xref: /freebsd/sys/netinet/ip_output.c (revision 6f63e88c0166ed3e5f2805a9e667c7d24d304cf1)
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_mpath.h"
42 #include "opt_ratelimit.h"
43 #include "opt_route.h"
44 #include "opt_rss.h"
45 #include "opt_sctp.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/kernel.h>
50 #include <sys/ktls.h>
51 #include <sys/lock.h>
52 #include <sys/malloc.h>
53 #include <sys/mbuf.h>
54 #include <sys/priv.h>
55 #include <sys/proc.h>
56 #include <sys/protosw.h>
57 #include <sys/rmlock.h>
58 #include <sys/sdt.h>
59 #include <sys/socket.h>
60 #include <sys/socketvar.h>
61 #include <sys/sysctl.h>
62 #include <sys/ucred.h>
63 
64 #include <net/if.h>
65 #include <net/if_var.h>
66 #include <net/if_llatbl.h>
67 #include <net/netisr.h>
68 #include <net/pfil.h>
69 #include <net/route.h>
70 #include <net/route/nhop.h>
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_fib.h>
80 #include <netinet/in_pcb.h>
81 #include <netinet/in_rss.h>
82 #include <netinet/in_var.h>
83 #include <netinet/ip_var.h>
84 #include <netinet/ip_options.h>
85 
86 #include <netinet/udp.h>
87 #include <netinet/udp_var.h>
88 
89 #ifdef SCTP
90 #include <netinet/sctp.h>
91 #include <netinet/sctp_crc32.h>
92 #endif
93 
94 #include <netipsec/ipsec_support.h>
95 
96 #include <machine/in_cksum.h>
97 
98 #include <security/mac/mac_framework.h>
99 
100 #ifdef MBUF_STRESS_TEST
101 static int mbuf_frag_size = 0;
102 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
103 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
104 #endif
105 
106 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
107 
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 #ifdef SCTP
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 #ifdef SCTP
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_ext_pgs.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 #endif
247 #ifdef RATELIMIT
248 	if (inp != NULL && mst == NULL) {
249 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
250 		    (inp->inp_snd_tag != NULL &&
251 		    inp->inp_snd_tag->ifp != ifp))
252 			in_pcboutput_txrtlmt(inp, ifp, m);
253 
254 		if (inp->inp_snd_tag != NULL)
255 			mst = inp->inp_snd_tag;
256 	}
257 #endif
258 	if (stamp_tag && mst != NULL) {
259 		KASSERT(m->m_pkthdr.rcvif == NULL,
260 		    ("trying to add a send tag to a forwarded packet"));
261 		if (mst->ifp != ifp) {
262 			error = EAGAIN;
263 			goto done;
264 		}
265 
266 		/* stamp send tag on mbuf */
267 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
268 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
269 	}
270 
271 	error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
272 
273 done:
274 	/* Check for route change invalidating send tags. */
275 #ifdef KERN_TLS
276 	if (tls != NULL) {
277 		if (error == EAGAIN)
278 			error = ktls_output_eagain(inp, tls);
279 		ktls_free(tls);
280 	}
281 #endif
282 #ifdef RATELIMIT
283 	if (error == EAGAIN)
284 		in_pcboutput_eagain(inp);
285 #endif
286 	return (error);
287 }
288 
289 /* rte<>ro_flags translation */
290 static inline void
291 rt_update_ro_flags(struct route *ro)
292 {
293 	int nh_flags = ro->ro_nh->nh_flags;
294 
295 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
296 
297 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
298 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
299 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
300 }
301 
302 /*
303  * IP output.  The packet in mbuf chain m contains a skeletal IP
304  * header (with len, off, ttl, proto, tos, src, dst).
305  * The mbuf chain containing the packet will be freed.
306  * The mbuf opt, if present, will not be freed.
307  * If route ro is present and has ro_rt initialized, route lookup would be
308  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
309  * then result of route lookup is stored in ro->ro_rt.
310  *
311  * In the IP forwarding case, the packet will arrive with options already
312  * inserted, so must have a NULL opt pointer.
313  */
314 int
315 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
316     struct ip_moptions *imo, struct inpcb *inp)
317 {
318 	struct rm_priotracker in_ifa_tracker;
319 	struct ip *ip;
320 	struct ifnet *ifp = NULL;	/* keep compiler happy */
321 	struct mbuf *m0;
322 	int hlen = sizeof (struct ip);
323 	int mtu;
324 	int error = 0;
325 	struct sockaddr_in *dst, sin;
326 	const struct sockaddr_in *gw;
327 	struct in_ifaddr *ia;
328 	struct in_addr src;
329 	int isbroadcast;
330 	uint16_t ip_len, ip_off;
331 	uint32_t fibnum;
332 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
333 	int no_route_but_check_spd = 0;
334 #endif
335 
336 	M_ASSERTPKTHDR(m);
337 	NET_EPOCH_ASSERT();
338 
339 	if (inp != NULL) {
340 		INP_LOCK_ASSERT(inp);
341 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
342 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
343 			m->m_pkthdr.flowid = inp->inp_flowid;
344 			M_HASHTYPE_SET(m, inp->inp_flowtype);
345 		}
346 #ifdef NUMA
347 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
348 #endif
349 	}
350 
351 	if (opt) {
352 		int len = 0;
353 		m = ip_insertoptions(m, opt, &len);
354 		if (len != 0)
355 			hlen = len; /* ip->ip_hl is updated above */
356 	}
357 	ip = mtod(m, struct ip *);
358 	ip_len = ntohs(ip->ip_len);
359 	ip_off = ntohs(ip->ip_off);
360 
361 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
362 		ip->ip_v = IPVERSION;
363 		ip->ip_hl = hlen >> 2;
364 		ip_fillid(ip);
365 	} else {
366 		/* Header already set, fetch hlen from there */
367 		hlen = ip->ip_hl << 2;
368 	}
369 	if ((flags & IP_FORWARDING) == 0)
370 		IPSTAT_INC(ips_localout);
371 
372 	/*
373 	 * dst/gw handling:
374 	 *
375 	 * gw is readonly but can point either to dst OR rt_gateway,
376 	 * therefore we need restore gw if we're redoing lookup.
