xref: /freebsd/sys/netinet/ip_output.c (revision 036d2e814bf0f5d88ffb4b24c159320894541757)
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 #ifdef RADIX_MPATH
71 #include <net/radix_mpath.h>
72 #endif
73 #include <net/rss_config.h>
74 #include <net/vnet.h>
75 
76 #include <netinet/in.h>
77 #include <netinet/in_fib.h>
78 #include <netinet/in_kdtrace.h>
79 #include <netinet/in_systm.h>
80 #include <netinet/ip.h>
81 #include <netinet/in_pcb.h>
82 #include <netinet/in_rss.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip_var.h>
85 #include <netinet/ip_options.h>
86 
87 #include <netinet/udp.h>
88 #include <netinet/udp_var.h>
89 
90 #ifdef SCTP
91 #include <netinet/sctp.h>
92 #include <netinet/sctp_crc32.h>
93 #endif
94 
95 #include <netipsec/ipsec_support.h>
96 
97 #include <machine/in_cksum.h>
98 
99 #include <security/mac/mac_framework.h>
100 
101 #ifdef MBUF_STRESS_TEST
102 static int mbuf_frag_size = 0;
103 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
104 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
105 #endif
106 
107 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
108 
109 
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 = EPERM;
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 #ifdef SCTP
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 #ifdef SCTP
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_in *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_ext.ext_pgs->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 #endif
248 #ifdef RATELIMIT
249 	if (inp != NULL && mst == NULL) {
250 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
251 		    (inp->inp_snd_tag != NULL &&
252 		    inp->inp_snd_tag->ifp != ifp))
253 			in_pcboutput_txrtlmt(inp, ifp, m);
254 
255 		if (inp->inp_snd_tag != NULL)
256 			mst = inp->inp_snd_tag;
257 	}
258 #endif
259 	if (stamp_tag && mst != NULL) {
260 		KASSERT(m->m_pkthdr.rcvif == NULL,
261 		    ("trying to add a send tag to a forwarded packet"));
262 		if (mst->ifp != ifp) {
263 			error = EAGAIN;
264 			goto done;
265 		}
266 
267 		/* stamp send tag on mbuf */
268 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
269 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
270 	}
271 
272 	error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
273 
274 done:
275 	/* Check for route change invalidating send tags. */
276 #ifdef KERN_TLS
277 	if (tls != NULL) {
278 		if (error == EAGAIN)
279 			error = ktls_output_eagain(inp, tls);
280 		ktls_free(tls);
281 	}
282 #endif
283 #ifdef RATELIMIT
284 	if (error == EAGAIN)
285 		in_pcboutput_eagain(inp);
286 #endif
287 	return (error);
288 }
289 
290 /*
291  * IP output.  The packet in mbuf chain m contains a skeletal IP
292  * header (with len, off, ttl, proto, tos, src, dst).
293  * The mbuf chain containing the packet will be freed.
294  * The mbuf opt, if present, will not be freed.
295  * If route ro is present and has ro_rt initialized, route lookup would be
296  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
297  * then result of route lookup is stored in ro->ro_rt.
298  *
299  * In the IP forwarding case, the packet will arrive with options already
300  * inserted, so must have a NULL opt pointer.
301  */
302 int
303 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
304     struct ip_moptions *imo, struct inpcb *inp)
305 {
306 	struct rm_priotracker in_ifa_tracker;
307 	struct epoch_tracker et;
308 	struct ip *ip;
309 	struct ifnet *ifp = NULL;	/* keep compiler happy */
310 	struct mbuf *m0;
311 	int hlen = sizeof (struct ip);
312 	int mtu;
313 	int error = 0;
314 	struct sockaddr_in *dst, sin;
315 	const struct sockaddr_in *gw;
316 	struct in_ifaddr *ia;
317 	struct in_addr src;
318 	int isbroadcast;
319 	uint16_t ip_len, ip_off;
320 	uint32_t fibnum;
321 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
322 	int no_route_but_check_spd = 0;
323 #endif
324 
325 	M_ASSERTPKTHDR(m);
326 
327 	if (inp != NULL) {
328 		INP_LOCK_ASSERT(inp);
329 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
330 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
331 			m->m_pkthdr.flowid = inp->inp_flowid;
332 			M_HASHTYPE_SET(m, inp->inp_flowtype);
333 		}
334 #ifdef NUMA
335 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
336 #endif
337 	}
338 
339 	if (opt) {
340 		int len = 0;
341 		m = ip_insertoptions(m, opt, &len);
342 		if (len != 0)
343 			hlen = len; /* ip->ip_hl is updated above */
344 	}
345 	ip = mtod(m, struct ip *);
346 	ip_len = ntohs(ip->ip_len);
347 	ip_off = ntohs(ip->ip_off);
348 
349 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
350 		ip->ip_v = IPVERSION;
351 		ip->ip_hl = hlen >> 2;
352 		ip_fillid(ip);
353 	} else {
354 		/* Header already set, fetch hlen from there */
355 		hlen = ip->ip_hl << 2;
356 	}
357 	if ((flags & IP_FORWARDING) == 0)
358 		IPSTAT_INC(ips_localout);
359 
360 	/*
361 	 * dst/gw handling:
362 	 *
363 	 * gw is readonly but can point either to dst OR rt_gateway,
364 	 * therefore we need restore gw if we're redoing lookup.
365 	 */
366 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
367 	if (ro != NULL)
368 		dst = (struct sockaddr_in *)&ro->ro_dst;
369 	else
370 		dst = &sin;
371 	if (ro == NULL || ro->ro_rt == NULL) {
372 		bzero(dst, sizeof(*dst));
373 		dst->sin_family = AF_INET;
374 		dst->sin_len = sizeof(*dst);
375 		dst->sin_addr = ip->ip_dst;
376 	}
377 	gw = dst;
378 	NET_EPOCH_ENTER(et);
379 again:
380 	/*
381 	 * Validate route against routing table additions;
382 	 * a better/more specific route might have been added.
