xref: /freebsd/sys/netinet/ip_output.c (revision bdcbfde31e8e9b343f113a1956384bdf30d1ed62)
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 
32 #include <sys/cdefs.h>
33 #include "opt_inet.h"
34 #include "opt_ipsec.h"
35 #include "opt_kern_tls.h"
36 #include "opt_mbuf_stress_test.h"
37 #include "opt_ratelimit.h"
38 #include "opt_route.h"
39 #include "opt_rss.h"
40 #include "opt_sctp.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/ktls.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mbuf.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/protosw.h>
52 #include <sys/sdt.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sysctl.h>
56 #include <sys/ucred.h>
57 
58 #include <net/if.h>
59 #include <net/if_var.h>
60 #include <net/if_private.h>
61 #include <net/if_vlan_var.h>
62 #include <net/if_llatbl.h>
63 #include <net/ethernet.h>
64 #include <net/netisr.h>
65 #include <net/pfil.h>
66 #include <net/route.h>
67 #include <net/route/nhop.h>
68 #include <net/rss_config.h>
69 #include <net/vnet.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_fib.h>
73 #include <netinet/in_kdtrace.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/ip.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/in_rss.h>
79 #include <netinet/in_var.h>
80 #include <netinet/ip_var.h>
81 #include <netinet/ip_options.h>
82 
83 #include <netinet/udp.h>
84 #include <netinet/udp_var.h>
85 
86 #if defined(SCTP) || defined(SCTP_SUPPORT)
87 #include <netinet/sctp.h>
88 #include <netinet/sctp_crc32.h>
89 #endif
90 
91 #include <netipsec/ipsec_support.h>
92 
93 #include <machine/in_cksum.h>
94 
95 #include <security/mac/mac_framework.h>
96 
97 #ifdef MBUF_STRESS_TEST
98 static int mbuf_frag_size = 0;
99 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
100 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
101 #endif
102 
103 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
104 
105 extern int in_mcast_loop;
106 
107 static inline int
108 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
109     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
110 {
111 	struct m_tag *fwd_tag = NULL;
112 	struct mbuf *m;
113 	struct in_addr odst;
114 	struct ip *ip;
115 
116 	m = *mp;
117 	ip = mtod(m, struct ip *);
118 
119 	/* Run through list of hooks for output packets. */
120 	odst.s_addr = ip->ip_dst.s_addr;
121 	switch (pfil_mbuf_out(V_inet_pfil_head, mp, ifp, inp)) {
122 	case PFIL_DROPPED:
123 		*error = EACCES;
124 		/* FALLTHROUGH */
125 	case PFIL_CONSUMED:
126 		return 1; /* Finished */
127 	case PFIL_PASS:
128 		*error = 0;
129 	}
130 	m = *mp;
131 	ip = mtod(m, struct ip *);
132 
133 	/* See if destination IP address was changed by packet filter. */
134 	if (odst.s_addr != ip->ip_dst.s_addr) {
135 		m->m_flags |= M_SKIP_FIREWALL;
136 		/* If destination is now ourself drop to ip_input(). */
137 		if (in_localip(ip->ip_dst)) {
138 			m->m_flags |= M_FASTFWD_OURS;
139 			if (m->m_pkthdr.rcvif == NULL)
140 				m->m_pkthdr.rcvif = V_loif;
141 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
142 				m->m_pkthdr.csum_flags |=
143 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
144 				m->m_pkthdr.csum_data = 0xffff;
145 			}
146 			m->m_pkthdr.csum_flags |=
147 				CSUM_IP_CHECKED | CSUM_IP_VALID;
148 #if defined(SCTP) || defined(SCTP_SUPPORT)
149 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
150 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
151 #endif
152 			*error = netisr_queue(NETISR_IP, m);
153 			return 1; /* Finished */
154 		}
155 
156 		bzero(dst, sizeof(*dst));
157 		dst->sin_family = AF_INET;
158 		dst->sin_len = sizeof(*dst);
159 		dst->sin_addr = ip->ip_dst;
160 
161 		return -1; /* Reloop */
162 	}
163 	/* See if fib was changed by packet filter. */
164 	if ((*fibnum) != M_GETFIB(m)) {
165 		m->m_flags |= M_SKIP_FIREWALL;
166 		*fibnum = M_GETFIB(m);
167 		return -1; /* Reloop for FIB change */
168 	}
169 
170 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
171 	if (m->m_flags & M_FASTFWD_OURS) {
172 		if (m->m_pkthdr.rcvif == NULL)
173 			m->m_pkthdr.rcvif = V_loif;
174 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
175 			m->m_pkthdr.csum_flags |=
176 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
177 			m->m_pkthdr.csum_data = 0xffff;
178 		}
179 #if defined(SCTP) || defined(SCTP_SUPPORT)
180 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
181 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
182 #endif
183 		m->m_pkthdr.csum_flags |=
184 			CSUM_IP_CHECKED | CSUM_IP_VALID;
185 
186 		*error = netisr_queue(NETISR_IP, m);
187 		return 1; /* Finished */
188 	}
189 	/* Or forward to some other address? */
190 	if ((m->m_flags & M_IP_NEXTHOP) &&
191 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
192 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
193 		m->m_flags |= M_SKIP_FIREWALL;
194 		m->m_flags &= ~M_IP_NEXTHOP;
195 		m_tag_delete(m, fwd_tag);
196 
197 		return -1; /* Reloop for CHANGE of dst */
198 	}
199 
200 	return 0;
201 }
202 
203 static int
204 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
205     const struct sockaddr *gw, struct route *ro, bool stamp_tag)
206 {
207 #ifdef KERN_TLS
208 	struct ktls_session *tls = NULL;
209 #endif
210 	struct m_snd_tag *mst;
211 	int error;
212 
213 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
214 	mst = NULL;
215 
216 #ifdef KERN_TLS
217 	/*
218 	 * If this is an unencrypted TLS record, save a reference to
219 	 * the record.  This local reference is used to call
220 	 * ktls_output_eagain after the mbuf has been freed (thus
221 	 * dropping the mbuf's reference) in if_output.