377 	 */
378 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
379 	if (ro != NULL)
380 		dst = (struct sockaddr_in *)&ro->ro_dst;
381 	else
382 		dst = &sin;
383 	if (ro == NULL || ro->ro_nh == NULL) {
384 		bzero(dst, sizeof(*dst));
385 		dst->sin_family = AF_INET;
386 		dst->sin_len = sizeof(*dst);
387 		dst->sin_addr = ip->ip_dst;
388 	}
389 	gw = dst;
390 again:
391 	/*
392 	 * Validate route against routing table additions;
393 	 * a better/more specific route might have been added.
394 	 */
395 	if (inp != NULL && ro != NULL && ro->ro_nh != NULL)
396 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
397 	/*
398 	 * If there is a cached route,
399 	 * check that it is to the same destination
400 	 * and is still up.  If not, free it and try again.
401 	 * The address family should also be checked in case of sharing the
402 	 * cache with IPv6.
403 	 * Also check whether routing cache needs invalidation.
404 	 */
405 	if (ro != NULL && ro->ro_nh != NULL &&
406 	    ((!NH_IS_VALID(ro->ro_nh)) || !RT_LINK_IS_UP(ro->ro_nh->nh_ifp) ||
407 	    dst->sin_family != AF_INET ||
408 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
409 		RO_INVALIDATE_CACHE(ro);
410 	ia = NULL;
411 	/*
412 	 * If routing to interface only, short circuit routing lookup.
413 	 * The use of an all-ones broadcast address implies this; an
414 	 * interface is specified by the broadcast address of an interface,
415 	 * or the destination address of a ptp interface.
416 	 */
417 	if (flags & IP_SENDONES) {
418 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
419 						      M_GETFIB(m)))) == NULL &&
420 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
421 						    M_GETFIB(m)))) == NULL) {
422 			IPSTAT_INC(ips_noroute);
423 			error = ENETUNREACH;
424 			goto bad;
425 		}
426 		ip->ip_dst.s_addr = INADDR_BROADCAST;
427 		dst->sin_addr = ip->ip_dst;
428 		ifp = ia->ia_ifp;
429 		mtu = ifp->if_mtu;
430 		ip->ip_ttl = 1;
431 		isbroadcast = 1;
432 		src = IA_SIN(ia)->sin_addr;
433 	} else if (flags & IP_ROUTETOIF) {
434 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
435 						    M_GETFIB(m)))) == NULL &&
436 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
437 						M_GETFIB(m)))) == NULL) {
438 			IPSTAT_INC(ips_noroute);
439 			error = ENETUNREACH;
440 			goto bad;
441 		}
442 		ifp = ia->ia_ifp;
443 		mtu = ifp->if_mtu;
444 		ip->ip_ttl = 1;
445 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
446 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
447 		src = IA_SIN(ia)->sin_addr;
448 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
449 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
450 		/*
451 		 * Bypass the normal routing lookup for multicast
452 		 * packets if the interface is specified.
453 		 */
454 		ifp = imo->imo_multicast_ifp;
455 		mtu = ifp->if_mtu;
456 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
457 		isbroadcast = 0;	/* fool gcc */
458 		/* Interface may have no addresses. */
459 		if (ia != NULL)
460 			src = IA_SIN(ia)->sin_addr;
461 		else
462 			src.s_addr = INADDR_ANY;
463 	} else if (ro != NULL) {
464 		if (ro->ro_nh == NULL) {
465 			/*
466 			 * We want to do any cloning requested by the link
467 			 * layer, as this is probably required in all cases
468 			 * for correct operation (as it is for ARP).
469 			 */
470 			uint32_t flowid;
471 #ifdef RADIX_MPATH
472 			flowid = ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr);
473 #else
474 			flowid = m->m_pkthdr.flowid;
475 #endif
476 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
477 			    NHR_REF, flowid);
478 
479 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh)) ||
480 			    !RT_LINK_IS_UP(ro->ro_nh->nh_ifp)) {
481 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
482 				/*
483 				 * There is no route for this packet, but it is
484 				 * possible that a matching SPD entry exists.
485 				 */
486 				no_route_but_check_spd = 1;
487 				mtu = 0; /* Silence GCC warning. */
488 				goto sendit;
489 #endif
490 				IPSTAT_INC(ips_noroute);
491 				error = EHOSTUNREACH;
492 				goto bad;
493 			}
494 		}
495 		ia = ifatoia(ro->ro_nh->nh_ifa);
496 		ifp = ro->ro_nh->nh_ifp;
497 		counter_u64_add(ro->ro_nh->nh_pksent, 1);
498 		rt_update_ro_flags(ro);
499 		if (ro->ro_nh->nh_flags & NHF_GATEWAY)
500 			gw = &ro->ro_nh->gw4_sa;
501 		if (ro->ro_nh->nh_flags & NHF_HOST)
502 			isbroadcast = (ro->ro_nh->nh_flags & NHF_BROADCAST);
503 		else if (ifp->if_flags & IFF_BROADCAST)
504 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
505 		else
506 			isbroadcast = 0;
507 		if (ro->ro_nh->nh_flags & NHF_HOST)
508 			mtu = ro->ro_nh->nh_mtu;
509 		else
510 			mtu = ifp->if_mtu;
511 		src = IA_SIN(ia)->sin_addr;
512 	} else {
513 		struct nhop4_extended nh;
514 
515 		bzero(&nh, sizeof(nh));
516 		if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh) !=
517 		    0) {
518 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
519 			/*
520 			 * There is no route for this packet, but it is
521 			 * possible that a matching SPD entry exists.