383 	 */
384 	if (inp != NULL && ro != NULL && ro->ro_rt != NULL)
385 		RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
386 	/*
387 	 * If there is a cached route,
388 	 * check that it is to the same destination
389 	 * and is still up.  If not, free it and try again.
390 	 * The address family should also be checked in case of sharing the
391 	 * cache with IPv6.
392 	 * Also check whether routing cache needs invalidation.
393 	 */
394 	if (ro != NULL && ro->ro_rt != NULL &&
395 	    ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
396 	    ro->ro_rt->rt_ifp == NULL || !RT_LINK_IS_UP(ro->ro_rt->rt_ifp) ||
397 	    dst->sin_family != AF_INET ||
398 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
399 		RO_INVALIDATE_CACHE(ro);
400 	ia = NULL;
401 	/*
402 	 * If routing to interface only, short circuit routing lookup.
403 	 * The use of an all-ones broadcast address implies this; an
404 	 * interface is specified by the broadcast address of an interface,
405 	 * or the destination address of a ptp interface.
406 	 */
407 	if (flags & IP_SENDONES) {
408 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
409 						      M_GETFIB(m)))) == NULL &&
410 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
411 						    M_GETFIB(m)))) == NULL) {
412 			IPSTAT_INC(ips_noroute);
413 			error = ENETUNREACH;
414 			goto bad;
415 		}
416 		ip->ip_dst.s_addr = INADDR_BROADCAST;
417 		dst->sin_addr = ip->ip_dst;
418 		ifp = ia->ia_ifp;
419 		mtu = ifp->if_mtu;
420 		ip->ip_ttl = 1;
421 		isbroadcast = 1;
422 		src = IA_SIN(ia)->sin_addr;
423 	} else if (flags & IP_ROUTETOIF) {
424 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
425 						    M_GETFIB(m)))) == NULL &&
426 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
427 						M_GETFIB(m)))) == NULL) {
428 			IPSTAT_INC(ips_noroute);
429 			error = ENETUNREACH;
430 			goto bad;
431 		}
432 		ifp = ia->ia_ifp;
433 		mtu = ifp->if_mtu;
434 		ip->ip_ttl = 1;
435 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
436 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
437 		src = IA_SIN(ia)->sin_addr;
438 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
439 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
440 		/*
441 		 * Bypass the normal routing lookup for multicast
442 		 * packets if the interface is specified.
443 		 */
444 		ifp = imo->imo_multicast_ifp;
445 		mtu = ifp->if_mtu;
446 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
447 		isbroadcast = 0;	/* fool gcc */
448 		/* Interface may have no addresses. */
449 		if (ia != NULL)
450 			src = IA_SIN(ia)->sin_addr;
451 		else
452 			src.s_addr = INADDR_ANY;
453 	} else if (ro != NULL) {
454 		if (ro->ro_rt == NULL) {
455 			/*
456 			 * We want to do any cloning requested by the link
457 			 * layer, as this is probably required in all cases
458 			 * for correct operation (as it is for ARP).
459 			 */
460 #ifdef RADIX_MPATH
461 			rtalloc_mpath_fib(ro,
462 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
463 			    fibnum);
464 #else
465 			in_rtalloc_ign(ro, 0, fibnum);
466 #endif
467 			if (ro->ro_rt == NULL ||
468 			    (ro->ro_rt->rt_flags & RTF_UP) == 0 ||
469 			    ro->ro_rt->rt_ifp == NULL ||
470 			    !RT_LINK_IS_UP(ro->ro_rt->rt_ifp)) {
471 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
472 				/*
473 				 * There is no route for this packet, but it is
474 				 * possible that a matching SPD entry exists.
475 				 */
476 				no_route_but_check_spd = 1;
477 				mtu = 0; /* Silence GCC warning. */
478 				goto sendit;
479 #endif
480 				IPSTAT_INC(ips_noroute);
481 				error = EHOSTUNREACH;
482 				goto bad;
483 			}
484 		}
485 		ia = ifatoia(ro->ro_rt->rt_ifa);
486 		ifp = ro->ro_rt->rt_ifp;
487 		counter_u64_add(ro->ro_rt->rt_pksent, 1);
488 		rt_update_ro_flags(ro);
489 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
490 			gw = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
491 		if (ro->ro_rt->rt_flags & RTF_HOST)
492 			isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
493 		else if (ifp->if_flags & IFF_BROADCAST)
494 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
495 		else
496 			isbroadcast = 0;
497 		if (ro->ro_rt->rt_flags & RTF_HOST)
498 			mtu = ro->ro_rt->rt_mtu;
499 		else
500 			mtu = ifp->if_mtu;
501 		src = IA_SIN(ia)->sin_addr;
502 	} else {
503 		struct nhop4_extended nh;
504 
505 		bzero(&nh, sizeof(nh));
506 		if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh) !=
507 		    0) {
508 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
509 			/*
510 			 * There is no route for this packet, but it is
511 			 * possible that a matching SPD entry exists.
512 			 */
513 			no_route_but_check_spd = 1;
514 			mtu = 0; /* Silence GCC warning. */
515 			goto sendit;
516 #endif
517 			IPSTAT_INC(ips_noroute);
518 			error = EHOSTUNREACH;
519 			goto bad;
520 		}
521 		ifp = nh.nh_ifp;
522 		mtu = nh.nh_mtu;
523 		/*
524 		 * We are rewriting here dst to be gw actually, contradicting
525 		 * comment at the beginning of the function. However, in this
526 		 * case we are always dealing with on stack dst.
527 		 * In case if pfil(9) sends us back to beginning of the
528 		 * function, the dst would be rewritten by ip_output_pfil().