222 	 */
223 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
224 		tls = ktls_hold(m->m_next->m_epg_tls);
225 		mst = tls->snd_tag;
226 
227 		/*
228 		 * If a TLS session doesn't have a valid tag, it must
229 		 * have had an earlier ifp mismatch, so drop this
230 		 * packet.
231 		 */
232 		if (mst == NULL) {
233 			m_freem(m);
234 			error = EAGAIN;
235 			goto done;
236 		}
237 		/*
238 		 * Always stamp tags that include NIC ktls.
239 		 */
240 		stamp_tag = true;
241 	}
242 #endif
243 #ifdef RATELIMIT
244 	if (inp != NULL && mst == NULL) {
245 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
246 		    (inp->inp_snd_tag != NULL &&
247 		    inp->inp_snd_tag->ifp != ifp))
248 			in_pcboutput_txrtlmt(inp, ifp, m);
249 
250 		if (inp->inp_snd_tag != NULL)
251 			mst = inp->inp_snd_tag;
252 	}
253 #endif
254 	if (stamp_tag && mst != NULL) {
255 		KASSERT(m->m_pkthdr.rcvif == NULL,
256 		    ("trying to add a send tag to a forwarded packet"));
257 		if (mst->ifp != ifp) {
258 			m_freem(m);
259 			error = EAGAIN;
260 			goto done;
261 		}
262 
263 		/* stamp send tag on mbuf */
264 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
265 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
266 	}
267 
268 	error = (*ifp->if_output)(ifp, m, gw, ro);
269 
270 done:
271 	/* Check for route change invalidating send tags. */
272 #ifdef KERN_TLS
273 	if (tls != NULL) {
274 		if (error == EAGAIN)
275 			error = ktls_output_eagain(inp, tls);
276 		ktls_free(tls);
277 	}
278 #endif
279 #ifdef RATELIMIT
280 	if (error == EAGAIN)
281 		in_pcboutput_eagain(inp);
282 #endif
283 	return (error);
284 }
285 
286 /* rte<>ro_flags translation */
287 static inline void
288 rt_update_ro_flags(struct route *ro, const struct nhop_object *nh)
289 {
290 	int nh_flags = nh->nh_flags;
291 
292 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
293 
294 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
295 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
296 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
297 }
298 
299 /*
300  * IP output.  The packet in mbuf chain m contains a skeletal IP
301  * header (with len, off, ttl, proto, tos, src, dst).
302  * The mbuf chain containing the packet will be freed.
303  * The mbuf opt, if present, will not be freed.
304  * If route ro is present and has ro_rt initialized, route lookup would be
305  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
306  * then result of route lookup is stored in ro->ro_rt.
307  *
308  * In the IP forwarding case, the packet will arrive with options already
309  * inserted, so must have a NULL opt pointer.
310  */
311 int
312 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
313     struct ip_moptions *imo, struct inpcb *inp)
314 {
315 	struct ip *ip;
316 	struct ifnet *ifp = NULL;	/* keep compiler happy */
317 	struct mbuf *m0;
318 	int hlen = sizeof (struct ip);
319 	int mtu = 0;
320 	int error = 0;
321 	int vlan_pcp = -1;
322 	struct sockaddr_in *dst;
323 	const struct sockaddr *gw;
324 	struct in_ifaddr *ia = NULL;
325 	struct in_addr src;
326 	int isbroadcast;
327 	uint16_t ip_len, ip_off;
328 	struct route iproute;
329 	uint32_t fibnum;
330 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
331 	int no_route_but_check_spd = 0;
332 #endif
333 
334 	M_ASSERTPKTHDR(m);
335 	NET_EPOCH_ASSERT();
336 
337 	if (inp != NULL) {
338 		INP_LOCK_ASSERT(inp);
339 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
340 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
341 			m->m_pkthdr.flowid = inp->inp_flowid;
342 			M_HASHTYPE_SET(m, inp->inp_flowtype);
343 		}
344 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
345 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
346 			    INP_2PCP_SHIFT;
347 #ifdef NUMA
348 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
349 #endif
350 	}
351 
352 	if (opt) {
353 		int len = 0;
354 		m = ip_insertoptions(m, opt, &len);
355 		if (len != 0)
356 			hlen = len; /* ip->ip_hl is updated above */
357 	}
358 	ip = mtod(m, struct ip *);
359 	ip_len = ntohs(ip->ip_len);
360 	ip_off = ntohs(ip->ip_off);
361 
362 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
363 		ip->ip_v = IPVERSION;
364 		ip->ip_hl = hlen >> 2;
365 		ip_fillid(ip);
366 	} else {
367 		/* Header already set, fetch hlen from there */
368 		hlen = ip->ip_hl << 2;
369 	}
370 	if ((flags & IP_FORWARDING) == 0)
371 		IPSTAT_INC(ips_localout);
372 
373 	/*
374 	 * dst/gw handling:
375 	 *
376 	 * gw is readonly but can point either to dst OR rt_gateway,
377 	 * therefore we need restore gw if we're redoing lookup.
