xref: /freebsd/sys/netinet/ip_output.c (revision 21b492ed51aa6ff8008a8aa83333b1de30288a15)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1990, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_inet.h"
38 #include "opt_ipsec.h"
39 #include "opt_kern_tls.h"
40 #include "opt_mbuf_stress_test.h"
41 #include "opt_ratelimit.h"
42 #include "opt_route.h"
43 #include "opt_rss.h"
44 #include "opt_sctp.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
49 #include <sys/ktls.h>
50 #include <sys/lock.h>
51 #include <sys/malloc.h>
52 #include <sys/mbuf.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/protosw.h>
56 #include <sys/rmlock.h>
57 #include <sys/sdt.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
60 #include <sys/sysctl.h>
61 #include <sys/ucred.h>
62 
63 #include <net/if.h>
64 #include <net/if_var.h>
65 #include <net/if_vlan_var.h>
66 #include <net/if_llatbl.h>
67 #include <net/ethernet.h>
68 #include <net/netisr.h>
69 #include <net/pfil.h>
70 #include <net/route.h>
71 #include <net/route/nhop.h>
72 #include <net/rss_config.h>
73 #include <net/vnet.h>
74 
75 #include <netinet/in.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_kdtrace.h>
78 #include <netinet/in_systm.h>
79 #include <netinet/ip.h>
80 #include <netinet/in_fib.h>
81 #include <netinet/in_pcb.h>
82 #include <netinet/in_rss.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip_var.h>
85 #include <netinet/ip_options.h>
86 #include <netinet/ip_mroute.h>
87 
88 #include <netinet/udp.h>
89 #include <netinet/udp_var.h>
90 
91 #if defined(SCTP) || defined(SCTP_SUPPORT)
92 #include <netinet/sctp.h>
93 #include <netinet/sctp_crc32.h>
94 #endif
95 
96 #include <netipsec/ipsec_support.h>
97 
98 #include <machine/in_cksum.h>
99 
100 #include <security/mac/mac_framework.h>
101 
102 #ifdef MBUF_STRESS_TEST
103 static int mbuf_frag_size = 0;
104 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
105 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
106 #endif
107 
108 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
109 
110 extern int in_mcast_loop;
111 extern	struct protosw inetsw[];
112 
113 static inline int
114 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
115     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
116 {
117 	struct m_tag *fwd_tag = NULL;
118 	struct mbuf *m;
119 	struct in_addr odst;
120 	struct ip *ip;
121 	int pflags = PFIL_OUT;
122 
123 	if (flags & IP_FORWARDING)
124 		pflags |= PFIL_FWD;
125 
126 	m = *mp;
127 	ip = mtod(m, struct ip *);
128 
129 	/* Run through list of hooks for output packets. */
130 	odst.s_addr = ip->ip_dst.s_addr;
131 	switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
132 	case PFIL_DROPPED:
133 		*error = EACCES;
134 		/* FALLTHROUGH */
135 	case PFIL_CONSUMED:
136 		return 1; /* Finished */
137 	case PFIL_PASS:
138 		*error = 0;
139 	}
140 	m = *mp;
141 	ip = mtod(m, struct ip *);
142 
143 	/* See if destination IP address was changed by packet filter. */
144 	if (odst.s_addr != ip->ip_dst.s_addr) {
145 		m->m_flags |= M_SKIP_FIREWALL;
146 		/* If destination is now ourself drop to ip_input(). */
147 		if (in_localip(ip->ip_dst)) {
148 			m->m_flags |= M_FASTFWD_OURS;
149 			if (m->m_pkthdr.rcvif == NULL)
150 				m->m_pkthdr.rcvif = V_loif;
151 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
152 				m->m_pkthdr.csum_flags |=
153 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
154 				m->m_pkthdr.csum_data = 0xffff;
155 			}
156 			m->m_pkthdr.csum_flags |=
157 				CSUM_IP_CHECKED | CSUM_IP_VALID;
158 #if defined(SCTP) || defined(SCTP_SUPPORT)
159 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
160 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
161 #endif
162 			*error = netisr_queue(NETISR_IP, m);
163 			return 1; /* Finished */
164 		}
165 
166 		bzero(dst, sizeof(*dst));
167 		dst->sin_family = AF_INET;
168 		dst->sin_len = sizeof(*dst);
169 		dst->sin_addr = ip->ip_dst;
170 
171 		return -1; /* Reloop */
172 	}
173 	/* See if fib was changed by packet filter. */
174 	if ((*fibnum) != M_GETFIB(m)) {
175 		m->m_flags |= M_SKIP_FIREWALL;
176 		*fibnum = M_GETFIB(m);
177 		return -1; /* Reloop for FIB change */
178 	}
179 
180 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
181 	if (m->m_flags & M_FASTFWD_OURS) {
182 		if (m->m_pkthdr.rcvif == NULL)
183 			m->m_pkthdr.rcvif = V_loif;
184 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
185 			m->m_pkthdr.csum_flags |=
186 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
187 			m->m_pkthdr.csum_data = 0xffff;
188 		}
189 #if defined(SCTP) || defined(SCTP_SUPPORT)
190 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
191 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
192 #endif
193 		m->m_pkthdr.csum_flags |=
194 			CSUM_IP_CHECKED | CSUM_IP_VALID;
195 
196 		*error = netisr_queue(NETISR_IP, m);
197 		return 1; /* Finished */
198 	}
199 	/* Or forward to some other address? */
200 	if ((m->m_flags & M_IP_NEXTHOP) &&
201 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
202 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
203 		m->m_flags |= M_SKIP_FIREWALL;
204 		m->m_flags &= ~M_IP_NEXTHOP;
205 		m_tag_delete(m, fwd_tag);
206 
207 		return -1; /* Reloop for CHANGE of dst */
208 	}
209 
210 	return 0;
211 }
212 
213 static int
214 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
215     const struct sockaddr *gw, struct route *ro, bool stamp_tag)
216 {
217 #ifdef KERN_TLS
218 	struct ktls_session *tls = NULL;
219 #endif
220 	struct m_snd_tag *mst;
221 	int error;
222 
223 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
224 	mst = NULL;
225 
226 #ifdef KERN_TLS
227 	/*
228 	 * If this is an unencrypted TLS record, save a reference to
229 	 * the record.  This local reference is used to call
230 	 * ktls_output_eagain after the mbuf has been freed (thus
231 	 * dropping the mbuf's reference) in if_output.