529 		 */
530 		MPASS(dst == &sin);
531 		dst->sin_addr = nh.nh_addr;
532 		ia = nh.nh_ia;
533 		src = nh.nh_src;
534 		isbroadcast = (((nh.nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
535 		    (NHF_HOST | NHF_BROADCAST)) ||
536 		    ((ifp->if_flags & IFF_BROADCAST) &&
537 		    in_ifaddr_broadcast(dst->sin_addr, ia)));
538 	}
539 
540 	/* Catch a possible divide by zero later. */
541 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (rt_flags=0x%08x) ifp=%p",
542 	    __func__, mtu, ro,
543 	    (ro != NULL && ro->ro_rt != NULL) ? ro->ro_rt->rt_flags : 0, ifp));
544 
545 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
546 		m->m_flags |= M_MCAST;
547 		/*
548 		 * IP destination address is multicast.  Make sure "gw"
549 		 * still points to the address in "ro".  (It may have been
550 		 * changed to point to a gateway address, above.)
551 		 */
552 		gw = dst;
553 		/*
554 		 * See if the caller provided any multicast options
555 		 */
556 		if (imo != NULL) {
557 			ip->ip_ttl = imo->imo_multicast_ttl;
558 			if (imo->imo_multicast_vif != -1)
559 				ip->ip_src.s_addr =
560 				    ip_mcast_src ?
561 				    ip_mcast_src(imo->imo_multicast_vif) :
562 				    INADDR_ANY;
563 		} else
564 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
565 		/*
566 		 * Confirm that the outgoing interface supports multicast.
567 		 */
568 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
569 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
570 				IPSTAT_INC(ips_noroute);
571 				error = ENETUNREACH;
572 				goto bad;
573 			}
574 		}
575 		/*
576 		 * If source address not specified yet, use address
577 		 * of outgoing interface.
578 		 */
579 		if (ip->ip_src.s_addr == INADDR_ANY)
580 			ip->ip_src = src;
581 
582 		if ((imo == NULL && in_mcast_loop) ||
583 		    (imo && imo->imo_multicast_loop)) {
584 			/*
585 			 * Loop back multicast datagram if not expressly
586 			 * forbidden to do so, even if we are not a member
587 			 * of the group; ip_input() will filter it later,
588 			 * thus deferring a hash lookup and mutex acquisition
589 			 * at the expense of a cheap copy using m_copym().
590 			 */
591 			ip_mloopback(ifp, m, hlen);
592 		} else {
593 			/*
594 			 * If we are acting as a multicast router, perform
595 			 * multicast forwarding as if the packet had just
596 			 * arrived on the interface to which we are about
597 			 * to send.  The multicast forwarding function
598 			 * recursively calls this function, using the
599 			 * IP_FORWARDING flag to prevent infinite recursion.
600 			 *
601 			 * Multicasts that are looped back by ip_mloopback(),
602 			 * above, will be forwarded by the ip_input() routine,
603 			 * if necessary.
604 			 */
605 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
606 				/*
607 				 * If rsvp daemon is not running, do not
608 				 * set ip_moptions. This ensures that the packet
609 				 * is multicast and not just sent down one link
610 				 * as prescribed by rsvpd.
611 				 */
612 				if (!V_rsvp_on)
613 					imo = NULL;
614 				if (ip_mforward &&
615 				    ip_mforward(ip, ifp, m, imo) != 0) {
616 					m_freem(m);
617 					goto done;
618 				}
619 			}
620 		}
621 
622 		/*
623 		 * Multicasts with a time-to-live of zero may be looped-
624 		 * back, above, but must not be transmitted on a network.
625 		 * Also, multicasts addressed to the loopback interface
626 		 * are not sent -- the above call to ip_mloopback() will
627 		 * loop back a copy. ip_input() will drop the copy if
628 		 * this host does not belong to the destination group on
629 		 * the loopback interface.
630 		 */
631 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
632 			m_freem(m);
633 			goto done;
634 		}
635 
636 		goto sendit;
637 	}
638 
639 	/*
640 	 * If the source address is not specified yet, use the address
641 	 * of the outoing interface.
642 	 */
643 	if (ip->ip_src.s_addr == INADDR_ANY)
644 		ip->ip_src = src;
645 
646 	/*
647 	 * Look for broadcast address and
648 	 * verify user is allowed to send
649 	 * such a packet.
650 	 */
651 	if (isbroadcast) {
652 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
653 			error = EADDRNOTAVAIL;
654 			goto bad;
655 		}
656 		if ((flags & IP_ALLOWBROADCAST) == 0) {
657 			error = EACCES;
658 			goto bad;
659 		}
660 		/* don't allow broadcast messages to be fragmented */
661 		if (ip_len > mtu) {
662 			error = EMSGSIZE;
663 			goto bad;
664 		}
665 		m->m_flags |= M_BCAST;
666 	} else {
667 		m->m_flags &= ~M_BCAST;
668 	}
669 
670 sendit:
671 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
672 	if (IPSEC_ENABLED(ipv4)) {
673 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
674 			if (error == EINPROGRESS)
675 				error = 0;
676 			goto done;
677 		}
678 	}
679 	/*
680 	 * Check if there was a route for this packet; return error if not.
