
916 		 */
917 		m->m_data += max_linkhdr;
918 		mhip = mtod(m, struct ip *);
919 		*mhip = *ip;
920 		if (hlen > sizeof (struct ip)) {
921 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
922 			mhip->ip_v = IPVERSION;
923 			mhip->ip_hl = mhlen >> 2;
924 		}
925 		m->m_len = mhlen;
926 		/* XXX do we need to add ip_off below ? */
927 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
928 		if (off + len >= ip_len)
929 			len = ip_len - off;
930 		else
931 			mhip->ip_off |= IP_MF;
932 		mhip->ip_len = htons((u_short)(len + mhlen));
933 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
934 		if (m->m_next == NULL) {	/* copy failed */
935 			m_free(m);
936 			error = ENOBUFS;	/* ??? */
937 			IPSTAT_INC(ips_odropped);
938 			goto done;
939 		}
940 		m->m_pkthdr.len = mhlen + len;
941 #ifdef MAC
942 		mac_netinet_fragment(m0, m);
943 #endif
944 		mhip->ip_off = htons(mhip->ip_off);
945 		mhip->ip_sum = 0;
946 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
947 			mhip->ip_sum = in_cksum(m, mhlen);
948 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
949 		}
950 		*mnext = m;
951 		mnext = &m->m_nextpkt;
952 	}
953 	IPSTAT_ADD(ips_ofragments, nfrags);
954 
955 	/*
956 	 * Update first fragment by trimming what's been copied out
957 	 * and updating header.
958 	 */
959 	m_adj(m0, hlen + firstlen - ip_len);
960 	m0->m_pkthdr.len = hlen + firstlen;
961 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
962 	ip->ip_off = htons(ip_off | IP_MF);
963 	ip->ip_sum = 0;
964 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
965 		ip->ip_sum = in_cksum(m0, hlen);
966 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
967 	}
968 
969 done:
970 	*m_frag = m0;
971 	return error;
972 }
973 
974 void
975 in_delayed_cksum(struct mbuf *m)
976 {
977 	struct ip *ip;
978 	struct udphdr *uh;
979 	uint16_t cklen, csum, offset;
980 
981 	ip = mtod(m, struct ip *);
982 	offset = ip->ip_hl << 2 ;
983 
984 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
985 		/* if udp header is not in the first mbuf copy udplen */
986 		if (offset + sizeof(struct udphdr) > m->m_len) {
987 			m_copydata(m, offset + offsetof(struct udphdr,
988 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
989 			cklen = ntohs(cklen);
990 		} else {
991 			uh = (struct udphdr *)mtodo(m, offset);
992 			cklen = ntohs(uh->uh_ulen);
993 		}
994 		csum = in_cksum_skip(m, cklen + offset, offset);
995 		if (csum == 0)
996 			csum = 0xffff;
997 	} else {
998 		cklen = ntohs(ip->ip_len);
999 		csum = in_cksum_skip(m, cklen, offset);
1000 	}
1001 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1002 
1003 	if (offset + sizeof(csum) > m->m_len)
1004 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1005 	else
1006 		*(u_short *)mtodo(m, offset) = csum;
1007 }
1008 
1009 /*
1010  * IP socket option processing.