232 	 */
233 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
234 		tls = ktls_hold(m->m_next->m_epg_tls);
235 		mst = tls->snd_tag;
236 
237 		/*
238 		 * If a TLS session doesn't have a valid tag, it must
239 		 * have had an earlier ifp mismatch, so drop this
240 		 * packet.
241 		 */
242 		if (mst == NULL) {
243 			error = EAGAIN;
244 			goto done;
245 		}
246 		/*
247 		 * Always stamp tags that include NIC ktls.
248 		 */
249 		stamp_tag = true;
250 	}
251 #endif
252 #ifdef RATELIMIT
253 	if (inp != NULL && mst == NULL) {
254 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
255 		    (inp->inp_snd_tag != NULL &&
256 		    inp->inp_snd_tag->ifp != ifp))
257 			in_pcboutput_txrtlmt(inp, ifp, m);
258 
259 		if (inp->inp_snd_tag != NULL)
260 			mst = inp->inp_snd_tag;
261 	}
262 #endif
263 	if (stamp_tag && mst != NULL) {
264 		KASSERT(m->m_pkthdr.rcvif == NULL,
265 		    ("trying to add a send tag to a forwarded packet"));
266 		if (mst->ifp != ifp) {
267 			error = EAGAIN;
268 			goto done;
269 		}
270 
271 		/* stamp send tag on mbuf */
272 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
273 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
274 	}
275 
276 	error = (*ifp->if_output)(ifp, m, gw, ro);
277 
278 done:
279 	/* Check for route change invalidating send tags. */
280 #ifdef KERN_TLS
281 	if (tls != NULL) {
282 		if (error == EAGAIN)
283 			error = ktls_output_eagain(inp, tls);
284 		ktls_free(tls);
285 	}
286 #endif
287 #ifdef RATELIMIT
288 	if (error == EAGAIN)
289 		in_pcboutput_eagain(inp);
290 #endif
291 	return (error);
292 }
293 
294 /* rte<>ro_flags translation */
295 static inline void
296 rt_update_ro_flags(struct route *ro, const struct nhop_object *nh)
297 {
298 	int nh_flags = nh->nh_flags;
299 
300 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
301 
302 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
303 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
304 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
305 }
306 
307 /*
308  * IP output.  The packet in mbuf chain m contains a skeletal IP
309  * header (with len, off, ttl, proto, tos, src, dst).
310  * The mbuf chain containing the packet will be freed.
311  * The mbuf opt, if present, will not be freed.
312  * If route ro is present and has ro_rt initialized, route lookup would be
313  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
314  * then result of route lookup is stored in ro->ro_rt.
315  *
316  * In the IP forwarding case, the packet will arrive with options already
317  * inserted, so must have a NULL opt pointer.
318  */
319 int
320 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
321     struct ip_moptions *imo, struct inpcb *inp)
322 {
323 	MROUTER_RLOCK_TRACKER;
324 	struct rm_priotracker in_ifa_tracker;
325 	struct ip *ip;
326 	struct ifnet *ifp = NULL;	/* keep compiler happy */
327 	struct mbuf *m0;
328 	int hlen = sizeof (struct ip);
329 	int mtu = 0;
330 	int error = 0;
331 	int vlan_pcp = -1;
332 	struct sockaddr_in *dst;
333 	const struct sockaddr *gw;
334 	struct in_ifaddr *ia = NULL;
335 	struct in_addr src;
336 	int isbroadcast;
337 	uint16_t ip_len, ip_off;
338 	struct route iproute;
339 	uint32_t fibnum;
340 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
341 	int no_route_but_check_spd = 0;
342 #endif
343 
344 	M_ASSERTPKTHDR(m);
345 	NET_EPOCH_ASSERT();
346 
347 	if (inp != NULL) {
348 		INP_LOCK_ASSERT(inp);
349 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
350 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
351 			m->m_pkthdr.flowid = inp->inp_flowid;
352 			M_HASHTYPE_SET(m, inp->inp_flowtype);
353 		}
354 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
355 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
356 			    INP_2PCP_SHIFT;
357 #ifdef NUMA
358 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
359 #endif
360 	}
361 
362 	if (opt) {
363 		int len = 0;
364 		m = ip_insertoptions(m, opt, &len);
365 		if (len != 0)
366 			hlen = len; /* ip->ip_hl is updated above */
367 	}
368 	ip = mtod(m, struct ip *);
369 	ip_len = ntohs(ip->ip_len);
370 	ip_off = ntohs(ip->ip_off);
371 
372 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
373 		ip->ip_v = IPVERSION;
374 		ip->ip_hl = hlen >> 2;
375 		ip_fillid(ip);
376 	} else {
377 		/* Header already set, fetch hlen from there */
378 		hlen = ip->ip_hl << 2;
379 	}
380 	if ((flags & IP_FORWARDING) == 0)
381 		IPSTAT_INC(ips_localout);
382 
383 	/*
384 	 * dst/gw handling:
385 	 *
386 	 * gw is readonly but can point either to dst OR rt_gateway,
387 	 * therefore we need restore gw if we're redoing lookup.
388 	 */
389 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
390 	if (ro == NULL) {
391 		ro = &iproute;
392 		bzero(ro, sizeof (*ro));
393 	}
394 	dst = (struct sockaddr_in *)&ro->ro_dst;
395 	if (ro->ro_nh == NULL) {
396 		dst->sin_family = AF_INET;
397 		dst->sin_len = sizeof(*dst);
398 		dst->sin_addr = ip->ip_dst;
399 	}
400 	gw = (const struct sockaddr *)dst;
401 again:
402 	/*
403 	 * Validate route against routing table additions;
404 	 * a better/more specific route might have been added.
405 	 */
406 	if (inp != NULL && ro->ro_nh != NULL)
407 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
408 	/*
409 	 * If there is a cached route,
410 	 * check that it is to the same destination
411 	 * and is still up.  If not, free it and try again.
412 	 * The address family should also be checked in case of sharing the
413 	 * cache with IPv6.
414 	 * Also check whether routing cache needs invalidation.
415 	 */
416 	if (ro->ro_nh != NULL &&
417 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
418 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
419 		RO_INVALIDATE_CACHE(ro);
420 	ia = NULL;
421 	/*
422 	 * If routing to interface only, short circuit routing lookup.