681 	 */
682 	if (no_route_but_check_spd) {
683 		IPSTAT_INC(ips_noroute);
684 		error = EHOSTUNREACH;
685 		goto bad;
686 	}
687 	/* Update variables that are affected by ipsec4_output(). */
688 	ip = mtod(m, struct ip *);
689 	hlen = ip->ip_hl << 2;
690 #endif /* IPSEC */
691 
692 	/* Jump over all PFIL processing if hooks are not active. */
693 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
694 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
695 		    &error)) {
696 		case 1: /* Finished */
697 			goto done;
698 
699 		case 0: /* Continue normally */
700 			ip = mtod(m, struct ip *);
701 			break;
702 
703 		case -1: /* Need to try again */
704 			/* Reset everything for a new round */
705 			if (ro != NULL) {
706 				RO_RTFREE(ro);
707 				ro->ro_prepend = NULL;
708 			}
709 			gw = dst;
710 			ip = mtod(m, struct ip *);
711 			goto again;
712 
713 		}
714 	}
715 
716 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
717 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
718 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
719 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
720 			IPSTAT_INC(ips_badaddr);
721 			error = EADDRNOTAVAIL;
722 			goto bad;
723 		}
724 	}
725 
726 	m->m_pkthdr.csum_flags |= CSUM_IP;
727 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
728 		m = mb_unmapped_to_ext(m);
729 		if (m == NULL) {
730 			IPSTAT_INC(ips_odropped);
731 			error = ENOBUFS;
732 			goto bad;
733 		}
734 		in_delayed_cksum(m);
735 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
736 	} else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
737 		m = mb_unmapped_to_ext(m);
738 		if (m == NULL) {
739 			IPSTAT_INC(ips_odropped);
740 			error = ENOBUFS;
741 			goto bad;
742 		}
743 	}
744 #ifdef SCTP
745 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
746 		m = mb_unmapped_to_ext(m);
747 		if (m == NULL) {
748 			IPSTAT_INC(ips_odropped);
749 			error = ENOBUFS;
750 			goto bad;
751 		}
752 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
753 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
754 	}
755 #endif
756 
757 	/*
758 	 * If small enough for interface, or the interface will take
759 	 * care of the fragmentation for us, we can just send directly.
760 	 */
761 	if (ip_len <= mtu ||
762 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
763 		ip->ip_sum = 0;
764 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
765 			ip->ip_sum = in_cksum(m, hlen);
766 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
767 		}
768 
769 		/*
770 		 * Record statistics for this interface address.
771 		 * With CSUM_TSO the byte/packet count will be slightly
772 		 * incorrect because we count the IP+TCP headers only
773 		 * once instead of for every generated packet.
774 		 */
775 		if (!(flags & IP_FORWARDING) && ia) {
776 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
777 				counter_u64_add(ia->ia_ifa.ifa_opackets,
778 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
779 			else
780 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
781 
782 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
783 		}
784 #ifdef MBUF_STRESS_TEST
785 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
786 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
787 #endif
788 		/*
789 		 * Reset layer specific mbuf flags
790 		 * to avoid confusing lower layers.
791 		 */
792 		m_clrprotoflags(m);
793 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
794 		error = ip_output_send(inp, ifp, m, gw, ro,
795 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
796 		goto done;
797 	}
798 
799 	/* Balk when DF bit is set or the interface didn't support TSO. */
800 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
801 		error = EMSGSIZE;
802 		IPSTAT_INC(ips_cantfrag);
803 		goto bad;
804 	}
805 
806 	/*
807 	 * Too large for interface; fragment if possible. If successful,
808 	 * on return, m will point to a list of packets to be sent.
809 	 */
810 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
811 	if (error)
812 		goto bad;
813 	for (; m; m = m0) {
814 		m0 = m->m_nextpkt;
815 		m->m_nextpkt = 0;
816 		if (error == 0) {
817 			/* Record statistics for this interface address. */
818 			if (ia != NULL) {
819 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
820 				counter_u64_add(ia->ia_ifa.ifa_obytes,
821 				    m->m_pkthdr.len);
822 			}
823 			/*
824 			 * Reset layer specific mbuf flags
825 			 * to avoid confusing upper layers.
826 			 */
827 			m_clrprotoflags(m);
828 
829 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
830 			    mtod(m, struct ip *), NULL);
831 			error = ip_output_send(inp, ifp, m, gw, ro, true);
832 		} else
833 			m_freem(m);
834 	}
835 
836 	if (error == 0)
837 		IPSTAT_INC(ips_fragmented);
838 
839 done:
840 	NET_EPOCH_EXIT(et);
841 	return (error);
842  bad:
843 	m_freem(m);
844 	goto done;
845 }
846 
847 /*
848  * Create a chain of fragments which fit the given mtu. m_frag points to the
849  * mbuf to be fragmented; on return it points to the chain with the fragments.
850  * Return 0 if no error. If error, m_frag may contain a partially built
851  * chain of fragments that should be freed by the caller.
852  *
853  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
854  */
855 int
856 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
857     u_long if_hwassist_flags)
858 {
859 	int error = 0;
860 	int hlen = ip->ip_hl << 2;
861 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
862 	int off;
863 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
864 	int firstlen;
865 	struct mbuf **mnext;
866 	int nfrags;
867 	uint16_t ip_len, ip_off;
868 
869 	ip_len = ntohs(ip->ip_len);
870 	ip_off = ntohs(ip->ip_off);
871 
872 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
873 		IPSTAT_INC(ips_cantfrag);
874 		return EMSGSIZE;
875 	}
876 
877 	/*
878 	 * Must be able to put at least 8 bytes per fragment.
879 	 */
880 	if (len < 8)
881 		return EMSGSIZE;
882 
883 	/*
884 	 * If the interface will not calculate checksums on
885 	 * fragmented packets, then do it here.
886 	 */
887 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
888 		m0 = mb_unmapped_to_ext(m0);
889 		if (m0 == NULL) {
890 			error = ENOBUFS;
891 			IPSTAT_INC(ips_odropped);
892 			goto done;
893 		}
894 		in_delayed_cksum(m0);
895 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
896 	}
897 #ifdef SCTP
898 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
899 		m0 = mb_unmapped_to_ext(m0);
900 		if (m0 == NULL) {
901 			error = ENOBUFS;
902 			IPSTAT_INC(ips_odropped);
903 			goto done;
904 		}
905 		sctp_delayed_cksum(m0, hlen);
906 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
907 	}
908 #endif
909 	if (len > PAGE_SIZE) {
910 		/*
911 		 * Fragment large datagrams such that each segment
912 		 * contains a multiple of PAGE_SIZE amount of data,
913 		 * plus headers. This enables a receiver to perform
914 		 * page-flipping zero-copy optimizations.