378 	 */
379 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
380 	if (ro == NULL) {
381 		ro = &iproute;
382 		bzero(ro, sizeof (*ro));
383 	}
384 	dst = (struct sockaddr_in *)&ro->ro_dst;
385 	if (ro->ro_nh == NULL) {
386 		dst->sin_family = AF_INET;
387 		dst->sin_len = sizeof(*dst);
388 		dst->sin_addr = ip->ip_dst;
389 	}
390 	gw = (const struct sockaddr *)dst;
391 again:
392 	/*
393 	 * Validate route against routing table additions;
394 	 * a better/more specific route might have been added.
395 	 */
396 	if (inp != NULL && ro->ro_nh != NULL)
397 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
398 	/*
399 	 * If there is a cached route,
400 	 * check that it is to the same destination
401 	 * and is still up.  If not, free it and try again.
402 	 * The address family should also be checked in case of sharing the
403 	 * cache with IPv6.
404 	 * Also check whether routing cache needs invalidation.
405 	 */
406 	if (ro->ro_nh != NULL &&
407 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
408 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
409 		RO_INVALIDATE_CACHE(ro);
410 	ia = NULL;
411 	/*
412 	 * If routing to interface only, short circuit routing lookup.
413 	 * The use of an all-ones broadcast address implies this; an
414 	 * interface is specified by the broadcast address of an interface,
415 	 * or the destination address of a ptp interface.
416 	 */
417 	if (flags & IP_SENDONES) {
418 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
419 						      M_GETFIB(m)))) == NULL &&
420 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
421 						    M_GETFIB(m)))) == NULL) {
422 			IPSTAT_INC(ips_noroute);
423 			error = ENETUNREACH;
424 			goto bad;
425 		}
426 		ip->ip_dst.s_addr = INADDR_BROADCAST;
427 		dst->sin_addr = ip->ip_dst;
428 		ifp = ia->ia_ifp;
429 		mtu = ifp->if_mtu;
430 		ip->ip_ttl = 1;
431 		isbroadcast = 1;
432 		src = IA_SIN(ia)->sin_addr;
433 	} else if (flags & IP_ROUTETOIF) {
434 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
435 						    M_GETFIB(m)))) == NULL &&
436 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
437 						M_GETFIB(m)))) == NULL) {
438 			IPSTAT_INC(ips_noroute);
439 			error = ENETUNREACH;
440 			goto bad;
441 		}
442 		ifp = ia->ia_ifp;
443 		mtu = ifp->if_mtu;
444 		ip->ip_ttl = 1;
445 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
446 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
447 		src = IA_SIN(ia)->sin_addr;
448 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
449 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
450 		/*
451 		 * Bypass the normal routing lookup for multicast
452 		 * packets if the interface is specified.
453 		 */
454 		ifp = imo->imo_multicast_ifp;
455 		mtu = ifp->if_mtu;
456 		IFP_TO_IA(ifp, ia);
457 		isbroadcast = 0;	/* fool gcc */
458 		/* Interface may have no addresses. */
459 		if (ia != NULL)
460 			src = IA_SIN(ia)->sin_addr;
461 		else
462 			src.s_addr = INADDR_ANY;
463 	} else if (ro != &iproute) {
464 		if (ro->ro_nh == NULL) {
465 			/*
466 			 * We want to do any cloning requested by the link
467 			 * layer, as this is probably required in all cases
468 			 * for correct operation (as it is for ARP).
469 			 */
470 			uint32_t flowid;
471 			flowid = m->m_pkthdr.flowid;
472 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
473 			    NHR_REF, flowid);
474 
475 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
476 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
477 				/*
478 				 * There is no route for this packet, but it is
479 				 * possible that a matching SPD entry exists.
480 				 */
481 				no_route_but_check_spd = 1;
482 				goto sendit;
483 #endif
484 				IPSTAT_INC(ips_noroute);
485 				error = EHOSTUNREACH;
486 				goto bad;
487 			}
488 		}
489 		struct nhop_object *nh = ro->ro_nh;
490 
491 		ia = ifatoia(nh->nh_ifa);
492 		ifp = nh->nh_ifp;
493 		counter_u64_add(nh->nh_pksent, 1);
494 		rt_update_ro_flags(ro, nh);
495 		if (nh->nh_flags & NHF_GATEWAY)
496 			gw = &nh->gw_sa;
497 		if (nh->nh_flags & NHF_HOST)
498 			isbroadcast = (nh->nh_flags & NHF_BROADCAST);
499 		else if ((ifp->if_flags & IFF_BROADCAST) && (gw->sa_family == AF_INET))
500 			isbroadcast = in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia);
501 		else
502 			isbroadcast = 0;
503 		mtu = nh->nh_mtu;
504 		src = IA_SIN(ia)->sin_addr;
505 	} else {
506 		struct nhop_object *nh;
507 
508 		nh = fib4_lookup(M_GETFIB(m), dst->sin_addr, 0, NHR_NONE,
509 		    m->m_pkthdr.flowid);
510 		if (nh == NULL) {
511 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
512 			/*
513 			 * There is no route for this packet, but it is
514 			 * possible that a matching SPD entry exists.