423 	 * The use of an all-ones broadcast address implies this; an
424 	 * interface is specified by the broadcast address of an interface,
425 	 * or the destination address of a ptp interface.
426 	 */
427 	if (flags & IP_SENDONES) {
428 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
429 						      M_GETFIB(m)))) == NULL &&
430 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
431 						    M_GETFIB(m)))) == NULL) {
432 			IPSTAT_INC(ips_noroute);
433 			error = ENETUNREACH;
434 			goto bad;
435 		}
436 		ip->ip_dst.s_addr = INADDR_BROADCAST;
437 		dst->sin_addr = ip->ip_dst;
438 		ifp = ia->ia_ifp;
439 		mtu = ifp->if_mtu;
440 		ip->ip_ttl = 1;
441 		isbroadcast = 1;
442 		src = IA_SIN(ia)->sin_addr;
443 	} else if (flags & IP_ROUTETOIF) {
444 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
445 						    M_GETFIB(m)))) == NULL &&
446 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
447 						M_GETFIB(m)))) == NULL) {
448 			IPSTAT_INC(ips_noroute);
449 			error = ENETUNREACH;
450 			goto bad;
451 		}
452 		ifp = ia->ia_ifp;
453 		mtu = ifp->if_mtu;
454 		ip->ip_ttl = 1;
455 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
456 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
457 		src = IA_SIN(ia)->sin_addr;
458 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
459 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
460 		/*
461 		 * Bypass the normal routing lookup for multicast
462 		 * packets if the interface is specified.
463 		 */
464 		ifp = imo->imo_multicast_ifp;
465 		mtu = ifp->if_mtu;
466 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
467 		isbroadcast = 0;	/* fool gcc */
468 		/* Interface may have no addresses. */
469 		if (ia != NULL)
470 			src = IA_SIN(ia)->sin_addr;
471 		else
472 			src.s_addr = INADDR_ANY;
473 	} else if (ro != &iproute) {
474 		if (ro->ro_nh == NULL) {
475 			/*
476 			 * We want to do any cloning requested by the link
477 			 * layer, as this is probably required in all cases
478 			 * for correct operation (as it is for ARP).
479 			 */
480 			uint32_t flowid;
481 			flowid = m->m_pkthdr.flowid;
482 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
483 			    NHR_REF, flowid);
484 
485 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
486 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
487 				/*
488 				 * There is no route for this packet, but it is
489 				 * possible that a matching SPD entry exists.
490 				 */
491 				no_route_but_check_spd = 1;
492 				goto sendit;
493 #endif
494 				IPSTAT_INC(ips_noroute);
495 				error = EHOSTUNREACH;
496 				goto bad;
497 			}
498 		}
499 		struct nhop_object *nh = ro->ro_nh;
500 
501 		ia = ifatoia(nh->nh_ifa);
502 		ifp = nh->nh_ifp;
503 		counter_u64_add(nh->nh_pksent, 1);
504 		rt_update_ro_flags(ro, nh);
505 		if (nh->nh_flags & NHF_GATEWAY)
506 			gw = &nh->gw_sa;
507 		if (nh->nh_flags & NHF_HOST)
508 			isbroadcast = (nh->nh_flags & NHF_BROADCAST);
509 		else if ((ifp->if_flags & IFF_BROADCAST) && (gw->sa_family == AF_INET))
510 			isbroadcast = in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia);
511 		else
512 			isbroadcast = 0;
513 		mtu = nh->nh_mtu;
514 		src = IA_SIN(ia)->sin_addr;
515 	} else {
516 		struct nhop_object *nh;
517 
518 		nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE,
519 		    m->m_pkthdr.flowid);
520 		if (nh == NULL) {
521 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
522 			/*
523 			 * There is no route for this packet, but it is
524 			 * possible that a matching SPD entry exists.
525 			 */
526 			no_route_but_check_spd = 1;
527 			goto sendit;
528 #endif
529 			IPSTAT_INC(ips_noroute);
530 			error = EHOSTUNREACH;
531 			goto bad;
532 		}
533 		ifp = nh->nh_ifp;
534 		mtu = nh->nh_mtu;
535 		rt_update_ro_flags(ro, nh);
536 		if (nh->nh_flags & NHF_GATEWAY)
537 			gw = &nh->gw_sa;
538 		ia = ifatoia(nh->nh_ifa);
539 		src = IA_SIN(ia)->sin_addr;
540 		isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
541 		    (NHF_HOST | NHF_BROADCAST)) ||
542 		    ((ifp->if_flags & IFF_BROADCAST) &&
543 		    (gw->sa_family == AF_INET) &&
544 		    in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia)));
545 	}
546 
547 	/* Catch a possible divide by zero later. */
548 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
549 	    __func__, mtu, ro,
550 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
551 
552 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
553 		m->m_flags |= M_MCAST;
554 		/*
555 		 * IP destination address is multicast.  Make sure "gw"
556 		 * still points to the address in "ro".  (It may have been
557 		 * changed to point to a gateway address, above.)
558 		 */
559 		gw = (const struct sockaddr *)dst;
560 		/*
561 		 * See if the caller provided any multicast options
562 		 */
563 		if (imo != NULL) {
564 			ip->ip_ttl = imo->imo_multicast_ttl;
565 			if (imo->imo_multicast_vif != -1)
566 				ip->ip_src.s_addr =
567 				    ip_mcast_src ?
568 				    ip_mcast_src(imo->imo_multicast_vif) :
569 				    INADDR_ANY;
570 		} else
571 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
572 		/*
573 		 * Confirm that the outgoing interface supports multicast.
574 		 */
575 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
576 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
577 				IPSTAT_INC(ips_noroute);
578 				error = ENETUNREACH;
579 				goto bad;
580 			}
581 		}
582 		/*
583 		 * If source address not specified yet, use address
584 		 * of outgoing interface.
585 		 */
586 		if (ip->ip_src.s_addr == INADDR_ANY)
587 			ip->ip_src = src;
588 
589 		if ((imo == NULL && in_mcast_loop) ||
590 		    (imo && imo->imo_multicast_loop)) {
591 			/*
592 			 * Loop back multicast datagram if not expressly
593 			 * forbidden to do so, even if we are not a member
594 			 * of the group; ip_input() will filter it later,
595 			 * thus deferring a hash lookup and mutex acquisition
596 			 * at the expense of a cheap copy using m_copym().