915 		 *
916 		 * XXX When does this help given that sender and receiver
917 		 * could have different page sizes, and also mtu could
918 		 * be less than the receiver's page size ?
919 		 */
920 		int newlen;
921 
922 		off = MIN(mtu, m0->m_pkthdr.len);
923 
924 		/*
925 		 * firstlen (off - hlen) must be aligned on an
926 		 * 8-byte boundary
927 		 */
928 		if (off < hlen)
929 			goto smart_frag_failure;
930 		off = ((off - hlen) & ~7) + hlen;
931 		newlen = (~PAGE_MASK) & mtu;
932 		if ((newlen + sizeof (struct ip)) > mtu) {
933 			/* we failed, go back the default */
934 smart_frag_failure:
935 			newlen = len;
936 			off = hlen + len;
937 		}
938 		len = newlen;
939 
940 	} else {
941 		off = hlen + len;
942 	}
943 
944 	firstlen = off - hlen;
945 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
946 
947 	/*
948 	 * Loop through length of segment after first fragment,
949 	 * make new header and copy data of each part and link onto chain.
950 	 * Here, m0 is the original packet, m is the fragment being created.
951 	 * The fragments are linked off the m_nextpkt of the original
952 	 * packet, which after processing serves as the first fragment.
953 	 */
954 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
955 		struct ip *mhip;	/* ip header on the fragment */
956 		struct mbuf *m;
957 		int mhlen = sizeof (struct ip);
958 
959 		m = m_gethdr(M_NOWAIT, MT_DATA);
960 		if (m == NULL) {
961 			error = ENOBUFS;
962 			IPSTAT_INC(ips_odropped);
963 			goto done;
964 		}
965 		/*
966 		 * Make sure the complete packet header gets copied
967 		 * from the originating mbuf to the newly created
968 		 * mbuf. This also ensures that existing firewall
969 		 * classification(s), VLAN tags and so on get copied
970 		 * to the resulting fragmented packet(s):
971 		 */
972 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
973 			m_free(m);
974 			error = ENOBUFS;
975 			IPSTAT_INC(ips_odropped);
976 			goto done;
977 		}
978 		/*
979 		 * In the first mbuf, leave room for the link header, then
980 		 * copy the original IP header including options. The payload
981 		 * goes into an additional mbuf chain returned by m_copym().
982 		 */
983 		m->m_data += max_linkhdr;
984 		mhip = mtod(m, struct ip *);
985 		*mhip = *ip;
986 		if (hlen > sizeof (struct ip)) {
987 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
988 			mhip->ip_v = IPVERSION;
989 			mhip->ip_hl = mhlen >> 2;
990 		}
991 		m->m_len = mhlen;
992 		/* XXX do we need to add ip_off below ? */
993 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
994 		if (off + len >= ip_len)
995 			len = ip_len - off;
996 		else
997 			mhip->ip_off |= IP_MF;
998 		mhip->ip_len = htons((u_short)(len + mhlen));
999 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1000 		if (m->m_next == NULL) {	/* copy failed */
1001 			m_free(m);
1002 			error = ENOBUFS;	/* ??? */
1003 			IPSTAT_INC(ips_odropped);
1004 			goto done;
1005 		}
1006 		m->m_pkthdr.len = mhlen + len;
1007 #ifdef MAC
1008 		mac_netinet_fragment(m0, m);
1009 #endif
1010 		mhip->ip_off = htons(mhip->ip_off);
1011 		mhip->ip_sum = 0;
1012 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1013 			mhip->ip_sum = in_cksum(m, mhlen);
1014 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1015 		}
1016 		*mnext = m;
1017 		mnext = &m->m_nextpkt;
1018 	}
1019 	IPSTAT_ADD(ips_ofragments, nfrags);
1020 
1021 	/*
1022 	 * Update first fragment by trimming what's been copied out
1023 	 * and updating header.
1024 	 */
1025 	m_adj(m0, hlen + firstlen - ip_len);
1026 	m0->m_pkthdr.len = hlen + firstlen;
1027 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1028 	ip->ip_off = htons(ip_off | IP_MF);
1029 	ip->ip_sum = 0;
1030 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1031 		ip->ip_sum = in_cksum(m0, hlen);
1032 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1033 	}
1034 
1035 done:
1036 	*m_frag = m0;
1037 	return error;
1038 }
1039 
1040 void
1041 in_delayed_cksum(struct mbuf *m)
1042 {
1043 	struct ip *ip;
1044 	struct udphdr *uh;
1045 	uint16_t cklen, csum, offset;
1046 
1047 	ip = mtod(m, struct ip *);
1048 	offset = ip->ip_hl << 2 ;
1049 
1050 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1051 		/* if udp header is not in the first mbuf copy udplen */
1052 		if (offset + sizeof(struct udphdr) > m->m_len) {
1053 			m_copydata(m, offset + offsetof(struct udphdr,
1054 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1055 			cklen = ntohs(cklen);
1056 		} else {
1057 			uh = (struct udphdr *)mtodo(m, offset);
1058 			cklen = ntohs(uh->uh_ulen);
1059 		}
1060 		csum = in_cksum_skip(m, cklen + offset, offset);
1061 		if (csum == 0)
1062 			csum = 0xffff;
1063 	} else {
1064 		cklen = ntohs(ip->ip_len);
1065 		csum = in_cksum_skip(m, cklen, offset);
1066 	}
1067 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1068 
1069 	if (offset + sizeof(csum) > m->m_len)
1070 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1071 	else
1072 		*(u_short *)mtodo(m, offset) = csum;
1073 }
1074 
1075 /*
1076  * IP socket option processing.