515 			 */
516 			no_route_but_check_spd = 1;
517 			goto sendit;
518 #endif
519 			IPSTAT_INC(ips_noroute);
520 			error = EHOSTUNREACH;
521 			goto bad;
522 		}
523 		ifp = nh->nh_ifp;
524 		mtu = nh->nh_mtu;
525 		rt_update_ro_flags(ro, nh);
526 		if (nh->nh_flags & NHF_GATEWAY)
527 			gw = &nh->gw_sa;
528 		ia = ifatoia(nh->nh_ifa);
529 		src = IA_SIN(ia)->sin_addr;
530 		isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
531 		    (NHF_HOST | NHF_BROADCAST)) ||
532 		    ((ifp->if_flags & IFF_BROADCAST) &&
533 		    (gw->sa_family == AF_INET) &&
534 		    in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia)));
535 	}
536 
537 	/* Catch a possible divide by zero later. */
538 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
539 	    __func__, mtu, ro,
540 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
541 
542 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
543 		m->m_flags |= M_MCAST;
544 		/*
545 		 * IP destination address is multicast.  Make sure "gw"
546 		 * still points to the address in "ro".  (It may have been
547 		 * changed to point to a gateway address, above.)
548 		 */
549 		gw = (const struct sockaddr *)dst;
550 		/*
551 		 * See if the caller provided any multicast options
552 		 */
553 		if (imo != NULL) {
554 			ip->ip_ttl = imo->imo_multicast_ttl;
555 			if (imo->imo_multicast_vif != -1)
556 				ip->ip_src.s_addr =
557 				    ip_mcast_src ?
558 				    ip_mcast_src(imo->imo_multicast_vif) :
559 				    INADDR_ANY;
560 		} else
561 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
562 		/*
563 		 * Confirm that the outgoing interface supports multicast.
564 		 */
565 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
566 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
567 				IPSTAT_INC(ips_noroute);
568 				error = ENETUNREACH;
569 				goto bad;
570 			}
571 		}
572 		/*
573 		 * If source address not specified yet, use address
574 		 * of outgoing interface.
575 		 */
576 		if (ip->ip_src.s_addr == INADDR_ANY)
577 			ip->ip_src = src;
578 
579 		if ((imo == NULL && in_mcast_loop) ||
580 		    (imo && imo->imo_multicast_loop)) {
581 			/*
582 			 * Loop back multicast datagram if not expressly
583 			 * forbidden to do so, even if we are not a member
584 			 * of the group; ip_input() will filter it later,
585 			 * thus deferring a hash lookup and mutex acquisition
586 			 * at the expense of a cheap copy using m_copym().
587 			 */
588 			ip_mloopback(ifp, m, hlen);
589 		} else {
590 			/*
591 			 * If we are acting as a multicast router, perform
592 			 * multicast forwarding as if the packet had just
593 			 * arrived on the interface to which we are about
594 			 * to send.  The multicast forwarding function
595 			 * recursively calls this function, using the
596 			 * IP_FORWARDING flag to prevent infinite recursion.
597 			 *
598 			 * Multicasts that are looped back by ip_mloopback(),
599 			 * above, will be forwarded by the ip_input() routine,
600 			 * if necessary.
601 			 */
602 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
603 				/*
604 				 * If rsvp daemon is not running, do not
605 				 * set ip_moptions. This ensures that the packet
606 				 * is multicast and not just sent down one link
607 				 * as prescribed by rsvpd.
608 				 */
609 				if (!V_rsvp_on)
610 					imo = NULL;
611 				if (ip_mforward &&
612 				    ip_mforward(ip, ifp, m, imo) != 0) {
613 					m_freem(m);
614 					goto done;
615 				}
616 			}
617 		}
618 
619 		/*
620 		 * Multicasts with a time-to-live of zero may be looped-
621 		 * back, above, but must not be transmitted on a network.
622 		 * Also, multicasts addressed to the loopback interface
623 		 * are not sent -- the above call to ip_mloopback() will
624 		 * loop back a copy. ip_input() will drop the copy if
625 		 * this host does not belong to the destination group on
626 		 * the loopback interface.
627 		 */
628 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
629 			m_freem(m);
630 			goto done;
631 		}
632 
633 		goto sendit;
634 	}
635 
636 	/*
637 	 * If the source address is not specified yet, use the address
638 	 * of the outoing interface.
639 	 */
640 	if (ip->ip_src.s_addr == INADDR_ANY)
641 		ip->ip_src = src;
642 
643 	/*
644 	 * Look for broadcast address and
645 	 * verify user is allowed to send
646 	 * such a packet.
647 	 */
648 	if (isbroadcast) {
649 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
650 			error = EADDRNOTAVAIL;
651 			goto bad;
652 		}
653 		if ((flags & IP_ALLOWBROADCAST) == 0) {
654 			error = EACCES;
655 			goto bad;
656 		}
657 		/* don't allow broadcast messages to be fragmented */
658 		if (ip_len > mtu) {
659 			error = EMSGSIZE;
660 			goto bad;
661 		}
662 		m->m_flags |= M_BCAST;
663 	} else {
664 		m->m_flags &= ~M_BCAST;
665 	}
666 
667 sendit:
668 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
669 	if (IPSEC_ENABLED(ipv4)) {
670 		m = mb_unmapped_to_ext(m);
671 		if (m == NULL) {
672 			IPSTAT_INC(ips_odropped);
673 			error = ENOBUFS;
674 			goto bad;
675 		}
676 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
677 			if (error == EINPROGRESS)
678 				error = 0;
679 			goto done;
680 		}
681 	}
682 	/*
683 	 * Check if there was a route for this packet; return error if not.