597 			 */
598 			ip_mloopback(ifp, m, hlen);
599 		} else {
600 			/*
601 			 * If we are acting as a multicast router, perform
602 			 * multicast forwarding as if the packet had just
603 			 * arrived on the interface to which we are about
604 			 * to send.  The multicast forwarding function
605 			 * recursively calls this function, using the
606 			 * IP_FORWARDING flag to prevent infinite recursion.
607 			 *
608 			 * Multicasts that are looped back by ip_mloopback(),
609 			 * above, will be forwarded by the ip_input() routine,
610 			 * if necessary.
611 			 */
612 			MROUTER_RLOCK();
613 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
614 				/*
615 				 * If rsvp daemon is not running, do not
616 				 * set ip_moptions. This ensures that the packet
617 				 * is multicast and not just sent down one link
618 				 * as prescribed by rsvpd.
619 				 */
620 				if (!V_rsvp_on)
621 					imo = NULL;
622 				if (ip_mforward &&
623 				    ip_mforward(ip, ifp, m, imo) != 0) {
624 					MROUTER_RUNLOCK();
625 					m_freem(m);
626 					goto done;
627 				}
628 			}
629 			MROUTER_RUNLOCK();
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 = (const struct sockaddr *)dst;
720 			ip = mtod(m, struct ip *);
721 			goto again;
722 		}
723 	}
724 
725 	if (vlan_pcp > -1)
726 		EVL_APPLY_PRI(m, vlan_pcp);
727 
728 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
729 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
730 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
731 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
732 			IPSTAT_INC(ips_badaddr);
733 			error = EADDRNOTAVAIL;
734 			goto bad;
735 		}
736 	}
737 
738 	m->m_pkthdr.csum_flags |= CSUM_IP;
739 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
740 		m = mb_unmapped_to_ext(m);
741 		if (m == NULL) {
742 			IPSTAT_INC(ips_odropped);
743 			error = ENOBUFS;
744 			goto bad;
745 		}
746 		in_delayed_cksum(m);
747 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
748 	} else if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
749 		m = mb_unmapped_to_ext(m);
750 		if (m == NULL) {
751 			IPSTAT_INC(ips_odropped);
752 			error = ENOBUFS;
753 			goto bad;
754 		}
755 	}
756 #if defined(SCTP) || defined(SCTP_SUPPORT)
757 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
758 		m = mb_unmapped_to_ext(m);
759 		if (m == NULL) {
760 			IPSTAT_INC(ips_odropped);
761 			error = ENOBUFS;
762 			goto bad;
763 		}
764 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
765 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
766 	}
767 #endif
768 
769 	/*
770 	 * If small enough for interface, or the interface will take
771 	 * care of the fragmentation for us, we can just send directly.
772 	 * Note that if_vxlan could have requested TSO even though the outer
773 	 * frame is UDP.  It is correct to not fragment such datagrams and
774 	 * instead just pass them on to the driver.
775 	 */
776 	if (ip_len <= mtu ||
777 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
778 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
779 		ip->ip_sum = 0;
780 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
781 			ip->ip_sum = in_cksum(m, hlen);
782 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
783 		}
784 
785 		/*
786 		 * Record statistics for this interface address.
787 		 * With CSUM_TSO the byte/packet count will be slightly
788 		 * incorrect because we count the IP+TCP headers only
789 		 * once instead of for every generated packet.
790 		 */
791 		if (!(flags & IP_FORWARDING) && ia) {
792 			if (m->m_pkthdr.csum_flags &
793 			    (CSUM_TSO | CSUM_INNER_TSO))
794 				counter_u64_add(ia->ia_ifa.ifa_opackets,
795 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
796 			else
797 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
798 
799 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
800 		}
801 #ifdef MBUF_STRESS_TEST
802 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
803 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
804 #endif
805 		/*
806 		 * Reset layer specific mbuf flags
807 		 * to avoid confusing lower layers.
808 		 */
809 		m_clrprotoflags(m);
810 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
811 		error = ip_output_send(inp, ifp, m, gw, ro,
812 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
813 		goto done;
814 	}
815 
816 	/* Balk when DF bit is set or the interface didn't support TSO. */
817 	if ((ip_off & IP_DF) ||
818 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
819 		error = EMSGSIZE;
820 		IPSTAT_INC(ips_cantfrag);
821 		goto bad;
822 	}
823 
824 	/*
825 	 * Too large for interface; fragment if possible. If successful,
826 	 * on return, m will point to a list of packets to be sent.
827 	 */
828 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
829 	if (error)
830 		goto bad;
831 	for (; m; m = m0) {
832 		m0 = m->m_nextpkt;
833 		m->m_nextpkt = 0;
834 		if (error == 0) {
835 			/* Record statistics for this interface address. */
836 			if (ia != NULL) {
837 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
838 				counter_u64_add(ia->ia_ifa.ifa_obytes,
839 				    m->m_pkthdr.len);
840 			}
841 			/*
842 			 * Reset layer specific mbuf flags
843 			 * to avoid confusing upper layers.
844 			 */
845 			m_clrprotoflags(m);
846 
847 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
848 			    mtod(m, struct ip *), NULL);
849 			error = ip_output_send(inp, ifp, m, gw, ro, true);
850 		} else
851 			m_freem(m);
852 	}
853 
854 	if (error == 0)
855 		IPSTAT_INC(ips_fragmented);
856 
857 done:
858 	return (error);
859  bad:
860 	m_freem(m);
861 	goto done;
862 }
863 
864 /*
865  * Create a chain of fragments which fit the given mtu. m_frag points to the
866  * mbuf to be fragmented; on return it points to the chain with the fragments.
867  * Return 0 if no error. If error, m_frag may contain a partially built
868  * chain of fragments that should be freed by the caller.
869  *
870  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
871  */
872 int
873 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
874     u_long if_hwassist_flags)
875 {
876 	int error = 0;
877 	int hlen = ip->ip_hl << 2;
878 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
879 	int off;
880 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
881 	int firstlen;
882 	struct mbuf **mnext;
883 	int nfrags;
884 	uint16_t ip_len, ip_off;
885 
886 	ip_len = ntohs(ip->ip_len);
887 	ip_off = ntohs(ip->ip_off);
888 
889 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
890 		IPSTAT_INC(ips_cantfrag);
891 		return EMSGSIZE;
892 	}
893 
894 	/*
895 	 * Must be able to put at least 8 bytes per fragment.