1077  */
1078 int
1079 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1080 {
1081 	struct	inpcb *inp = sotoinpcb(so);
1082 	int	error, optval;
1083 #ifdef	RSS
1084 	uint32_t rss_bucket;
1085 	int retval;
1086 #endif
1087 
1088 	error = optval = 0;
1089 	if (sopt->sopt_level != IPPROTO_IP) {
1090 		error = EINVAL;
1091 
1092 		if (sopt->sopt_level == SOL_SOCKET &&
1093 		    sopt->sopt_dir == SOPT_SET) {
1094 			switch (sopt->sopt_name) {
1095 			case SO_REUSEADDR:
1096 				INP_WLOCK(inp);
1097 				if ((so->so_options & SO_REUSEADDR) != 0)
1098 					inp->inp_flags2 |= INP_REUSEADDR;
1099 				else
1100 					inp->inp_flags2 &= ~INP_REUSEADDR;
1101 				INP_WUNLOCK(inp);
1102 				error = 0;
1103 				break;
1104 			case SO_REUSEPORT:
1105 				INP_WLOCK(inp);
1106 				if ((so->so_options & SO_REUSEPORT) != 0)
1107 					inp->inp_flags2 |= INP_REUSEPORT;
1108 				else
1109 					inp->inp_flags2 &= ~INP_REUSEPORT;
1110 				INP_WUNLOCK(inp);
1111 				error = 0;
1112 				break;
1113 			case SO_REUSEPORT_LB:
1114 				INP_WLOCK(inp);
1115 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1116 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1117 				else
1118 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1119 				INP_WUNLOCK(inp);
1120 				error = 0;
1121 				break;
1122 			case SO_SETFIB:
1123 				INP_WLOCK(inp);
1124 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1125 				INP_WUNLOCK(inp);
1126 				error = 0;
1127 				break;
1128 			case SO_MAX_PACING_RATE:
1129 #ifdef RATELIMIT
1130 				INP_WLOCK(inp);
1131 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1132 				INP_WUNLOCK(inp);
1133 				error = 0;
1134 #else
1135 				error = EOPNOTSUPP;
1136 #endif
1137 				break;
1138 			default:
1139 				break;
1140 			}
1141 		}
1142 		return (error);
1143 	}
1144 
1145 	switch (sopt->sopt_dir) {
1146 	case SOPT_SET:
1147 		switch (sopt->sopt_name) {
1148 		case IP_OPTIONS:
1149 #ifdef notyet
1150 		case IP_RETOPTS:
1151 #endif
1152 		{
1153 			struct mbuf *m;
1154 			if (sopt->sopt_valsize > MLEN) {
1155 				error = EMSGSIZE;
1156 				break;
1157 			}
1158 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1159 			if (m == NULL) {
1160 				error = ENOBUFS;
1161 				break;
1162 			}
1163 			m->m_len = sopt->sopt_valsize;
1164 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1165 					    m->m_len);
1166 			if (error) {
1167 				m_free(m);
1168 				break;
1169 			}
1170 			INP_WLOCK(inp);
1171 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1172 			INP_WUNLOCK(inp);
1173 			return (error);
1174 		}
1175 
1176 		case IP_BINDANY:
1177 			if (sopt->sopt_td != NULL) {
1178 				error = priv_check(sopt->sopt_td,
1179 				    PRIV_NETINET_BINDANY);
1180 				if (error)
1181 					break;
1182 			}
1183 			/* FALLTHROUGH */
1184 		case IP_BINDMULTI:
1185 #ifdef	RSS
1186 		case IP_RSS_LISTEN_BUCKET:
1187 #endif
1188 		case IP_TOS:
1189 		case IP_TTL:
1190 		case IP_MINTTL:
1191 		case IP_RECVOPTS:
1192 		case IP_RECVRETOPTS:
1193 		case IP_ORIGDSTADDR:
1194 		case IP_RECVDSTADDR:
1195 		case IP_RECVTTL:
1196 		case IP_RECVIF:
1197 		case IP_ONESBCAST:
1198 		case IP_DONTFRAG:
1199 		case IP_RECVTOS:
1200 		case IP_RECVFLOWID:
1201 #ifdef	RSS
1202 		case IP_RECVRSSBUCKETID:
1203 #endif
1204 			error = sooptcopyin(sopt, &optval, sizeof optval,
1205 					    sizeof optval);
1206 			if (error)
1207 				break;
1208 
1209 			switch (sopt->sopt_name) {
1210 			case IP_TOS:
1211 				inp->inp_ip_tos = optval;
1212 				break;
1213 
1214 			case IP_TTL:
1215 				inp->inp_ip_ttl = optval;
1216 				break;
1217 
1218 			case IP_MINTTL:
1219 				if (optval >= 0 && optval <= MAXTTL)
1220 					inp->inp_ip_minttl = optval;
1221 				else
1222 					error = EINVAL;
1223 				break;
1224 
1225 #define	OPTSET(bit) do {						\
1226 	INP_WLOCK(inp);							\
1227 	if (optval)							\
1228 		inp->inp_flags |= bit;					\
1229 	else								\
1230 		inp->inp_flags &= ~bit;					\
1231 	INP_WUNLOCK(inp);						\
1232 } while (0)
1233 
1234 #define	OPTSET2(bit, val) do {						\
1235 	INP_WLOCK(inp);							\
1236 	if (val)							\
1237 		inp->inp_flags2 |= bit;					\
1238 	else								\
1239 		inp->inp_flags2 &= ~bit;				\
1240 	INP_WUNLOCK(inp);						\
1241 } while (0)
1242 
1243 			case IP_RECVOPTS:
1244 				OPTSET(INP_RECVOPTS);
1245 				break;
1246 
1247 			case IP_RECVRETOPTS:
1248 				OPTSET(INP_RECVRETOPTS);
1249 				break;