684 	 */
685 	if (no_route_but_check_spd) {
686 		IPSTAT_INC(ips_noroute);
687 		error = EHOSTUNREACH;
688 		goto bad;
689 	}
690 	/* Update variables that are affected by ipsec4_output(). */
691 	ip = mtod(m, struct ip *);
692 	hlen = ip->ip_hl << 2;
693 #endif /* IPSEC */
694 
695 	/* Jump over all PFIL processing if hooks are not active. */
696 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
697 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
698 		    &error)) {
699 		case 1: /* Finished */
700 			goto done;
701 
702 		case 0: /* Continue normally */
703 			ip = mtod(m, struct ip *);
704 			ip_len = ntohs(ip->ip_len);
705 			break;
706 
707 		case -1: /* Need to try again */
708 			/* Reset everything for a new round */
709 			if (ro != NULL) {
710 				RO_NHFREE(ro);
711 				ro->ro_prepend = NULL;
712 			}
713 			gw = (const struct sockaddr *)dst;
714 			ip = mtod(m, struct ip *);
715 			goto again;
716 		}
717 	}
718 
719 	if (vlan_pcp > -1)
720 		EVL_APPLY_PRI(m, vlan_pcp);
721 
722 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
723 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
724 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
725 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
726 			IPSTAT_INC(ips_badaddr);
727 			error = EADDRNOTAVAIL;
728 			goto bad;
729 		}
730 	}
731 
732 	/* Ensure the packet data is mapped if the interface requires it. */
733 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
734 		m = mb_unmapped_to_ext(m);
735 		if (m == NULL) {
736 			IPSTAT_INC(ips_odropped);
737 			error = ENOBUFS;
738 			goto bad;
739 		}
740 	}
741 
742 	m->m_pkthdr.csum_flags |= CSUM_IP;
743 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
744 		in_delayed_cksum(m);
745 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
746 	}
747 #if defined(SCTP) || defined(SCTP_SUPPORT)
748 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
749 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
750 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
751 	}
752 #endif
753 
754 	/*
755 	 * If small enough for interface, or the interface will take
756 	 * care of the fragmentation for us, we can just send directly.
757 	 * Note that if_vxlan could have requested TSO even though the outer
758 	 * frame is UDP.  It is correct to not fragment such datagrams and
759 	 * instead just pass them on to the driver.
760 	 */
761 	if (ip_len <= mtu ||
762 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
763 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
764 		ip->ip_sum = 0;
765 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
766 			ip->ip_sum = in_cksum(m, hlen);
767 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
768 		}
769 
770 		/*
771 		 * Record statistics for this interface address.
772 		 * With CSUM_TSO the byte/packet count will be slightly
773 		 * incorrect because we count the IP+TCP headers only
774 		 * once instead of for every generated packet.
775 		 */
776 		if (!(flags & IP_FORWARDING) && ia) {
777 			if (m->m_pkthdr.csum_flags &
778 			    (CSUM_TSO | CSUM_INNER_TSO))
779 				counter_u64_add(ia->ia_ifa.ifa_opackets,
780 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
781 			else
782 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
783 
784 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
785 		}
786 #ifdef MBUF_STRESS_TEST
787 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
788 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
789 #endif
790 		/*
791 		 * Reset layer specific mbuf flags
792 		 * to avoid confusing lower layers.
793 		 */
794 		m_clrprotoflags(m);
795 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
796 		error = ip_output_send(inp, ifp, m, gw, ro,
797 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
798 		goto done;
799 	}
800 
801 	/* Balk when DF bit is set or the interface didn't support TSO. */
802 	if ((ip_off & IP_DF) ||
803 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
804 		error = EMSGSIZE;
805 		IPSTAT_INC(ips_cantfrag);
806 		goto bad;
807 	}
808 
809 	/*
810 	 * Too large for interface; fragment if possible. If successful,
811 	 * on return, m will point to a list of packets to be sent.
812 	 */
813 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
814 	if (error)
815 		goto bad;
816 	for (; m; m = m0) {
817 		m0 = m->m_nextpkt;
818 		m->m_nextpkt = 0;
819 		if (error == 0) {
820 			/* Record statistics for this interface address. */
821 			if (ia != NULL) {
822 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
823 				counter_u64_add(ia->ia_ifa.ifa_obytes,
824 				    m->m_pkthdr.len);
825 			}
826 			/*
827 			 * Reset layer specific mbuf flags
828 			 * to avoid confusing upper layers.
829 			 */
830 			m_clrprotoflags(m);
831 
832 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
833 			    mtod(m, struct ip *), NULL);
834 			error = ip_output_send(inp, ifp, m, gw, ro, true);
835 		} else
836 			m_freem(m);
837 	}
838 
839 	if (error == 0)
840 		IPSTAT_INC(ips_fragmented);
841 
842 done:
843 	return (error);
844  bad:
845 	m_freem(m);
846 	goto done;
847 }
848 
849 /*
850  * Create a chain of fragments which fit the given mtu. m_frag points to the
851  * mbuf to be fragmented; on return it points to the chain with the fragments.
852  * Return 0 if no error. If error, m_frag may contain a partially built
853  * chain of fragments that should be freed by the caller.
854  *
855  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
856  */
857 int
858 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
859     u_long if_hwassist_flags)
860 {
861 	int error = 0;
862 	int hlen = ip->ip_hl << 2;
863 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
864 	int off;
865 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
866 	int firstlen;
867 	struct mbuf **mnext;
868 	int nfrags;
869 	uint16_t ip_len, ip_off;
870 
871 	ip_len = ntohs(ip->ip_len);
872 	ip_off = ntohs(ip->ip_off);
873 
874 	/*
875 	 * Packet shall not have "Don't Fragment" flag and have at least 8
876 	 * bytes of payload.