896 	 */
897 	if (len < 8)
898 		return EMSGSIZE;
899 
900 	/*
901 	 * If the interface will not calculate checksums on
902 	 * fragmented packets, then do it here.
903 	 */
904 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
905 		m0 = mb_unmapped_to_ext(m0);
906 		if (m0 == NULL) {
907 			error = ENOBUFS;
908 			IPSTAT_INC(ips_odropped);
909 			goto done;
910 		}
911 		in_delayed_cksum(m0);
912 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
913 	}
914 #if defined(SCTP) || defined(SCTP_SUPPORT)
915 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
916 		m0 = mb_unmapped_to_ext(m0);
917 		if (m0 == NULL) {
918 			error = ENOBUFS;
919 			IPSTAT_INC(ips_odropped);
920 			goto done;
921 		}
922 		sctp_delayed_cksum(m0, hlen);
923 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
924 	}
925 #endif
926 	if (len > PAGE_SIZE) {
927 		/*
928 		 * Fragment large datagrams such that each segment
929 		 * contains a multiple of PAGE_SIZE amount of data,
930 		 * plus headers. This enables a receiver to perform
931 		 * page-flipping zero-copy optimizations.
932 		 *
933 		 * XXX When does this help given that sender and receiver
934 		 * could have different page sizes, and also mtu could
935 		 * be less than the receiver's page size ?
936 		 */
937 		int newlen;
938 
939 		off = MIN(mtu, m0->m_pkthdr.len);
940 
941 		/*
942 		 * firstlen (off - hlen) must be aligned on an
943 		 * 8-byte boundary
944 		 */
945 		if (off < hlen)
946 			goto smart_frag_failure;
947 		off = ((off - hlen) & ~7) + hlen;
948 		newlen = (~PAGE_MASK) & mtu;
949 		if ((newlen + sizeof (struct ip)) > mtu) {
950 			/* we failed, go back the default */
951 smart_frag_failure:
952 			newlen = len;
953 			off = hlen + len;
954 		}
955 		len = newlen;
956 
957 	} else {
958 		off = hlen + len;
959 	}
960 
961 	firstlen = off - hlen;
962 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
963 
964 	/*
965 	 * Loop through length of segment after first fragment,
966 	 * make new header and copy data of each part and link onto chain.
967 	 * Here, m0 is the original packet, m is the fragment being created.
968 	 * The fragments are linked off the m_nextpkt of the original
969 	 * packet, which after processing serves as the first fragment.
970 	 */
971 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
972 		struct ip *mhip;	/* ip header on the fragment */
973 		struct mbuf *m;
974 		int mhlen = sizeof (struct ip);
975 
976 		m = m_gethdr(M_NOWAIT, MT_DATA);
977 		if (m == NULL) {
978 			error = ENOBUFS;
979 			IPSTAT_INC(ips_odropped);
980 			goto done;
981 		}
982 		/*
983 		 * Make sure the complete packet header gets copied
984 		 * from the originating mbuf to the newly created
985 		 * mbuf. This also ensures that existing firewall
986 		 * classification(s), VLAN tags and so on get copied
987 		 * to the resulting fragmented packet(s):
988 		 */
989 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
990 			m_free(m);
991 			error = ENOBUFS;
992 			IPSTAT_INC(ips_odropped);
993 			goto done;
994 		}
995 		/*
996 		 * In the first mbuf, leave room for the link header, then
997 		 * copy the original IP header including options. The payload
998 		 * goes into an additional mbuf chain returned by m_copym().
999 		 */
1000 		m->m_data += max_linkhdr;
1001 		mhip = mtod(m, struct ip *);
1002 		*mhip = *ip;
1003 		if (hlen > sizeof (struct ip)) {
1004 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1005 			mhip->ip_v = IPVERSION;
1006 			mhip->ip_hl = mhlen >> 2;
1007 		}
1008 		m->m_len = mhlen;
1009 		/* XXX do we need to add ip_off below ? */
1010 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1011 		if (off + len >= ip_len)
1012 			len = ip_len - off;
1013 		else
1014 			mhip->ip_off |= IP_MF;
1015 		mhip->ip_len = htons((u_short)(len + mhlen));
1016 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1017 		if (m->m_next == NULL) {	/* copy failed */
1018 			m_free(m);
1019 			error = ENOBUFS;	/* ??? */
1020 			IPSTAT_INC(ips_odropped);
1021 			goto done;
1022 		}
1023 		m->m_pkthdr.len = mhlen + len;
1024 #ifdef MAC
1025 		mac_netinet_fragment(m0, m);
1026 #endif
1027 		mhip->ip_off = htons(mhip->ip_off);
1028 		mhip->ip_sum = 0;
1029 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1030 			mhip->ip_sum = in_cksum(m, mhlen);
1031 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1032 		}
1033 		*mnext = m;
1034 		mnext = &m->m_nextpkt;
1035 	}
1036 	IPSTAT_ADD(ips_ofragments, nfrags);
1037 
1038 	/*
1039 	 * Update first fragment by trimming what's been copied out
1040 	 * and updating header.
1041 	 */
1042 	m_adj(m0, hlen + firstlen - ip_len);
1043 	m0->m_pkthdr.len = hlen + firstlen;
1044 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1045 	ip->ip_off = htons(ip_off | IP_MF);
1046 	ip->ip_sum = 0;
1047 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1048 		ip->ip_sum = in_cksum(m0, hlen);
1049 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1050 	}
1051 
1052 done:
1053 	*m_frag = m0;
1054 	return error;
1055 }
1056 
1057 void
1058 in_delayed_cksum(struct mbuf *m)
1059 {
1060 	struct ip *ip;
1061 	struct udphdr *uh;
1062 	uint16_t cklen, csum, offset;
1063 
1064 	ip = mtod(m, struct ip *);
1065 	offset = ip->ip_hl << 2 ;
1066 
1067 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1068 		/* if udp header is not in the first mbuf copy udplen */
1069 		if (offset + sizeof(struct udphdr) > m->m_len) {
1070 			m_copydata(m, offset + offsetof(struct udphdr,
1071 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1072 			cklen = ntohs(cklen);
1073 		} else {
1074 			uh = (struct udphdr *)mtodo(m, offset);
1075 			cklen = ntohs(uh->uh_ulen);
1076 		}
1077 		csum = in_cksum_skip(m, cklen + offset, offset);
1078 		if (csum == 0)
1079 			csum = 0xffff;
1080 	} else {
1081 		cklen = ntohs(ip->ip_len);
1082 		csum = in_cksum_skip(m, cklen, offset);
1083 	}
1084 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1085 
1086 	if (offset + sizeof(csum) > m->m_len)
1087 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1088 	else
1089 		*(u_short *)mtodo(m, offset) = csum;
1090 }
1091 
1092 /*
1093  * IP socket option processing.