1250 
1251 			case IP_RECVDSTADDR:
1252 				OPTSET(INP_RECVDSTADDR);
1253 				break;
1254 
1255 			case IP_ORIGDSTADDR:
1256 				OPTSET2(INP_ORIGDSTADDR, optval);
1257 				break;
1258 
1259 			case IP_RECVTTL:
1260 				OPTSET(INP_RECVTTL);
1261 				break;
1262 
1263 			case IP_RECVIF:
1264 				OPTSET(INP_RECVIF);
1265 				break;
1266 
1267 			case IP_ONESBCAST:
1268 				OPTSET(INP_ONESBCAST);
1269 				break;
1270 			case IP_DONTFRAG:
1271 				OPTSET(INP_DONTFRAG);
1272 				break;
1273 			case IP_BINDANY:
1274 				OPTSET(INP_BINDANY);
1275 				break;
1276 			case IP_RECVTOS:
1277 				OPTSET(INP_RECVTOS);
1278 				break;
1279 			case IP_BINDMULTI:
1280 				OPTSET2(INP_BINDMULTI, optval);
1281 				break;
1282 			case IP_RECVFLOWID:
1283 				OPTSET2(INP_RECVFLOWID, optval);
1284 				break;
1285 #ifdef	RSS
1286 			case IP_RSS_LISTEN_BUCKET:
1287 				if ((optval >= 0) &&
1288 				    (optval < rss_getnumbuckets())) {
1289 					inp->inp_rss_listen_bucket = optval;
1290 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1291 				} else {
1292 					error = EINVAL;
1293 				}
1294 				break;
1295 			case IP_RECVRSSBUCKETID:
1296 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1297 				break;
1298 #endif
1299 			}
1300 			break;
1301 #undef OPTSET
1302 #undef OPTSET2
1303 
1304 		/*
1305 		 * Multicast socket options are processed by the in_mcast
1306 		 * module.
1307 		 */
1308 		case IP_MULTICAST_IF:
1309 		case IP_MULTICAST_VIF:
1310 		case IP_MULTICAST_TTL:
1311 		case IP_MULTICAST_LOOP:
1312 		case IP_ADD_MEMBERSHIP:
1313 		case IP_DROP_MEMBERSHIP:
1314 		case IP_ADD_SOURCE_MEMBERSHIP:
1315 		case IP_DROP_SOURCE_MEMBERSHIP:
1316 		case IP_BLOCK_SOURCE:
1317 		case IP_UNBLOCK_SOURCE:
1318 		case IP_MSFILTER:
1319 		case MCAST_JOIN_GROUP:
1320 		case MCAST_LEAVE_GROUP:
1321 		case MCAST_JOIN_SOURCE_GROUP:
1322 		case MCAST_LEAVE_SOURCE_GROUP:
1323 		case MCAST_BLOCK_SOURCE:
1324 		case MCAST_UNBLOCK_SOURCE:
1325 			error = inp_setmoptions(inp, sopt);
1326 			break;
1327 
1328 		case IP_PORTRANGE:
1329 			error = sooptcopyin(sopt, &optval, sizeof optval,
1330 					    sizeof optval);
1331 			if (error)
1332 				break;
1333 
1334 			INP_WLOCK(inp);
1335 			switch (optval) {
1336 			case IP_PORTRANGE_DEFAULT:
1337 				inp->inp_flags &= ~(INP_LOWPORT);
1338 				inp->inp_flags &= ~(INP_HIGHPORT);
1339 				break;
1340 
1341 			case IP_PORTRANGE_HIGH:
1342 				inp->inp_flags &= ~(INP_LOWPORT);
1343 				inp->inp_flags |= INP_HIGHPORT;
1344 				break;
1345 
1346 			case IP_PORTRANGE_LOW:
1347 				inp->inp_flags &= ~(INP_HIGHPORT);
1348 				inp->inp_flags |= INP_LOWPORT;
1349 				break;
1350 
1351 			default:
1352 				error = EINVAL;
1353 				break;
1354 			}
1355 			INP_WUNLOCK(inp);
1356 			break;
1357 
1358 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1359 		case IP_IPSEC_POLICY:
1360 			if (IPSEC_ENABLED(ipv4)) {
1361 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1362 				break;
1363 			}
1364 			/* FALLTHROUGH */
1365 #endif /* IPSEC */
1366 
1367 		default:
1368 			error = ENOPROTOOPT;
1369 			break;
1370 		}
1371 		break;
1372 
1373 	case SOPT_GET:
1374 		switch (sopt->sopt_name) {
1375 		case IP_OPTIONS:
1376 		case IP_RETOPTS:
1377 			INP_RLOCK(inp);
1378 			if (inp->inp_options) {
1379 				struct mbuf *options;
1380 
1381 				options = m_copym(inp->inp_options, 0,
1382 				    M_COPYALL, M_NOWAIT);
1383 				INP_RUNLOCK(inp);
1384 				if (options != NULL) {
1385 					error = sooptcopyout(sopt,
1386 							     mtod(options, char *),
1387 							     options->m_len);
1388 					m_freem(options);
1389 				} else
1390 					error = ENOMEM;
1391 			} else {
1392 				INP_RUNLOCK(inp);
1393 				sopt->sopt_valsize = 0;
1394 			}
1395 			break;
1396 
1397 		case IP_TOS:
1398 		case IP_TTL:
1399 		case IP_MINTTL:
1400 		case IP_RECVOPTS:
1401 		case IP_RECVRETOPTS:
1402 		case IP_ORIGDSTADDR:
1403 		case IP_RECVDSTADDR:
1404 		case IP_RECVTTL:
1405 		case IP_RECVIF:
1406 		case IP_PORTRANGE:
1407 		case IP_ONESBCAST:
1408 		case IP_DONTFRAG:
1409 		case IP_BINDANY:
1410 		case IP_RECVTOS:
1411 		case IP_BINDMULTI:
1412 		case IP_FLOWID:
1413 		case IP_FLOWTYPE:
1414 		case IP_RECVFLOWID:
1415 #ifdef	RSS
1416 		case IP_RSSBUCKETID:
1417 		case IP_RECVRSSBUCKETID:
1418 #endif
1419 			switch (sopt->sopt_name) {
1420 
1421 			case IP_TOS:
1422 				optval = inp->inp_ip_tos;
1423 				break;
1424 