877 	 */
878 	if (__predict_false((ip_off & IP_DF) || len < 8)) {
879 		IPSTAT_INC(ips_cantfrag);
880 		return (EMSGSIZE);
881 	}
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 		in_delayed_cksum(m0);
889 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
890 	}
891 #if defined(SCTP) || defined(SCTP_SUPPORT)
892 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
893 		sctp_delayed_cksum(m0, hlen);
894 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
895 	}
896 #endif
897 	if (len > PAGE_SIZE) {
898 		/*
899 		 * Fragment large datagrams such that each segment
900 		 * contains a multiple of PAGE_SIZE amount of data,
901 		 * plus headers. This enables a receiver to perform
902 		 * page-flipping zero-copy optimizations.
903 		 *
904 		 * XXX When does this help given that sender and receiver
905 		 * could have different page sizes, and also mtu could
906 		 * be less than the receiver's page size ?
907 		 */
908 		int newlen;
909 
910 		off = MIN(mtu, m0->m_pkthdr.len);
911 
912 		/*
913 		 * firstlen (off - hlen) must be aligned on an
914 		 * 8-byte boundary
915 		 */
916 		if (off < hlen)
917 			goto smart_frag_failure;
918 		off = ((off - hlen) & ~7) + hlen;
919 		newlen = (~PAGE_MASK) & mtu;
920 		if ((newlen + sizeof (struct ip)) > mtu) {
921 			/* we failed, go back the default */
922 smart_frag_failure:
923 			newlen = len;
924 			off = hlen + len;
925 		}
926 		len = newlen;
927 
928 	} else {
929 		off = hlen + len;
930 	}
931 
932 	firstlen = off - hlen;
933 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
934 
935 	/*
936 	 * Loop through length of segment after first fragment,
937 	 * make new header and copy data of each part and link onto chain.
938 	 * Here, m0 is the original packet, m is the fragment being created.
939 	 * The fragments are linked off the m_nextpkt of the original
940 	 * packet, which after processing serves as the first fragment.
941 	 */
942 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
943 		struct ip *mhip;	/* ip header on the fragment */
944 		struct mbuf *m;
945 		int mhlen = sizeof (struct ip);
946 
947 		m = m_gethdr(M_NOWAIT, MT_DATA);
948 		if (m == NULL) {
949 			error = ENOBUFS;
950 			IPSTAT_INC(ips_odropped);
951 			goto done;
952 		}
953 		/*
954 		 * Make sure the complete packet header gets copied
955 		 * from the originating mbuf to the newly created
956 		 * mbuf. This also ensures that existing firewall
957 		 * classification(s), VLAN tags and so on get copied
958 		 * to the resulting fragmented packet(s):
959 		 */
960 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
961 			m_free(m);
962 			error = ENOBUFS;
963 			IPSTAT_INC(ips_odropped);
964 			goto done;
965 		}
966 		/*
967 		 * In the first mbuf, leave room for the link header, then
968 		 * copy the original IP header including options. The payload
969 		 * goes into an additional mbuf chain returned by m_copym().
970 		 */
971 		m->m_data += max_linkhdr;
972 		mhip = mtod(m, struct ip *);
973 		*mhip = *ip;
974 		if (hlen > sizeof (struct ip)) {
975 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
976 			mhip->ip_v = IPVERSION;
977 			mhip->ip_hl = mhlen >> 2;
978 		}
979 		m->m_len = mhlen;
980 		/* XXX do we need to add ip_off below ? */
981 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
982 		if (off + len >= ip_len)
983 			len = ip_len - off;
984 		else
985 			mhip->ip_off |= IP_MF;
986 		mhip->ip_len = htons((u_short)(len + mhlen));
987 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
988 		if (m->m_next == NULL) {	/* copy failed */
989 			m_free(m);
990 			error = ENOBUFS;	/* ??? */
991 			IPSTAT_INC(ips_odropped);
992 			goto done;
993 		}
994 		m->m_pkthdr.len = mhlen + len;
995 #ifdef MAC
996 		mac_netinet_fragment(m0, m);
997 #endif
998 		mhip->ip_off = htons(mhip->ip_off);
999 		mhip->ip_sum = 0;
1000 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1001 			mhip->ip_sum = in_cksum(m, mhlen);
1002 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1003 		}
1004 		*mnext = m;
1005 		mnext = &m->m_nextpkt;
1006 	}
1007 	IPSTAT_ADD(ips_ofragments, nfrags);
1008 
1009 	/*
1010 	 * Update first fragment by trimming what's been copied out
1011 	 * and updating header.
1012 	 */
1013 	m_adj(m0, hlen + firstlen - ip_len);
1014 	m0->m_pkthdr.len = hlen + firstlen;
1015 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1016 	ip->ip_off = htons(ip_off | IP_MF);
1017 	ip->ip_sum = 0;
1018 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1019 		ip->ip_sum = in_cksum(m0, hlen);
1020 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1021 	}
1022 
1023 done:
1024 	*m_frag = m0;
1025 	return error;
1026 }
1027 
1028 void
1029 in_delayed_cksum(struct mbuf *m)
1030 {
1031 	struct ip *ip;
1032 	struct udphdr *uh;
1033 	uint16_t cklen, csum, offset;
1034 
1035 	ip = mtod(m, struct ip *);
1036 	offset = ip->ip_hl << 2 ;
1037 
1038 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1039 		/* if udp header is not in the first mbuf copy udplen */
1040 		if (offset + sizeof(struct udphdr) > m->m_len) {
1041 			m_copydata(m, offset + offsetof(struct udphdr,
1042 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1043 			cklen = ntohs(cklen);
1044 		} else {
1045 			uh = (struct udphdr *)mtodo(m, offset);
1046 			cklen = ntohs(uh->uh_ulen);
1047 		}
1048 		csum = in_cksum_skip(m, cklen + offset, offset);
1049 		if (csum == 0)
1050 			csum = 0xffff;
1051 	} else {
1052 		cklen = ntohs(ip->ip_len);
1053 		csum = in_cksum_skip(m, cklen, offset);
1054 	}
1055 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1056 
1057 	if (offset + sizeof(csum) > m->m_len)
1058 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1059 	else
1060 		*(u_short *)mtodo(m, offset) = csum;
1061 }
1062 
1063 /*
1064  * IP socket option processing.