522 			 */
523 			no_route_but_check_spd = 1;
524 			mtu = 0; /* Silence GCC warning. */
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 		dst->sin_addr = nh.nh_addr;
542 		ia = nh.nh_ia;
543 		src = nh.nh_src;
544 		isbroadcast = (((nh.nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
545 		    (NHF_HOST | NHF_BROADCAST)) ||
546 		    ((ifp->if_flags & IFF_BROADCAST) &&
547 		    in_ifaddr_broadcast(dst->sin_addr, ia)));
548 	}
549 
550 	/* Catch a possible divide by zero later. */
551 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
552 	    __func__, mtu, ro,
553 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
554 
555 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
556 		m->m_flags |= M_MCAST;
557 		/*
558 		 * IP destination address is multicast.  Make sure "gw"
559 		 * still points to the address in "ro".  (It may have been
560 		 * changed to point to a gateway address, above.)
561 		 */
562 		gw = dst;
563 		/*
564 		 * See if the caller provided any multicast options
565 		 */
566 		if (imo != NULL) {
567 			ip->ip_ttl = imo->imo_multicast_ttl;
568 			if (imo->imo_multicast_vif != -1)
569 				ip->ip_src.s_addr =
570 				    ip_mcast_src ?
571 				    ip_mcast_src(imo->imo_multicast_vif) :
572 				    INADDR_ANY;
573 		} else
574 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
575 		/*
576 		 * Confirm that the outgoing interface supports multicast.
577 		 */
578 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
579 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
580 				IPSTAT_INC(ips_noroute);
581 				error = ENETUNREACH;
582 				goto bad;
583 			}
584 		}
585 		/*
586 		 * If source address not specified yet, use address
587 		 * of outgoing interface.
588 		 */
589 		if (ip->ip_src.s_addr == INADDR_ANY)
590 			ip->ip_src = src;
591 
592 		if ((imo == NULL && in_mcast_loop) ||
593 		    (imo && imo->imo_multicast_loop)) {
594 			/*
595 			 * Loop back multicast datagram if not expressly
596 			 * forbidden to do so, even if we are not a member
597 			 * of the group; ip_input() will filter it later,
598 			 * thus deferring a hash lookup and mutex acquisition
599 			 * at the expense of a cheap copy using m_copym().
600 			 */
601 			ip_mloopback(ifp, m, hlen);
602 		} else {
603 			/*
604 			 * If we are acting as a multicast router, perform
605 			 * multicast forwarding as if the packet had just
606 			 * arrived on the interface to which we are about
607 			 * to send.  The multicast forwarding function
608 			 * recursively calls this function, using the
609 			 * IP_FORWARDING flag to prevent infinite recursion.
610 			 *
611 			 * Multicasts that are looped back by ip_mloopback(),
612 			 * above, will be forwarded by the ip_input() routine,
613 			 * if necessary.
614 			 */
615 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
616 				/*
617 				 * If rsvp daemon is not running, do not
618 				 * set ip_moptions. This ensures that the packet
619 				 * is multicast and not just sent down one link
620 				 * as prescribed by rsvpd.
621 				 */
622 				if (!V_rsvp_on)
623 					imo = NULL;
624 				if (ip_mforward &&
625 				    ip_mforward(ip, ifp, m, imo) != 0) {
626 					m_freem(m);
627 					goto done;
628 				}
629 			}
630 		}
631 
632 		/*
633 		 * Multicasts with a time-to-live of zero may be looped-
634 		 * back, above, but must not be transmitted on a network.
635 		 * Also, multicasts addressed to the loopback interface
636 		 * are not sent -- the above call to ip_mloopback() will
637 		 * loop back a copy. ip_input() will drop the copy if
638 		 * this host does not belong to the destination group on
639 		 * the loopback interface.
640 		 */
641 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
642 			m_freem(m);
643 			goto done;
644 		}
645 
646 		goto sendit;
647 	}
648 
649 	/*
650 	 * If the source address is not specified yet, use the address
651 	 * of the outoing interface.
652 	 */
653 	if (ip->ip_src.s_addr == INADDR_ANY)
654 		ip->ip_src = src;
655 
656 	/*
657 	 * Look for broadcast address and
658 	 * verify user is allowed to send
659 	 * such a packet.
660 	 */
661 	if (isbroadcast) {
662 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
663 			error = EADDRNOTAVAIL;
664 			goto bad;
665 		}
666 		if ((flags & IP_ALLOWBROADCAST) == 0) {
667 			error = EACCES;
668 			goto bad;
669 		}
670 		/* don't allow broadcast messages to be fragmented */
671 		if (ip_len > mtu) {
672 			error = EMSGSIZE;
673 			goto bad;
674 		}
675 		m->m_flags |= M_BCAST;
676 	} else {
677 		m->m_flags &= ~M_BCAST;
678 	}
679 
680 sendit:
681 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
682 	if (IPSEC_ENABLED(ipv4)) {
683 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
684 			if (error == EINPROGRESS)
685 				error = 0;
686 			goto done;
687 		}
688 	}
689 	/*
690 	 * Check if there was a route for this packet; return error if not.