1425 			case IP_TTL:
1426 				optval = inp->inp_ip_ttl;
1427 				break;
1428 
1429 			case IP_MINTTL:
1430 				optval = inp->inp_ip_minttl;
1431 				break;
1432 
1433 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1434 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1435 
1436 			case IP_RECVOPTS:
1437 				optval = OPTBIT(INP_RECVOPTS);
1438 				break;
1439 
1440 			case IP_RECVRETOPTS:
1441 				optval = OPTBIT(INP_RECVRETOPTS);
1442 				break;
1443 
1444 			case IP_RECVDSTADDR:
1445 				optval = OPTBIT(INP_RECVDSTADDR);
1446 				break;
1447 
1448 			case IP_ORIGDSTADDR:
1449 				optval = OPTBIT2(INP_ORIGDSTADDR);
1450 				break;
1451 
1452 			case IP_RECVTTL:
1453 				optval = OPTBIT(INP_RECVTTL);
1454 				break;
1455 
1456 			case IP_RECVIF:
1457 				optval = OPTBIT(INP_RECVIF);
1458 				break;
1459 
1460 			case IP_PORTRANGE:
1461 				if (inp->inp_flags & INP_HIGHPORT)
1462 					optval = IP_PORTRANGE_HIGH;
1463 				else if (inp->inp_flags & INP_LOWPORT)
1464 					optval = IP_PORTRANGE_LOW;
1465 				else
1466 					optval = 0;
1467 				break;
1468 
1469 			case IP_ONESBCAST:
1470 				optval = OPTBIT(INP_ONESBCAST);
1471 				break;
1472 			case IP_DONTFRAG:
1473 				optval = OPTBIT(INP_DONTFRAG);
1474 				break;
1475 			case IP_BINDANY:
1476 				optval = OPTBIT(INP_BINDANY);
1477 				break;
1478 			case IP_RECVTOS:
1479 				optval = OPTBIT(INP_RECVTOS);
1480 				break;
1481 			case IP_FLOWID:
1482 				optval = inp->inp_flowid;
1483 				break;
1484 			case IP_FLOWTYPE:
1485 				optval = inp->inp_flowtype;
1486 				break;
1487 			case IP_RECVFLOWID:
1488 				optval = OPTBIT2(INP_RECVFLOWID);
1489 				break;
1490 #ifdef	RSS
1491 			case IP_RSSBUCKETID:
1492 				retval = rss_hash2bucket(inp->inp_flowid,
1493 				    inp->inp_flowtype,
1494 				    &rss_bucket);
1495 				if (retval == 0)
1496 					optval = rss_bucket;
1497 				else
1498 					error = EINVAL;
1499 				break;
1500 			case IP_RECVRSSBUCKETID:
1501 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1502 				break;
1503 #endif
1504 			case IP_BINDMULTI:
1505 				optval = OPTBIT2(INP_BINDMULTI);
1506 				break;
1507 			}
1508 			error = sooptcopyout(sopt, &optval, sizeof optval);
1509 			break;
1510 
1511 		/*
1512 		 * Multicast socket options are processed by the in_mcast
1513 		 * module.
1514 		 */
1515 		case IP_MULTICAST_IF:
1516 		case IP_MULTICAST_VIF:
1517 		case IP_MULTICAST_TTL:
1518 		case IP_MULTICAST_LOOP:
1519 		case IP_MSFILTER:
1520 			error = inp_getmoptions(inp, sopt);
1521 			break;
1522 
1523 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1524 		case IP_IPSEC_POLICY:
1525 			if (IPSEC_ENABLED(ipv4)) {
1526 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1527 				break;
1528 			}
1529 			/* FALLTHROUGH */
1530 #endif /* IPSEC */
1531 
1532 		default:
1533 			error = ENOPROTOOPT;
1534 			break;
1535 		}
1536 		break;
1537 	}
1538 	return (error);
1539 }
1540 
1541 /*
1542  * Routine called from ip_output() to loop back a copy of an IP multicast
1543  * packet to the input queue of a specified interface.  Note that this
1544  * calls the output routine of the loopback "driver", but with an interface
1545  * pointer that might NOT be a loopback interface -- evil, but easier than
1546  * replicating that code here.
1547  */
1548 static void
1549 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1550 {
1551 	struct ip *ip;
1552 	struct mbuf *copym;
1553 
1554 	/*
1555 	 * Make a deep copy of the packet because we're going to
1556 	 * modify the pack in order to generate checksums.
1557 	 */
1558 	copym = m_dup(m, M_NOWAIT);
1559 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1560 		copym = m_pullup(copym, hlen);
1561 	if (copym != NULL) {
1562 		/* If needed, compute the checksum and mark it as valid. */
1563 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1564 			in_delayed_cksum(copym);
1565 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1566 			copym->m_pkthdr.csum_flags |=
1567 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1568 			copym->m_pkthdr.csum_data = 0xffff;
1569 		}
1570 		/*
1571 		 * We don't bother to fragment if the IP length is greater
1572 		 * than the interface's MTU.  Can this possibly matter?
1573 		 */
1574 		ip = mtod(copym, struct ip *);
1575 		ip->ip_sum = 0;
1576 		ip->ip_sum = in_cksum(copym, hlen);
1577 		if_simloop(ifp, copym, AF_INET, 0);
1578 	}
1579 }
1580