1065  */
1066 int
1067 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1068 {
1069 	struct inpcb *inp = sotoinpcb(so);
1070 	int	error, optval;
1071 #ifdef	RSS
1072 	uint32_t rss_bucket;
1073 	int retval;
1074 #endif
1075 
1076 	error = optval = 0;
1077 	if (sopt->sopt_level != IPPROTO_IP) {
1078 		error = EINVAL;
1079 
1080 		if (sopt->sopt_level == SOL_SOCKET &&
1081 		    sopt->sopt_dir == SOPT_SET) {
1082 			switch (sopt->sopt_name) {
1083 			case SO_SETFIB:
1084 				INP_WLOCK(inp);
1085 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1086 				INP_WUNLOCK(inp);
1087 				error = 0;
1088 				break;
1089 			case SO_MAX_PACING_RATE:
1090 #ifdef RATELIMIT
1091 				INP_WLOCK(inp);
1092 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1093 				INP_WUNLOCK(inp);
1094 				error = 0;
1095 #else
1096 				error = EOPNOTSUPP;
1097 #endif
1098 				break;
1099 			default:
1100 				break;
1101 			}
1102 		}
1103 		return (error);
1104 	}
1105 
1106 	switch (sopt->sopt_dir) {
1107 	case SOPT_SET:
1108 		switch (sopt->sopt_name) {
1109 		case IP_OPTIONS:
1110 #ifdef notyet
1111 		case IP_RETOPTS:
1112 #endif
1113 		{
1114 			struct mbuf *m;
1115 			if (sopt->sopt_valsize > MLEN) {
1116 				error = EMSGSIZE;
1117 				break;
1118 			}
1119 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1120 			if (m == NULL) {
1121 				error = ENOBUFS;
1122 				break;
1123 			}
1124 			m->m_len = sopt->sopt_valsize;
1125 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1126 					    m->m_len);
1127 			if (error) {
1128 				m_free(m);
1129 				break;
1130 			}
1131 			INP_WLOCK(inp);
1132 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1133 			INP_WUNLOCK(inp);
1134 			return (error);
1135 		}
1136 
1137 		case IP_BINDANY:
1138 			if (sopt->sopt_td != NULL) {
1139 				error = priv_check(sopt->sopt_td,
1140 				    PRIV_NETINET_BINDANY);
1141 				if (error)
1142 					break;
1143 			}
1144 			/* FALLTHROUGH */
1145 		case IP_TOS:
1146 		case IP_TTL:
1147 		case IP_MINTTL:
1148 		case IP_RECVOPTS:
1149 		case IP_RECVRETOPTS:
1150 		case IP_ORIGDSTADDR:
1151 		case IP_RECVDSTADDR:
1152 		case IP_RECVTTL:
1153 		case IP_RECVIF:
1154 		case IP_ONESBCAST:
1155 		case IP_DONTFRAG:
1156 		case IP_RECVTOS:
1157 		case IP_RECVFLOWID:
1158 #ifdef	RSS
1159 		case IP_RECVRSSBUCKETID:
1160 #endif
1161 		case IP_VLAN_PCP:
1162 			error = sooptcopyin(sopt, &optval, sizeof optval,
1163 					    sizeof optval);
1164 			if (error)
1165 				break;
1166 
1167 			switch (sopt->sopt_name) {
1168 			case IP_TOS:
1169 				inp->inp_ip_tos = optval;
1170 				break;
1171 
1172 			case IP_TTL:
1173 				inp->inp_ip_ttl = optval;
1174 				break;
1175 
1176 			case IP_MINTTL:
1177 				if (optval >= 0 && optval <= MAXTTL)
1178 					inp->inp_ip_minttl = optval;
1179 				else
1180 					error = EINVAL;
1181 				break;
1182 
1183 #define	OPTSET(bit) do {						\
1184 	INP_WLOCK(inp);							\
1185 	if (optval)							\
1186 		inp->inp_flags |= bit;					\
1187 	else								\
1188 		inp->inp_flags &= ~bit;					\
1189 	INP_WUNLOCK(inp);						\
1190 } while (0)
1191 
1192 #define	OPTSET2(bit, val) do {						\
1193 	INP_WLOCK(inp);							\
1194 	if (val)							\
1195 		inp->inp_flags2 |= bit;					\
1196 	else								\
1197 		inp->inp_flags2 &= ~bit;				\
1198 	INP_WUNLOCK(inp);						\
1199 } while (0)
1200 
1201 			case IP_RECVOPTS:
1202 				OPTSET(INP_RECVOPTS);
1203 				break;
1204 
1205 			case IP_RECVRETOPTS:
1206 				OPTSET(INP_RECVRETOPTS);
1207 				break;
1208 
1209 			case IP_RECVDSTADDR:
1210 				OPTSET(INP_RECVDSTADDR);
1211 				break;
1212 
1213 			case IP_ORIGDSTADDR:
1214 				OPTSET2(INP_ORIGDSTADDR, optval);
1215 				break;
1216 
1217 			case IP_RECVTTL:
1218 				OPTSET(INP_RECVTTL);
1219 				break;
1220 
1221 			case IP_RECVIF:
1222 				OPTSET(INP_RECVIF);
1223 				break;