691 	 */
692 	if (no_route_but_check_spd) {
693 		IPSTAT_INC(ips_noroute);
694 		error = EHOSTUNREACH;
695 		goto bad;
696 	}
697 	/* Update variables that are affected by ipsec4_output(). */
698 	ip = mtod(m, struct ip *);
699 	hlen = ip->ip_hl << 2;
700 #endif /* IPSEC */
701 
702 	/* Jump over all PFIL processing if hooks are not active. */
703 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
704 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
705 		    &error)) {
706 		case 1: /* Finished */
707 			goto done;
708 
709 		case 0: /* Continue normally */
710 			ip = mtod(m, struct ip *);
711 			break;
712 
713 		case -1: /* Need to try again */
714 			/* Reset everything for a new round */
715 			if (ro != NULL) {
716 				RO_NHFREE(ro);
717 				ro->ro_prepend = NULL;
718 			}
719 			gw = dst;
720 			ip = mtod(m, struct ip *);
721 			goto again;
722 
723 		}
724 	}
725 
726 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
727 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
728 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
729 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
730 			IPSTAT_INC(ips_badaddr);
731 			error = EADDRNOTAVAIL;
732 			goto bad;
733 		}
734 	}
735 
736 	m->m_pkthdr.csum_flags |= CSUM_IP;
737 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
738 		m = mb_unmapped_to_ext(m);
739 		if (m == NULL) {
740 			IPSTAT_INC(ips_odropped);
741 			error = ENOBUFS;
742 			goto bad;
743 		}
744 		in_delayed_cksum(m);
745 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
746 	} else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
747 		m = mb_unmapped_to_ext(m);
748 		if (m == NULL) {
749 			IPSTAT_INC(ips_odropped);
750 			error = ENOBUFS;
751 			goto bad;
752 		}
753 	}
754 #ifdef SCTP
755 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
756 		m = mb_unmapped_to_ext(m);
757 		if (m == NULL) {
758 			IPSTAT_INC(ips_odropped);
759 			error = ENOBUFS;
760 			goto bad;
761 		}
762 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
763 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
764 	}
765 #endif
766 
767 	/*
768 	 * If small enough for interface, or the interface will take
769 	 * care of the fragmentation for us, we can just send directly.
770 	 */
771 	if (ip_len <= mtu ||
772 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
773 		ip->ip_sum = 0;
774 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
775 			ip->ip_sum = in_cksum(m, hlen);
776 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
777 		}
778 
779 		/*
780 		 * Record statistics for this interface address.
781 		 * With CSUM_TSO the byte/packet count will be slightly
782 		 * incorrect because we count the IP+TCP headers only
783 		 * once instead of for every generated packet.
784 		 */
785 		if (!(flags & IP_FORWARDING) && ia) {
786 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
787 				counter_u64_add(ia->ia_ifa.ifa_opackets,
788 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
789 			else
790 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
791 
792 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
793 		}
794 #ifdef MBUF_STRESS_TEST
795 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
796 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
797 #endif
798 		/*
799 		 * Reset layer specific mbuf flags
800 		 * to avoid confusing lower layers.
801 		 */
802 		m_clrprotoflags(m);
803 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
804 		error = ip_output_send(inp, ifp, m, gw, ro,
805 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
806 		goto done;
807 	}
808 
809 	/* Balk when DF bit is set or the interface didn't support TSO. */
810 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
811 		error = EMSGSIZE;
812 		IPSTAT_INC(ips_cantfrag);
813 		goto bad;
814 	}
815 
816 	/*
817 	 * Too large for interface; fragment if possible. If successful,
818 	 * on return, m will point to a list of packets to be sent.
819 	 */
820 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
821 	if (error)
822 		goto bad;
823 	for (; m; m = m0) {
824 		m0 = m->m_nextpkt;
825 		m->m_nextpkt = 0;
826 		if (error == 0) {
827 			/* Record statistics for this interface address. */
828 			if (ia != NULL) {
829 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
830 				counter_u64_add(ia->ia_ifa.ifa_obytes,
831 				    m->m_pkthdr.len);
832 			}
833 			/*
834 			 * Reset layer specific mbuf flags
835 			 * to avoid confusing upper layers.
836 			 */
837 			m_clrprotoflags(m);
838 
839 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
840 			    mtod(m, struct ip *), NULL);
841 			error = ip_output_send(inp, ifp, m, gw, ro, true);
842 		} else
843 			m_freem(m);
844 	}
845 
846 	if (error == 0)
847 		IPSTAT_INC(ips_fragmented);
848 
849 done:
850 	return (error);
851  bad:
852 	m_freem(m);
853 	goto done;
854 }
855 
856 /*
857  * Create a chain of fragments which fit the given mtu. m_frag points to the
858  * mbuf to be fragmented; on return it points to the chain with the fragments.
859  * Return 0 if no error. If error, m_frag may contain a partially built
860  * chain of fragments that should be freed by the caller.
861  *
862  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
863  */
864 int
865 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
866     u_long if_hwassist_flags)
867 {
868 	int error = 0;
869 	int hlen = ip->ip_hl << 2;
870 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
871 	int off;
872 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
873 	int firstlen;
874 	struct mbuf **mnext;
875 	int nfrags;
876 	uint16_t ip_len, ip_off;
877 
878 	ip_len = ntohs(ip->ip_len);
879 	ip_off = ntohs(ip->ip_off);
880 
881 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
882 		IPSTAT_INC(ips_cantfrag);
883 		return EMSGSIZE;
884 	}
885 
886 	/*
887 	 * Must be able to put at least 8 bytes per fragment.
888 	 */
889 	if (len < 8)
890 		return EMSGSIZE;
891 
892 	/*
893 	 * If the interface will not calculate checksums on
894 	 * fragmented packets, then do it here.
895 	 */
896 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
897 		m0 = mb_unmapped_to_ext(m0);
898 		if (m0 == NULL) {
899 			error = ENOBUFS;
900 			IPSTAT_INC(ips_odropped);
901 			goto done;
902 		}
903 		in_delayed_cksum(m0);
904 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
905 	}
906 #ifdef SCTP
907 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
908 		m0 = mb_unmapped_to_ext(m0);
909 		if (m0 == NULL) {
910 			error = ENOBUFS;
911 			IPSTAT_INC(ips_odropped);
912 			goto done;
913 		}
914 		sctp_delayed_cksum(m0, hlen);
915 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
916 	}
917 #endif
918 	if (len > PAGE_SIZE) {
919 		/*
920 		 * Fragment large datagrams such that each segment
921 		 * contains a multiple of PAGE_SIZE amount of data,
922 		 * plus headers. This enables a receiver to perform
923 		 * page-flipping zero-copy optimizations.