1011  */
1012 int
1013 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1014 {
1015 	struct	inpcb *inp = sotoinpcb(so);
1016 	int	error, optval;
1017 #ifdef	RSS
1018 	uint32_t rss_bucket;
1019 	int retval;
1020 #endif
1021 
1022 	error = optval = 0;
1023 	if (sopt->sopt_level != IPPROTO_IP) {
1024 		error = EINVAL;
1025 
1026 		if (sopt->sopt_level == SOL_SOCKET &&
1027 		    sopt->sopt_dir == SOPT_SET) {
1028 			switch (sopt->sopt_name) {
1029 			case SO_REUSEADDR:
1030 				INP_WLOCK(inp);
1031 				if ((so->so_options & SO_REUSEADDR) != 0)
1032 					inp->inp_flags2 |= INP_REUSEADDR;
1033 				else
1034 					inp->inp_flags2 &= ~INP_REUSEADDR;
1035 				INP_WUNLOCK(inp);
1036 				error = 0;
1037 				break;
1038 			case SO_REUSEPORT:
1039 				INP_WLOCK(inp);
1040 				if ((so->so_options & SO_REUSEPORT) != 0)
1041 					inp->inp_flags2 |= INP_REUSEPORT;
1042 				else
1043 					inp->inp_flags2 &= ~INP_REUSEPORT;
1044 				INP_WUNLOCK(inp);
1045 				error = 0;
1046 				break;
1047 			case SO_REUSEPORT_LB:
1048 				INP_WLOCK(inp);
1049 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1050 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1051 				else
1052 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1053 				INP_WUNLOCK(inp);
1054 				error = 0;
1055 				break;
1056 			case SO_SETFIB:
1057 				INP_WLOCK(inp);
1058 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1059 				INP_WUNLOCK(inp);
1060 				error = 0;
1061 				break;
1062 			case SO_MAX_PACING_RATE:
1063 #ifdef RATELIMIT
1064 				INP_WLOCK(inp);
1065 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1066 				INP_WUNLOCK(inp);
1067 				error = 0;
1068 #else
1069 				error = EOPNOTSUPP;
1070 #endif
1071 				break;
1072 			default:
1073 				break;
1074 			}
1075 		}
1076 		return (error);
1077 	}
1078 
1079 	switch (sopt->sopt_dir) {
1080 	case SOPT_SET:
1081 		switch (sopt->sopt_name) {
1082 		case IP_OPTIONS:
1083 #ifdef notyet
1084 		case IP_RETOPTS:
1085 #endif
1086 		{
1087 			struct mbuf *m;
1088 			if (sopt->sopt_valsize > MLEN) {
1089 				error = EMSGSIZE;
1090 				break;
1091 			}
1092 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1093 			if (m == NULL) {
1094 				error = ENOBUFS;
1095 				break;
1096 			}
1097 			m->m_len = sopt->sopt_valsize;
1098 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1099 					    m->m_len);
1100 			if (error) {
1101 				m_free(m);
1102 				break;
1103 			}
1104 			INP_WLOCK(inp);
1105 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1106 			INP_WUNLOCK(inp);
1107 			return (error);
1108 		}
1109 
1110 		case IP_BINDANY:
1111 			if (sopt->sopt_td != NULL) {
1112 				error = priv_check(sopt->sopt_td,
1113 				    PRIV_NETINET_BINDANY);
1114 				if (error)
1115 					break;
1116 			}
1117 			/* FALLTHROUGH */
1118 		case IP_BINDMULTI:
1119 #ifdef	RSS
1120 		case IP_RSS_LISTEN_BUCKET:
1121 #endif
1122 		case IP_TOS:
1123 		case IP_TTL:
1124 		case IP_MINTTL:
1125 		case IP_RECVOPTS:
1126 		case IP_RECVRETOPTS:
1127 		case IP_ORIGDSTADDR:
1128 		case IP_RECVDSTADDR:
1129 		case IP_RECVTTL:
1130 		case IP_RECVIF:
1131 		case IP_ONESBCAST:
1132 		case IP_DONTFRAG:
1133 		case IP_RECVTOS:
1134 		case IP_RECVFLOWID:
1135 #ifdef	RSS
1136 		case IP_RECVRSSBUCKETID:
1137 #endif
1138 			error = sooptcopyin(sopt, &optval, sizeof optval,