1094  */
1095 int
1096 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1097 {
1098 	struct inpcb *inp = sotoinpcb(so);
1099 	int	error, optval;
1100 #ifdef	RSS
1101 	uint32_t rss_bucket;
1102 	int retval;
1103 #endif
1104 
1105 	error = optval = 0;
1106 	if (sopt->sopt_level != IPPROTO_IP) {
1107 		error = EINVAL;
1108 
1109 		if (sopt->sopt_level == SOL_SOCKET &&
1110 		    sopt->sopt_dir == SOPT_SET) {
1111 			switch (sopt->sopt_name) {
1112 			case SO_REUSEADDR:
1113 				INP_WLOCK(inp);
1114 				if ((so->so_options & SO_REUSEADDR) != 0)
1115 					inp->inp_flags2 |= INP_REUSEADDR;
1116 				else
1117 					inp->inp_flags2 &= ~INP_REUSEADDR;
1118 				INP_WUNLOCK(inp);
1119 				error = 0;
1120 				break;
1121 			case SO_REUSEPORT:
1122 				INP_WLOCK(inp);
1123 				if ((so->so_options & SO_REUSEPORT) != 0)
1124 					inp->inp_flags2 |= INP_REUSEPORT;
1125 				else
1126 					inp->inp_flags2 &= ~INP_REUSEPORT;
1127 				INP_WUNLOCK(inp);
1128 				error = 0;
1129 				break;
1130 			case SO_REUSEPORT_LB:
1131 				INP_WLOCK(inp);
1132 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1133 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1134 				else
1135 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1136 				INP_WUNLOCK(inp);
1137 				error = 0;
1138 				break;
1139 			case SO_SETFIB:
1140 				INP_WLOCK(inp);
1141 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1142 				INP_WUNLOCK(inp);
1143 				error = 0;
1144 				break;
1145 			case SO_MAX_PACING_RATE:
1146 #ifdef RATELIMIT
1147 				INP_WLOCK(inp);
1148 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1149 				INP_WUNLOCK(inp);
1150 				error = 0;
1151 #else
1152 				error = EOPNOTSUPP;
1153 #endif
1154 				break;
1155 			default:
1156 				break;
1157 			}
1158 		}
1159 		return (error);
1160 	}
1161 
1162 	switch (sopt->sopt_dir) {
1163 	case SOPT_SET:
1164 		switch (sopt->sopt_name) {
1165 		case IP_OPTIONS:
1166 #ifdef notyet
1167 		case IP_RETOPTS:
1168 #endif
1169 		{
1170 			struct mbuf *m;
1171 			if (sopt->sopt_valsize > MLEN) {
1172 				error = EMSGSIZE;
1173 				break;
1174 			}
1175 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1176 			if (m == NULL) {
1177 				error = ENOBUFS;
1178 				break;
1179 			}
1180 			m->m_len = sopt->sopt_valsize;
1181 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1182 					    m->m_len);
1183 			if (error) {
1184 				m_free(m);
1185 				break;
1186 			}
1187 			INP_WLOCK(inp);
1188 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1189 			INP_WUNLOCK(inp);
1190 			return (error);
1191 		}
1192 
1193 		case IP_BINDANY:
1194 			if (sopt->sopt_td != NULL) {
1195 				error = priv_check(sopt->sopt_td,
1196 				    PRIV_NETINET_BINDANY);
1197 				if (error)
1198 					break;
1199 			}
1200 			/* FALLTHROUGH */
1201 		case IP_BINDMULTI:
1202 #ifdef	RSS
1203 		case IP_RSS_LISTEN_BUCKET:
1204 #endif
1205 		case IP_TOS:
1206 		case IP_TTL:
1207 		case IP_MINTTL:
1208 		case IP_RECVOPTS:
1209 		case IP_RECVRETOPTS:
1210 		case IP_ORIGDSTADDR:
1211 		case IP_RECVDSTADDR:
1212 		case IP_RECVTTL:
1213 		case IP_RECVIF:
1214 		case IP_ONESBCAST:
1215 		case IP_DONTFRAG:
1216 		case IP_RECVTOS:
1217 		case IP_RECVFLOWID:
1218 #ifdef	RSS
1219 		case IP_RECVRSSBUCKETID:
1220 #endif
1221 		case IP_VLAN_PCP:
1222 			error = sooptcopyin(sopt, &optval, sizeof optval,
1223 					    sizeof optval);
1224 			if (error)
1225 				break;
1226 
1227 			switch (sopt->sopt_name) {
1228 			case IP_TOS:
1229 				inp->inp_ip_tos = optval;
1230 				break;
1231 
1232 			case IP_TTL:
1233 				inp->inp_ip_ttl = optval;
1234 				break;
1235 
1236 			case IP_MINTTL:
1237 				if (optval >= 0 && optval <= MAXTTL)
1238 					inp->inp_ip_minttl = optval;
1239 				else
1240 					error = EINVAL;
1241 				break;
1242 
1243 #define	OPTSET(bit) do {						\
1244 	INP_WLOCK(inp);							\
1245 	if (optval)							\
1246 		inp->inp_flags |= bit;					\
1247 	else								\
1248 		inp->inp_flags &= ~bit;					\
1249 	INP_WUNLOCK(inp);						\
1250 } while (0)
1251 
1252 #define	OPTSET2(bit, val) do {						\
1253 	INP_WLOCK(inp);							\
1254 	if (val)							\
1255 		inp->inp_flags2 |= bit;					\
1256 	else								\
1257 		inp->inp_flags2 &= ~bit;				\
1258 	INP_WUNLOCK(inp);						\
1259 } while (0)
1260 
1261 			case IP_RECVOPTS:
1262 				OPTSET(INP_RECVOPTS);
1263 				break;
1264 
1265 			case IP_RECVRETOPTS:
1266 				OPTSET(INP_RECVRETOPTS);
1267 				break;
1268 
1269 			case IP_RECVDSTADDR:
1270 				OPTSET(INP_RECVDSTADDR);
1271 				break;
1272 
1273 			case IP_ORIGDSTADDR:
1274 				OPTSET2(INP_ORIGDSTADDR, optval);
1275 				break;
1276 
1277 			case IP_RECVTTL:
1278 				OPTSET(INP_RECVTTL);
1279 				break;
1280 
1281 			case IP_RECVIF:
1282 				OPTSET(INP_RECVIF);
1283 				break;
1284 
1285 			case IP_ONESBCAST:
1286 				OPTSET(INP_ONESBCAST);
1287 				break;
1288 			case IP_DONTFRAG:
1289 				OPTSET(INP_DONTFRAG);
1290 				break;
1291 			case IP_BINDANY:
1292 				OPTSET(INP_BINDANY);
1293 				break;
1294 			case IP_RECVTOS:
1295 				OPTSET(INP_RECVTOS);
1296 				break;
1297 			case IP_BINDMULTI:
1298 				OPTSET2(INP_BINDMULTI, optval);
1299 				break;
1300 			case IP_RECVFLOWID:
1301 				OPTSET2(INP_RECVFLOWID, optval);
1302 				break;
1303 #ifdef	RSS
1304 			case IP_RSS_LISTEN_BUCKET:
1305 				if ((optval >= 0) &&
1306 				    (optval < rss_getnumbuckets())) {
1307 					inp->inp_rss_listen_bucket = optval;
1308 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1309 				} else {
1310 					error = EINVAL;
1311 				}
1312 				break;
1313 			case IP_RECVRSSBUCKETID:
1314 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1315 				break;
1316 #endif
1317 			case IP_VLAN_PCP:
1318 				if ((optval >= -1) && (optval <=
1319 				    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1320 					if (optval == -1) {
1321 						INP_WLOCK(inp);
1322 						inp->inp_flags2 &=
1323 						    ~(INP_2PCP_SET |
1324 						      INP_2PCP_MASK);
1325 						INP_WUNLOCK(inp);
1326 					} else {
1327 						INP_WLOCK(inp);
1328 						inp->inp_flags2 |=
1329 						    INP_2PCP_SET;
1330 						inp->inp_flags2 &=
1331 						    ~INP_2PCP_MASK;
1332 						inp->inp_flags2 |=
1333 						    optval << INP_2PCP_SHIFT;
1334 						INP_WUNLOCK(inp);
1335 					}
1336 				} else
1337 					error = EINVAL;
1338 				break;
1339 			}
1340 			break;
1341 #undef OPTSET
1342 #undef OPTSET2
1343 
1344 		/*
1345 		 * Multicast socket options are processed by the in_mcast
1346 		 * module.
1347 		 */
1348 		case IP_MULTICAST_IF:
1349 		case IP_MULTICAST_VIF:
1350 		case IP_MULTICAST_TTL:
1351 		case IP_MULTICAST_LOOP:
1352 		case IP_ADD_MEMBERSHIP:
1353 		case IP_DROP_MEMBERSHIP:
1354 		case IP_ADD_SOURCE_MEMBERSHIP:
1355 		case IP_DROP_SOURCE_MEMBERSHIP:
1356 		case IP_BLOCK_SOURCE:
1357 		case IP_UNBLOCK_SOURCE:
1358 		case IP_MSFILTER:
1359 		case MCAST_JOIN_GROUP:
1360 		case MCAST_LEAVE_GROUP:
1361 		case MCAST_JOIN_SOURCE_GROUP:
1362 		case MCAST_LEAVE_SOURCE_GROUP:
1363 		case MCAST_BLOCK_SOURCE:
1364 		case MCAST_UNBLOCK_SOURCE:
1365 			error = inp_setmoptions(inp, sopt);
1366 			break;
1367 
1368 		case IP_PORTRANGE:
1369 			error = sooptcopyin(sopt, &optval, sizeof optval,
1370 					    sizeof optval);
1371 			if (error)
1372 				break;
1373 
1374 			INP_WLOCK(inp);
1375 			switch (optval) {
1376 			case IP_PORTRANGE_DEFAULT:
1377 				inp->inp_flags &= ~(INP_LOWPORT);
1378 				inp->inp_flags &= ~(INP_HIGHPORT);
1379 				break;
1380 
1381 			case IP_PORTRANGE_HIGH:
1382 				inp->inp_flags &= ~(INP_LOWPORT);
1383 				inp->inp_flags |= INP_HIGHPORT;
1384 				break;
1385 
1386 			case IP_PORTRANGE_LOW:
1387 				inp->inp_flags &= ~(INP_HIGHPORT);
1388 				inp->inp_flags |= INP_LOWPORT;
1389 				break;
1390 
1391 			default:
1392 				error = EINVAL;
1393 				break;
1394 			}
1395 			INP_WUNLOCK(inp);
1396 			break;
1397 
1398 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1399 		case IP_IPSEC_POLICY:
1400 			if (IPSEC_ENABLED(ipv4)) {
1401 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1402 				break;
1403 			}
1404 			/* FALLTHROUGH */
1405 #endif /* IPSEC */
1406 
1407 		default:
1408 			error = ENOPROTOOPT;
1409 			break;
1410 		}
1411 		break;
1412 
1413 	case SOPT_GET:
1414 		switch (sopt->sopt_name) {
1415 		case IP_OPTIONS:
1416 		case IP_RETOPTS:
1417 			INP_RLOCK(inp);
1418 			if (inp->inp_options) {
1419 				struct mbuf *options;
1420 
1421 				options = m_copym(inp->inp_options, 0,
1422 				    M_COPYALL, M_NOWAIT);
1423 				INP_RUNLOCK(inp);
1424 				if (options != NULL) {
1425 					error = sooptcopyout(sopt,
1426 							     mtod(options, char *),
1427 							     options->m_len);
1428 					m_freem(options);
1429 				} else
1430 					error = ENOMEM;
1431 			} else {
1432 				INP_RUNLOCK(inp);
1433 				sopt->sopt_valsize = 0;
1434 			}
1435 			break;
1436 
1437 		case IP_TOS:
1438 		case IP_TTL:
1439 		case IP_MINTTL:
1440 		case IP_RECVOPTS:
1441 		case IP_RECVRETOPTS:
1442 		case IP_ORIGDSTADDR:
1443 		case IP_RECVDSTADDR:
1444 		case IP_RECVTTL:
1445 		case IP_RECVIF:
1446 		case IP_PORTRANGE:
1447 		case IP_ONESBCAST:
1448 		case IP_DONTFRAG:
1449 		case IP_BINDANY:
1450 		case IP_RECVTOS:
1451 		case IP_BINDMULTI:
1452 		case IP_FLOWID:
1453 		case IP_FLOWTYPE:
1454 		case IP_RECVFLOWID:
1455 #ifdef	RSS
1456 		case IP_RSSBUCKETID:
1457 		case IP_RECVRSSBUCKETID:
1458 #endif
1459 		case IP_VLAN_PCP:
1460 			switch (sopt->sopt_name) {
1461 			case IP_TOS:
1462 				optval = inp->inp_ip_tos;
1463 				break;
1464 
1465 			case IP_TTL:
1466 				optval = inp->inp_ip_ttl;
1467 				break;
1468 
1469 			case IP_MINTTL:
1470 				optval = inp->inp_ip_minttl;
1471 				break;
1472 
1473 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1474 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1475 
1476 			case IP_RECVOPTS:
1477 				optval = OPTBIT(INP_RECVOPTS);
1478 				break;
1479 
1480 			case IP_RECVRETOPTS:
1481 				optval = OPTBIT(INP_RECVRETOPTS);
1482 				break;
1483 
1484 			case IP_RECVDSTADDR:
1485 				optval = OPTBIT(INP_RECVDSTADDR);
1486 				break;
1487 
1488 			case IP_ORIGDSTADDR:
1489 				optval = OPTBIT2(INP_ORIGDSTADDR);
1490 				break;
1491 
1492 			case IP_RECVTTL:
1493 				optval = OPTBIT(INP_RECVTTL);
1494 				break;
1495 
1496 			case IP_RECVIF:
1497 				optval = OPTBIT(INP_RECVIF);
1498 				break;
1499 
1500 			case IP_PORTRANGE:
1501 				if (inp->inp_flags & INP_HIGHPORT)
1502 					optval = IP_PORTRANGE_HIGH;
1503 				else if (inp->inp_flags & INP_LOWPORT)
1504 					optval = IP_PORTRANGE_LOW;
1505 				else
1506 					optval = 0;
1507 				break;
1508 
1509 			case IP_ONESBCAST:
1510 				optval = OPTBIT(INP_ONESBCAST);
1511 				break;
1512 			case IP_DONTFRAG:
1513 				optval = OPTBIT(INP_DONTFRAG);
1514 				break;
1515 			case IP_BINDANY:
1516 				optval = OPTBIT(INP_BINDANY);
1517 				break;
1518 			case IP_RECVTOS:
1519 				optval = OPTBIT(INP_RECVTOS);
1520 				break;
1521 			case IP_FLOWID:
1522 				optval = inp->inp_flowid;
1523 				break;
1524 			case IP_FLOWTYPE:
1525 				optval = inp->inp_flowtype;
1526 				break;
1527 			case IP_RECVFLOWID:
1528 				optval = OPTBIT2(INP_RECVFLOWID);
1529 				break;
1530 #ifdef	RSS
1531 			case IP_RSSBUCKETID:
1532 				retval = rss_hash2bucket(inp->inp_flowid,
1533 				    inp->inp_flowtype,
1534 				    &rss_bucket);
1535 				if (retval == 0)
1536 					optval = rss_bucket;
1537 				else
1538 					error = EINVAL;
1539 				break;
1540 			case IP_RECVRSSBUCKETID:
1541 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1542 				break;
1543 #endif
1544 			case IP_BINDMULTI:
1545 				optval = OPTBIT2(INP_BINDMULTI);
1546 				break;
1547 			case IP_VLAN_PCP:
1548 				if (OPTBIT2(INP_2PCP_SET)) {
1549 					optval = (inp->inp_flags2 &
1550 					    INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1551 				} else {
1552 					optval = -1;
1553 				}
1554 				break;
1555 			}
1556 			error = sooptcopyout(sopt, &optval, sizeof optval);
1557 			break;
1558 
1559 		/*
1560 		 * Multicast socket options are processed by the in_mcast
1561 		 * module.
1562 		 */
1563 		case IP_MULTICAST_IF:
1564 		case IP_MULTICAST_VIF:
1565 		case IP_MULTICAST_TTL:
1566 		case IP_MULTICAST_LOOP:
1567 		case IP_MSFILTER:
1568 			error = inp_getmoptions(inp, sopt);
1569 			break;
1570 
1571 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1572 		case IP_IPSEC_POLICY:
1573 			if (IPSEC_ENABLED(ipv4)) {
1574 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1575 				break;
1576 			}
1577 			/* FALLTHROUGH */
1578 #endif /* IPSEC */
1579 
1580 		default:
1581 			error = ENOPROTOOPT;
1582 			break;
1583 		}
1584 		break;
1585 	}
1586 	return (error);
1587 }
1588 
1589 /*
1590  * Routine called from ip_output() to loop back a copy of an IP multicast
1591  * packet to the input queue of a specified interface.  Note that this
1592  * calls the output routine of the loopback "driver", but with an interface
1593  * pointer that might NOT be a loopback interface -- evil, but easier than
1594  * replicating that code here.
1595  */
1596 static void
1597 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1598 {
1599 	struct ip *ip;
1600 	struct mbuf *copym;
1601 
1602 	/*
1603 	 * Make a deep copy of the packet because we're going to
1604 	 * modify the pack in order to generate checksums.
1605 	 */
1606 	copym = m_dup(m, M_NOWAIT);
1607 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1608 		copym = m_pullup(copym, hlen);
1609 	if (copym != NULL) {
1610 		/* If needed, compute the checksum and mark it as valid. */
1611 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1612 			in_delayed_cksum(copym);
1613 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1614 			copym->m_pkthdr.csum_flags |=
1615 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1616 			copym->m_pkthdr.csum_data = 0xffff;
1617 		}
1618 		/*
1619 		 * We don't bother to fragment if the IP length is greater
1620 		 * than the interface's MTU.  Can this possibly matter?
1621 		 */
1622 		ip = mtod(copym, struct ip *);
1623 		ip->ip_sum = 0;
1624 		ip->ip_sum = in_cksum(copym, hlen);
1625 		if_simloop(ifp, copym, AF_INET, 0);
1626 	}
1627 }
1628