924 		 *
925 		 * XXX When does this help given that sender and receiver
926 		 * could have different page sizes, and also mtu could
927 		 * be less than the receiver's page size ?
928 		 */
929 		int newlen;
930 
931 		off = MIN(mtu, m0->m_pkthdr.len);
932 
933 		/*
934 		 * firstlen (off - hlen) must be aligned on an
935 		 * 8-byte boundary
936 		 */
937 		if (off < hlen)
938 			goto smart_frag_failure;
939 		off = ((off - hlen) & ~7) + hlen;
940 		newlen = (~PAGE_MASK) & mtu;
941 		if ((newlen + sizeof (struct ip)) > mtu) {
942 			/* we failed, go back the default */
943 smart_frag_failure:
944 			newlen = len;
945 			off = hlen + len;
946 		}
947 		len = newlen;
948 
949 	} else {
950 		off = hlen + len;
951 	}
952 
953 	firstlen = off - hlen;
954 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
955 
956 	/*
957 	 * Loop through length of segment after first fragment,
958 	 * make new header and copy data of each part and link onto chain.
959 	 * Here, m0 is the original packet, m is the fragment being created.
960 	 * The fragments are linked off the m_nextpkt of the original
961 	 * packet, which after processing serves as the first fragment.
962 	 */
963 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
964 		struct ip *mhip;	/* ip header on the fragment */
965 		struct mbuf *m;
966 		int mhlen = sizeof (struct ip);
967 
968 		m = m_gethdr(M_NOWAIT, MT_DATA);
969 		if (m == NULL) {
970 			error = ENOBUFS;
971 			IPSTAT_INC(ips_odropped);
972 			goto done;
973 		}
974 		/*
975 		 * Make sure the complete packet header gets copied
976 		 * from the originating mbuf to the newly created
977 		 * mbuf. This also ensures that existing firewall
978 		 * classification(s), VLAN tags and so on get copied
979 		 * to the resulting fragmented packet(s):
980 		 */
981 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
982 			m_free(m);
983 			error = ENOBUFS;
984 			IPSTAT_INC(ips_odropped);
985 			goto done;
986 		}
987 		/*
988 		 * In the first mbuf, leave room for the link header, then
989 		 * copy the original IP header including options. The payload
990 		 * goes into an additional mbuf chain returned by m_copym().
991 		 */
992 		m->m_data += max_linkhdr;
993 		mhip = mtod(m, struct ip *);
994 		*mhip = *ip;
995 		if (hlen > sizeof (struct ip)) {
996 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
997 			mhip->ip_v = IPVERSION;
998 			mhip->ip_hl = mhlen >> 2;
999 		}
1000 		m->m_len = mhlen;
1001 		/* XXX do we need to add ip_off below ? */
1002 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1003 		if (off + len >= ip_len)
1004 			len = ip_len - off;
1005 		else
1006 			mhip->ip_off |= IP_MF;
1007 		mhip->ip_len = htons((u_short)(len + mhlen));
1008 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1009 		if (m->m_next == NULL) {	/* copy failed */
1010 			m_free(m);
1011 			error = ENOBUFS;	/* ??? */
1012 			IPSTAT_INC(ips_odropped);
1013 			goto done;
1014 		}
1015 		m->m_pkthdr.len = mhlen + len;
1016 #ifdef MAC
1017 		mac_netinet_fragment(m0, m);
1018 #endif
1019 		mhip->ip_off = htons(mhip->ip_off);
1020 		mhip->ip_sum = 0;
1021 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1022 			mhip->ip_sum = in_cksum(m, mhlen);
1023 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1024 		}
1025 		*mnext = m;
1026 		mnext = &m->m_nextpkt;
1027 	}
1028 	IPSTAT_ADD(ips_ofragments, nfrags);
1029 
1030 	/*
1031 	 * Update first fragment by trimming what's been copied out
1032 	 * and updating header.
1033 	 */
1034 	m_adj(m0, hlen + firstlen - ip_len);
1035 	m0->m_pkthdr.len = hlen + firstlen;
1036 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1037 	ip->ip_off = htons(ip_off | IP_MF);
1038 	ip->ip_sum = 0;
1039 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1040 		ip->ip_sum = in_cksum(m0, hlen);
1041 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1042 	}
1043 
1044 done:
1045 	*m_frag = m0;
1046 	return error;
1047 }
1048 
1049 void
1050 in_delayed_cksum(struct mbuf *m)
1051 {
1052 	struct ip *ip;
1053 	struct udphdr *uh;
1054 	uint16_t cklen, csum, offset;
1055 
1056 	ip = mtod(m, struct ip *);
1057 	offset = ip->ip_hl << 2 ;
1058 
1059 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1060 		/* if udp header is not in the first mbuf copy udplen */
1061 		if (offset + sizeof(struct udphdr) > m->m_len) {
1062 			m_copydata(m, offset + offsetof(struct udphdr,
1063 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1064 			cklen = ntohs(cklen);
1065 		} else {
1066 			uh = (struct udphdr *)mtodo(m, offset);
1067 			cklen = ntohs(uh->uh_ulen);
1068 		}
1069 		csum = in_cksum_skip(m, cklen + offset, offset);
1070 		if (csum == 0)
1071 			csum = 0xffff;
1072 	} else {
1073 		cklen = ntohs(ip->ip_len);
1074 		csum = in_cksum_skip(m, cklen, offset);
1075 	}
1076 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1077 
1078 	if (offset + sizeof(csum) > m->m_len)
1079 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1080 	else
1081 		*(u_short *)mtodo(m, offset) = csum;
1082 }
1083 
1084 /*
1085  * IP socket option processing.