1139 					    sizeof optval);
1140 			if (error)
1141 				break;
1142 
1143 			switch (sopt->sopt_name) {
1144 			case IP_TOS:
1145 				inp->inp_ip_tos = optval;
1146 				break;
1147 
1148 			case IP_TTL:
1149 				inp->inp_ip_ttl = optval;
1150 				break;
1151 
1152 			case IP_MINTTL:
1153 				if (optval >= 0 && optval <= MAXTTL)
1154 					inp->inp_ip_minttl = optval;
1155 				else
1156 					error = EINVAL;
1157 				break;
1158 
1159 #define	OPTSET(bit) do {						\
1160 	INP_WLOCK(inp);							\
1161 	if (optval)							\
1162 		inp->inp_flags |= bit;					\
1163 	else								\
1164 		inp->inp_flags &= ~bit;					\
1165 	INP_WUNLOCK(inp);						\
1166 } while (0)
1167 
1168 #define	OPTSET2(bit, val) do {						\
1169 	INP_WLOCK(inp);							\
1170 	if (val)							\
1171 		inp->inp_flags2 |= bit;					\
1172 	else								\
1173 		inp->inp_flags2 &= ~bit;				\
1174 	INP_WUNLOCK(inp);						\
1175 } while (0)
1176 
1177 			case IP_RECVOPTS:
1178 				OPTSET(INP_RECVOPTS);
1179 				break;
1180 
1181 			case IP_RECVRETOPTS:
1182 				OPTSET(INP_RECVRETOPTS);
1183 				break;
1184 
1185 			case IP_RECVDSTADDR:
1186 				OPTSET(INP_RECVDSTADDR);
1187 				break;
1188 
1189 			case IP_ORIGDSTADDR:
1190 				OPTSET2(INP_ORIGDSTADDR, optval);
1191 				break;
1192 
1193 			case IP_RECVTTL:
1194 				OPTSET(INP_RECVTTL);
1195 				break;
1196 
1197 			case IP_RECVIF:
1198 				OPTSET(INP_RECVIF);
1199 				break;
1200 
1201 			case IP_ONESBCAST:
1202 				OPTSET(INP_ONESBCAST);
1203 				break;
1204 			case IP_DONTFRAG:
1205 				OPTSET(INP_DONTFRAG);
1206 				break;
1207 			case IP_BINDANY:
1208 				OPTSET(INP_BINDANY);
1209 				break;
1210 			case IP_RECVTOS:
1211 				OPTSET(INP_RECVTOS);
1212 				break;
1213 			case IP_BINDMULTI:
1214 				OPTSET2(INP_BINDMULTI, optval);
1215 				break;
1216 			case IP_RECVFLOWID:
1217 				OPTSET2(INP_RECVFLOWID, optval);
1218 				break;
1219 #ifdef	RSS
1220 			case IP_RSS_LISTEN_BUCKET:
1221 				if ((optval >= 0) &&
1222 				    (optval < rss_getnumbuckets())) {
1223 					inp->inp_rss_listen_bucket = optval;
1224 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1225 				} else {
1226 					error = EINVAL;
1227 				}
1228 				break;
1229 			case IP_RECVRSSBUCKETID:
1230 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1231 				break;
1232 #endif
1233 			}
1234 			break;
1235 #undef OPTSET
1236 #undef OPTSET2
1237 
1238 		/*
1239 		 * Multicast socket options are processed by the in_mcast
1240 		 * module.
1241 		 */
1242 		case IP_MULTICAST_IF:
1243 		case IP_MULTICAST_VIF:
1244 		case IP_MULTICAST_TTL:
1245 		case IP_MULTICAST_LOOP:
1246 		case IP_ADD_MEMBERSHIP:
1247 		case IP_DROP_MEMBERSHIP:
1248 		case IP_ADD_SOURCE_MEMBERSHIP:
1249 		case IP_DROP_SOURCE_MEMBERSHIP:
1250 		case IP_BLOCK_SOURCE:
1251 		case IP_UNBLOCK_SOURCE:
1252 		case IP_MSFILTER:
1253 		case MCAST_JOIN_GROUP:
1254 		case MCAST_LEAVE_GROUP:
1255 		case MCAST_JOIN_SOURCE_GROUP:
1256 		case MCAST_LEAVE_SOURCE_GROUP:
1257 		case MCAST_BLOCK_SOURCE:
1258 		case MCAST_UNBLOCK_SOURCE:
1259 			error = inp_setmoptions(inp, sopt);
1260 			break;
1261 
1262 		case IP_PORTRANGE:
1263 			error = sooptcopyin(sopt, &optval, sizeof optval,
1264 					    sizeof optval);
1265 			if (error)
1266 				break;
1267 
1268 			INP_WLOCK(inp);