1086  */
1087 int
1088 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1089 {
1090 	struct inpcb *inp = sotoinpcb(so);
1091 	int	error, optval;
1092 #ifdef	RSS
1093 	uint32_t rss_bucket;
1094 	int retval;
1095 #endif
1096 
1097 	error = optval = 0;
1098 	if (sopt->sopt_level != IPPROTO_IP) {
1099 		error = EINVAL;
1100 
1101 		if (sopt->sopt_level == SOL_SOCKET &&
1102 		    sopt->sopt_dir == SOPT_SET) {
1103 			switch (sopt->sopt_name) {
1104 			case SO_REUSEADDR:
1105 				INP_WLOCK(inp);
1106 				if ((so->so_options & SO_REUSEADDR) != 0)
1107 					inp->inp_flags2 |= INP_REUSEADDR;
1108 				else
1109 					inp->inp_flags2 &= ~INP_REUSEADDR;
1110 				INP_WUNLOCK(inp);
1111 				error = 0;
1112 				break;
1113 			case SO_REUSEPORT:
1114 				INP_WLOCK(inp);
1115 				if ((so->so_options & SO_REUSEPORT) != 0)
1116 					inp->inp_flags2 |= INP_REUSEPORT;
1117 				else
1118 					inp->inp_flags2 &= ~INP_REUSEPORT;
1119 				INP_WUNLOCK(inp);
1120 				error = 0;
1121 				break;
1122 			case SO_REUSEPORT_LB:
1123 				INP_WLOCK(inp);
1124 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1125 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1126 				else
1127 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1128 				INP_WUNLOCK(inp);
1129 				error = 0;
1130 				break;
1131 			case SO_SETFIB:
1132 				INP_WLOCK(inp);
1133 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1134 				INP_WUNLOCK(inp);
1135 				error = 0;
1136 				break;
1137 			case SO_MAX_PACING_RATE:
1138 #ifdef RATELIMIT
1139 				INP_WLOCK(inp);
1140 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1141 				INP_WUNLOCK(inp);
1142 				error = 0;
1143 #else
1144 				error = EOPNOTSUPP;
1145 #endif
1146 				break;
1147 			default:
1148 				break;
1149 			}
1150 		}
1151 		return (error);
1152 	}
1153 
1154 	switch (sopt->sopt_dir) {
1155 	case SOPT_SET:
1156 		switch (sopt->sopt_name) {
1157 		case IP_OPTIONS:
1158 #ifdef notyet
1159 		case IP_RETOPTS:
1160 #endif
1161 		{
1162 			struct mbuf *m;
1163 			if (sopt->sopt_valsize > MLEN) {
1164 				error = EMSGSIZE;
1165 				break;
1166 			}
1167 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1168 			if (m == NULL) {
1169 				error = ENOBUFS;
1170 				break;
1171 			}
1172 			m->m_len = sopt->sopt_valsize;
1173 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1174 					    m->m_len);
1175 			if (error) {
1176 				m_free(m);
1177 				break;
1178 			}
1179 			INP_WLOCK(inp);
1180 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1181 			INP_WUNLOCK(inp);
1182 			return (error);
1183 		}
1184 
1185 		case IP_BINDANY:
1186 			if (sopt->sopt_td != NULL) {
1187 				error = priv_check(sopt->sopt_td,
1188 				    PRIV_NETINET_BINDANY);
1189 				if (error)
1190 					break;
1191 			}
1192 			/* FALLTHROUGH */
1193 		case IP_BINDMULTI:
1194 #ifdef	RSS
1195 		case IP_RSS_LISTEN_BUCKET:
1196 #endif
1197 		case IP_TOS:
1198 		case IP_TTL:
1199 		case IP_MINTTL:
1200 		case IP_RECVOPTS:
1201 		case IP_RECVRETOPTS:
1202 		case IP_ORIGDSTADDR:
1203 		case IP_RECVDSTADDR:
1204 		case IP_RECVTTL:
1205 		case IP_RECVIF:
1206 		case IP_ONESBCAST:
1207 		case IP_DONTFRAG:
1208 		case IP_RECVTOS:
1209 		case IP_RECVFLOWID:
1210 #ifdef	RSS
1211 		case IP_RECVRSSBUCKETID:
1212 #endif
1213 			error = sooptcopyin(sopt, &optval, sizeof optval,
1214 					    sizeof optval);
1215 			if (error)
1216 				break;
1217 
1218 			switch (sopt->sopt_name) {
1219 			case IP_TOS:
1220 				inp->inp_ip_tos = optval;
1221 				break;
1222 
1223 			case IP_TTL:
1224 				inp->inp_ip_ttl = optval;
1225 				break;
1226 
1227 			case IP_MINTTL:
1228 				if (optval >= 0 && optval <= MAXTTL)
1229 					inp->inp_ip_minttl = optval;
1230 				else
1231 					error = EINVAL;
1232 				break;
1233 
1234 #define	OPTSET(bit) do {						\
1235 	INP_WLOCK(inp);							\
1236 	if (optval)							\
1237 		inp->inp_flags |= bit;					\
1238 	else								\
1239 		inp->inp_flags &= ~bit;					\
1240 	INP_WUNLOCK(inp);						\
1241 } while (0)
1242 
1243 #define	OPTSET2(bit, val) do {						\
1244 	INP_WLOCK(inp);							\
1245 	if (val)							\
1246 		inp->inp_flags2 |= bit;					\
1247 	else								\
1248 		inp->inp_flags2 &= ~bit;				\
1249 	INP_WUNLOCK(inp);						\
1250 } while (0)
1251 
1252 			case IP_RECVOPTS:
1253 				OPTSET(INP_RECVOPTS);
1254 				break;
1255 
1256 			case IP_RECVRETOPTS:
1257 				OPTSET(INP_RECVRETOPTS);
1258 				break;
1259 
1260 			case IP_RECVDSTADDR:
1261 				OPTSET(INP_RECVDSTADDR);
1262 				break;
1263 
1264 			case IP_ORIGDSTADDR:
1265 				OPTSET2(INP_ORIGDSTADDR, optval);
1266 				break;
1267 
1268 			case IP_RECVTTL:
1269 				OPTSET(INP_RECVTTL);
1270 				break;
1271 
1272 			case IP_RECVIF:
1273 				OPTSET(INP_RECVIF);
1274 				break;
1275 
1276 			case IP_ONESBCAST:
1277 				OPTSET(INP_ONESBCAST);
1278 				break;
1279 			case IP_DONTFRAG:
1280 				OPTSET(INP_DONTFRAG);
1281 				break;
1282 			case IP_BINDANY:
1283 				OPTSET(INP_BINDANY);
1284 				break;
1285 			case IP_RECVTOS:
1286 				OPTSET(INP_RECVTOS);
1287 				break;
1288 			case IP_BINDMULTI:
1289 				OPTSET2(INP_BINDMULTI, optval);
1290 				break;
1291 			case IP_RECVFLOWID:
1292 				OPTSET2(INP_RECVFLOWID, optval);
1293 				break;
1294 #ifdef	RSS
1295 			case IP_RSS_LISTEN_BUCKET:
1296 				if ((optval >= 0) &&
1297 				    (optval < rss_getnumbuckets())) {
1298 					inp->inp_rss_listen_bucket = optval;
1299 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1300 				} else {
1301 					error = EINVAL;
1302 				}
1303 				break;
1304 			case IP_RECVRSSBUCKETID:
1305 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1306 				break;
1307 #endif
1308 			}
1309 			break;
1310 #undef OPTSET
1311 #undef OPTSET2
1312 
1313 		/*
1314 		 * Multicast socket options are processed by the in_mcast
1315 		 * module.
1316 		 */
1317 		case IP_MULTICAST_IF:
1318 		case IP_MULTICAST_VIF:
1319 		case IP_MULTICAST_TTL:
1320 		case IP_MULTICAST_LOOP:
1321 		case IP_ADD_MEMBERSHIP:
1322 		case IP_DROP_MEMBERSHIP:
1323 		case IP_ADD_SOURCE_MEMBERSHIP:
1324 		case IP_DROP_SOURCE_MEMBERSHIP:
1325 		case IP_BLOCK_SOURCE:
1326 		case IP_UNBLOCK_SOURCE:
1327 		case IP_MSFILTER:
1328 		case MCAST_JOIN_GROUP:
1329 		case MCAST_LEAVE_GROUP:
1330 		case MCAST_JOIN_SOURCE_GROUP:
1331 		case MCAST_LEAVE_SOURCE_GROUP:
1332 		case MCAST_BLOCK_SOURCE:
1333 		case MCAST_UNBLOCK_SOURCE:
1334 			error = inp_setmoptions(inp, sopt);
1335 			break;
1336 
1337 		case IP_PORTRANGE:
1338 			error = sooptcopyin(sopt, &optval, sizeof optval,
1339 					    sizeof optval);
1340 			if (error)
1341 				break;
1342 
1343 			INP_WLOCK(inp);
1344 			switch (optval) {
1345 			case IP_PORTRANGE_DEFAULT:
1346 				inp->inp_flags &= ~(INP_LOWPORT);
1347 				inp->inp_flags &= ~(INP_HIGHPORT);
1348 				break;
1349 
1350 			case IP_PORTRANGE_HIGH:
1351 				inp->inp_flags &= ~(INP_LOWPORT);
1352 				inp->inp_flags |= INP_HIGHPORT;
1353 				break;
1354 
1355 			case IP_PORTRANGE_LOW:
1356 				inp->inp_flags &= ~(INP_HIGHPORT);
1357 				inp->inp_flags |= INP_LOWPORT;
1358 				break;
1359 
1360 			default:
1361 				error = EINVAL;
1362 				break;
1363 			}
1364 			INP_WUNLOCK(inp);
1365 			break;
1366 
1367 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1368 		case IP_IPSEC_POLICY:
1369 			if (IPSEC_ENABLED(ipv4)) {
1370 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1371 				break;
1372 			}
1373 			/* FALLTHROUGH */
1374 #endif /* IPSEC */
1375 
1376 		default:
1377 			error = ENOPROTOOPT;
1378 			break;
1379 		}
1380 		break;
1381 
1382 	case SOPT_GET:
1383 		switch (sopt->sopt_name) {
1384 		case IP_OPTIONS:
1385 		case IP_RETOPTS:
1386 			INP_RLOCK(inp);
1387 			if (inp->inp_options) {
1388 				struct mbuf *options;
1389 
1390 				options = m_copym(inp->inp_options, 0,
1391 				    M_COPYALL, M_NOWAIT);
1392 				INP_RUNLOCK(inp);
1393 				if (options != NULL) {
1394 					error = sooptcopyout(sopt,
1395 							     mtod(options, char *),
1396 							     options->m_len);
1397 					m_freem(options);
1398 				} else
1399 					error = ENOMEM;
1400 			} else {
1401 				INP_RUNLOCK(inp);
1402 				sopt->sopt_valsize = 0;
1403 			}
1404 			break;
1405 
1406 		case IP_TOS:
1407 		case IP_TTL:
1408 		case IP_MINTTL:
1409 		case IP_RECVOPTS:
1410 		case IP_RECVRETOPTS:
1411 		case IP_ORIGDSTADDR:
1412 		case IP_RECVDSTADDR:
1413 		case IP_RECVTTL:
1414 		case IP_RECVIF:
1415 		case IP_PORTRANGE:
1416 		case IP_ONESBCAST:
1417 		case IP_DONTFRAG:
1418 		case IP_BINDANY:
1419 		case IP_RECVTOS:
1420 		case IP_BINDMULTI:
1421 		case IP_FLOWID:
1422 		case IP_FLOWTYPE:
1423 		case IP_RECVFLOWID:
1424 #ifdef	RSS
1425 		case IP_RSSBUCKETID:
1426 		case IP_RECVRSSBUCKETID:
1427 #endif
1428 			switch (sopt->sopt_name) {
1429 
1430 			case IP_TOS:
1431 				optval = inp->inp_ip_tos;
1432 				break;
1433 