1269 			switch (optval) {
1270 			case IP_PORTRANGE_DEFAULT:
1271 				inp->inp_flags &= ~(INP_LOWPORT);
1272 				inp->inp_flags &= ~(INP_HIGHPORT);
1273 				break;
1274 
1275 			case IP_PORTRANGE_HIGH:
1276 				inp->inp_flags &= ~(INP_LOWPORT);
1277 				inp->inp_flags |= INP_HIGHPORT;
1278 				break;
1279 
1280 			case IP_PORTRANGE_LOW:
1281 				inp->inp_flags &= ~(INP_HIGHPORT);
1282 				inp->inp_flags |= INP_LOWPORT;
1283 				break;
1284 
1285 			default:
1286 				error = EINVAL;
1287 				break;
1288 			}
1289 			INP_WUNLOCK(inp);
1290 			break;
1291 
1292 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1293 		case IP_IPSEC_POLICY:
1294 			if (IPSEC_ENABLED(ipv4)) {
1295 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1296 				break;
1297 			}
1298 			/* FALLTHROUGH */
1299 #endif /* IPSEC */
1300 
1301 		default:
1302 			error = ENOPROTOOPT;
1303 			break;
1304 		}
1305 		break;
1306 
1307 	case SOPT_GET:
1308 		switch (sopt->sopt_name) {
1309 		case IP_OPTIONS:
1310 		case IP_RETOPTS:
1311 			INP_RLOCK(inp);
1312 			if (inp->inp_options) {
1313 				struct mbuf *options;
1314 
1315 				options = m_copym(inp->inp_options, 0,
1316 				    M_COPYALL, M_NOWAIT);
1317 				INP_RUNLOCK(inp);
1318 				if (options != NULL) {
1319 					error = sooptcopyout(sopt,
1320 							     mtod(options, char *),
1321 							     options->m_len);
1322 					m_freem(options);
1323 				} else
1324 					error = ENOMEM;
1325 			} else {
1326 				INP_RUNLOCK(inp);
1327 				sopt->sopt_valsize = 0;
1328 			}
1329 			break;
1330 
1331 		case IP_TOS:
1332 		case IP_TTL:
1333 		case IP_MINTTL:
1334 		case IP_RECVOPTS:
1335 		case IP_RECVRETOPTS:
1336 		case IP_ORIGDSTADDR:
1337 		case IP_RECVDSTADDR:
1338 		case IP_RECVTTL:
1339 		case IP_RECVIF:
1340 		case IP_PORTRANGE:
1341 		case IP_ONESBCAST:
1342 		case IP_DONTFRAG:
1343 		case IP_BINDANY:
1344 		case IP_RECVTOS:
1345 		case IP_BINDMULTI:
1346 		case IP_FLOWID:
1347 		case IP_FLOWTYPE:
1348 		case IP_RECVFLOWID:
1349 #ifdef	RSS
1350 		case IP_RSSBUCKETID:
1351 		case IP_RECVRSSBUCKETID:
1352 #endif
1353 			switch (sopt->sopt_name) {
1354 
1355 			case IP_TOS:
1356 				optval = inp->inp_ip_tos;
1357 				break;
1358 
1359 			case IP_TTL:
1360 				optval = inp->inp_ip_ttl;
1361 				break;
1362 
1363 			case IP_MINTTL:
1364 				optval = inp->inp_ip_minttl;
1365 				break;
1366 
1367 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1368 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1369 
1370 			case IP_RECVOPTS:
1371 				optval = OPTBIT(INP_RECVOPTS);
1372 				break;
1373 
1374 			case IP_RECVRETOPTS:
1375 				optval = OPTBIT(INP_RECVRETOPTS);
1376 				break;
1377 
1378 			case IP_RECVDSTADDR:
1379 				optval = OPTBIT(INP_RECVDSTADDR);
1380 				break;
1381 
1382 			case IP_ORIGDSTADDR:
1383 				optval = OPTBIT2(INP_ORIGDSTADDR);
1384 				break;
1385 
1386 			case IP_RECVTTL:
1387 				optval = OPTBIT(INP_RECVTTL);
1388 				break;
1389 
1390 			case IP_RECVIF:
1391 				optval = OPTBIT(INP_RECVIF);
1392 				break;
1393 
1394 			case IP_PORTRANGE:
1395 				if (inp->inp_flags & INP_HIGHPORT)
1396 					optval = IP_PORTRANGE_HIGH;
1397 				else if (inp->inp_flags & INP_LOWPORT)
1398 					optval = IP_PORTRANGE_LOW;
1399 				else
1400 					optval = 0;
1401 				break;
1402 
1403 			case IP_ONESBCAST:
1404 				optval = OPTBIT(INP_ONESBCAST);
1405 				break;
1406 			case IP_DONTFRAG:
1407 				optval = OPTBIT(INP_DONTFRAG);
1408 				break;
1409 			case IP_BINDANY:
1410 				optval = OPTBIT(INP_BINDANY);
1411 				break;
1412 			case IP_RECVTOS:
1413 				optval = OPTBIT(INP_RECVTOS);
1414 				break;
1415 			case IP_FLOWID:
1416 				optval = inp->inp_flowid;
1417 				break;
1418 			case IP_FLOWTYPE:
1419 				optval = inp->inp_flowtype;
1420 				break;
1421 			case IP_RECVFLOWID:
1422 				optval = OPTBIT2(INP_RECVFLOWID);
1423 				break;
1424 #ifdef	RSS
1425 			case IP_RSSBUCKETID:
1426 				retval = rss_hash2bucket(inp->inp_flowid,
1427 				    inp->inp_flowtype,
1428 				    &rss_bucket);
1429 				if (retval == 0)
1430 					optval = rss_bucket;
1431 				else
1432 					error = EINVAL;
1433 				break;
1434 			case IP_RECVRSSBUCKETID:
1435 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1436 				break;
1437 #endif
1438 			case IP_BINDMULTI:
1439 				optval = OPTBIT2(INP_BINDMULTI);
1440 				break;
1441 			}
1442 			error = sooptcopyout(sopt, &optval, sizeof optval);
1443 			break;
1444 
1445 		/*
1446 		 * Multicast socket options are processed by the in_mcast
1447 		 * module.
1448 		 */
1449 		case IP_MULTICAST_IF:
1450 		case IP_MULTICAST_VIF:
1451 		case IP_MULTICAST_TTL:
1452 		case IP_MULTICAST_LOOP:
1453 		case IP_MSFILTER:
1454 			error = inp_getmoptions(inp, sopt);
1455 			break;
1456 
1457 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1458 		case IP_IPSEC_POLICY:
1459 			if (IPSEC_ENABLED(ipv4)) {
1460 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1461 				break;
1462 			}
1463 			/* FALLTHROUGH */
1464 #endif /* IPSEC */
1465 
1466 		default:
1467 			error = ENOPROTOOPT;
1468 			break;
1469 		}
1470 		break;
1471 	}
1472 	return (error);
1473 }
1474 
1475 /*
1476  * Routine called from ip_output() to loop back a copy of an IP multicast
1477  * packet to the input queue of a specified interface.  Note that this
1478  * calls the output routine of the loopback "driver", but with an interface
1479  * pointer that might NOT be a loopback interface -- evil, but easier than
1480  * replicating that code here.
1481  */
1482 static void
1483 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1484 {
1485 	struct ip *ip;
1486 	struct mbuf *copym;
1487 
1488 	/*
1489 	 * Make a deep copy of the packet because we're going to
1490 	 * modify the pack in order to generate checksums.
1491 	 */
1492 	copym = m_dup(m, M_NOWAIT);
1493 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1494 		copym = m_pullup(copym, hlen);
1495 	if (copym != NULL) {
1496 		/* If needed, compute the checksum and mark it as valid. */
1497 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1498 			in_delayed_cksum(copym);
1499 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1500 			copym->m_pkthdr.csum_flags |=
1501 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1502 			copym->m_pkthdr.csum_data = 0xffff;
1503 		}
1504 		/*
1505 		 * We don't bother to fragment if the IP length is greater
1506 		 * than the interface's MTU.  Can this possibly matter?
1507 		 */
1508 		ip = mtod(copym, struct ip *);
1509 		ip->ip_sum = 0;
1510 		ip->ip_sum = in_cksum(copym, hlen);
1511 		if_simloop(ifp, copym, AF_INET, 0);
1512 	}
1513 }
1514