1434 			case IP_TTL:
1435 				optval = inp->inp_ip_ttl;
1436 				break;
1437 
1438 			case IP_MINTTL:
1439 				optval = inp->inp_ip_minttl;
1440 				break;
1441 
1442 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1443 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1444 
1445 			case IP_RECVOPTS:
1446 				optval = OPTBIT(INP_RECVOPTS);
1447 				break;
1448 
1449 			case IP_RECVRETOPTS:
1450 				optval = OPTBIT(INP_RECVRETOPTS);
1451 				break;
1452 
1453 			case IP_RECVDSTADDR:
1454 				optval = OPTBIT(INP_RECVDSTADDR);
1455 				break;
1456 
1457 			case IP_ORIGDSTADDR:
1458 				optval = OPTBIT2(INP_ORIGDSTADDR);
1459 				break;
1460 
1461 			case IP_RECVTTL:
1462 				optval = OPTBIT(INP_RECVTTL);
1463 				break;
1464 
1465 			case IP_RECVIF:
1466 				optval = OPTBIT(INP_RECVIF);
1467 				break;
1468 
1469 			case IP_PORTRANGE:
1470 				if (inp->inp_flags & INP_HIGHPORT)
1471 					optval = IP_PORTRANGE_HIGH;
1472 				else if (inp->inp_flags & INP_LOWPORT)
1473 					optval = IP_PORTRANGE_LOW;
1474 				else
1475 					optval = 0;
1476 				break;
1477 
1478 			case IP_ONESBCAST:
1479 				optval = OPTBIT(INP_ONESBCAST);
1480 				break;
1481 			case IP_DONTFRAG:
1482 				optval = OPTBIT(INP_DONTFRAG);
1483 				break;
1484 			case IP_BINDANY:
1485 				optval = OPTBIT(INP_BINDANY);
1486 				break;
1487 			case IP_RECVTOS:
1488 				optval = OPTBIT(INP_RECVTOS);
1489 				break;
1490 			case IP_FLOWID:
1491 				optval = inp->inp_flowid;
1492 				break;
1493 			case IP_FLOWTYPE:
1494 				optval = inp->inp_flowtype;
1495 				break;
1496 			case IP_RECVFLOWID:
1497 				optval = OPTBIT2(INP_RECVFLOWID);
1498 				break;
1499 #ifdef	RSS
1500 			case IP_RSSBUCKETID:
1501 				retval = rss_hash2bucket(inp->inp_flowid,
1502 				    inp->inp_flowtype,
1503 				    &rss_bucket);
1504 				if (retval == 0)
1505 					optval = rss_bucket;
1506 				else
1507 					error = EINVAL;
1508 				break;
1509 			case IP_RECVRSSBUCKETID:
1510 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1511 				break;
1512 #endif
1513 			case IP_BINDMULTI:
1514 				optval = OPTBIT2(INP_BINDMULTI);
1515 				break;
1516 			}
1517 			error = sooptcopyout(sopt, &optval, sizeof optval);
1518 			break;
1519 
1520 		/*
1521 		 * Multicast socket options are processed by the in_mcast
1522 		 * module.
1523 		 */
1524 		case IP_MULTICAST_IF:
1525 		case IP_MULTICAST_VIF:
1526 		case IP_MULTICAST_TTL:
1527 		case IP_MULTICAST_LOOP:
1528 		case IP_MSFILTER:
1529 			error = inp_getmoptions(inp, sopt);
1530 			break;
1531 
1532 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1533 		case IP_IPSEC_POLICY:
1534 			if (IPSEC_ENABLED(ipv4)) {
1535 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1536 				break;
1537 			}
1538 			/* FALLTHROUGH */
1539 #endif /* IPSEC */
1540 
1541 		default:
1542 			error = ENOPROTOOPT;
1543 			break;
1544 		}
1545 		break;
1546 	}
1547 	return (error);
1548 }
1549 
1550 /*
1551  * Routine called from ip_output() to loop back a copy of an IP multicast
1552  * packet to the input queue of a specified interface.  Note that this
1553  * calls the output routine of the loopback "driver", but with an interface
1554  * pointer that might NOT be a loopback interface -- evil, but easier than
1555  * replicating that code here.
1556  */
1557 static void
1558 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1559 {
1560 	struct ip *ip;
1561 	struct mbuf *copym;
1562 
1563 	/*
1564 	 * Make a deep copy of the packet because we're going to
1565 	 * modify the pack in order to generate checksums.
1566 	 */
1567 	copym = m_dup(m, M_NOWAIT);
1568 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1569 		copym = m_pullup(copym, hlen);
1570 	if (copym != NULL) {
1571 		/* If needed, compute the checksum and mark it as valid. */
1572 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1573 			in_delayed_cksum(copym);
1574 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1575 			copym->m_pkthdr.csum_flags |=
1576 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1577 			copym->m_pkthdr.csum_data = 0xffff;
1578 		}
1579 		/*
1580 		 * We don't bother to fragment if the IP length is greater
1581 		 * than the interface's MTU.  Can this possibly matter?
1582 		 */
1583 		ip = mtod(copym, struct ip *);
1584 		ip->ip_sum = 0;
1585 		ip->ip_sum = in_cksum(copym, hlen);
1586 		if_simloop(ifp, copym, AF_INET, 0);
1587 	}
1588 }
1589