1224 
1225 			case IP_ONESBCAST:
1226 				OPTSET(INP_ONESBCAST);
1227 				break;
1228 			case IP_DONTFRAG:
1229 				OPTSET(INP_DONTFRAG);
1230 				break;
1231 			case IP_BINDANY:
1232 				OPTSET(INP_BINDANY);
1233 				break;
1234 			case IP_RECVTOS:
1235 				OPTSET(INP_RECVTOS);
1236 				break;
1237 			case IP_RECVFLOWID:
1238 				OPTSET2(INP_RECVFLOWID, optval);
1239 				break;
1240 #ifdef RSS
1241 			case IP_RECVRSSBUCKETID:
1242 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1243 				break;
1244 #endif
1245 			case IP_VLAN_PCP:
1246 				if ((optval >= -1) && (optval <=
1247 				    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1248 					if (optval == -1) {
1249 						INP_WLOCK(inp);
1250 						inp->inp_flags2 &=
1251 						    ~(INP_2PCP_SET |
1252 						      INP_2PCP_MASK);
1253 						INP_WUNLOCK(inp);
1254 					} else {
1255 						INP_WLOCK(inp);
1256 						inp->inp_flags2 |=
1257 						    INP_2PCP_SET;
1258 						inp->inp_flags2 &=
1259 						    ~INP_2PCP_MASK;
1260 						inp->inp_flags2 |=
1261 						    optval << INP_2PCP_SHIFT;
1262 						INP_WUNLOCK(inp);
1263 					}
1264 				} else
1265 					error = EINVAL;
1266 				break;
1267 			}
1268 			break;
1269 #undef OPTSET
1270 #undef OPTSET2
1271 
1272 		/*
1273 		 * Multicast socket options are processed by the in_mcast
1274 		 * module.
1275 		 */
1276 		case IP_MULTICAST_IF:
1277 		case IP_MULTICAST_VIF:
1278 		case IP_MULTICAST_TTL:
1279 		case IP_MULTICAST_LOOP:
1280 		case IP_ADD_MEMBERSHIP:
1281 		case IP_DROP_MEMBERSHIP:
1282 		case IP_ADD_SOURCE_MEMBERSHIP:
1283 		case IP_DROP_SOURCE_MEMBERSHIP:
1284 		case IP_BLOCK_SOURCE:
1285 		case IP_UNBLOCK_SOURCE:
1286 		case IP_MSFILTER:
1287 		case MCAST_JOIN_GROUP:
1288 		case MCAST_LEAVE_GROUP:
1289 		case MCAST_JOIN_SOURCE_GROUP:
1290 		case MCAST_LEAVE_SOURCE_GROUP:
1291 		case MCAST_BLOCK_SOURCE:
1292 		case MCAST_UNBLOCK_SOURCE:
1293 			error = inp_setmoptions(inp, sopt);
1294 			break;
1295 
1296 		case IP_PORTRANGE:
1297 			error = sooptcopyin(sopt, &optval, sizeof optval,
1298 					    sizeof optval);
1299 			if (error)
1300 				break;
1301 
1302 			INP_WLOCK(inp);
1303 			switch (optval) {
1304 			case IP_PORTRANGE_DEFAULT:
1305 				inp->inp_flags &= ~(INP_LOWPORT);
1306 				inp->inp_flags &= ~(INP_HIGHPORT);
1307 				break;
1308 
1309 			case IP_PORTRANGE_HIGH:
1310 				inp->inp_flags &= ~(INP_LOWPORT);
1311 				inp->inp_flags |= INP_HIGHPORT;
1312 				break;
1313 
1314 			case IP_PORTRANGE_LOW:
1315 				inp->inp_flags &= ~(INP_HIGHPORT);
1316 				inp->inp_flags |= INP_LOWPORT;
1317 				break;
1318 
1319 			default:
1320 				error = EINVAL;
1321 				break;
1322 			}
1323 			INP_WUNLOCK(inp);
1324 			break;
1325 
1326 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1327 		case IP_IPSEC_POLICY:
1328 			if (IPSEC_ENABLED(ipv4)) {
1329 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1330 				break;
1331 			}
1332 			/* FALLTHROUGH */
1333 #endif /* IPSEC */
1334 
1335 		default:
1336 			error = ENOPROTOOPT;
1337 			break;
1338 		}
1339 		break;
1340 
1341 	case SOPT_GET:
1342 		switch (sopt->sopt_name) {
1343 		case IP_OPTIONS:
1344 		case IP_RETOPTS:
1345 			INP_RLOCK(inp);
1346 			if (inp->inp_options) {
1347 				struct mbuf *options;
1348 
1349 				options = m_copym(inp->inp_options, 0,
1350 				    M_COPYALL, M_NOWAIT);
1351 				INP_RUNLOCK(inp);
1352 				if (options != NULL) {
1353 					error = sooptcopyout(sopt,
1354 							     mtod(options, char *),
1355 							     options->m_len);
1356 					m_freem(options);
1357 				} else
1358 					error = ENOMEM;
1359 			} else {
1360 				INP_RUNLOCK(inp);
1361 				sopt->sopt_valsize = 0;
1362 			}
1363 			break;
1364 
1365 		case IP_TOS:
1366 		case IP_TTL:
1367 		case IP_MINTTL:
1368 		case IP_RECVOPTS:
1369 		case IP_RECVRETOPTS:
1370 		case IP_ORIGDSTADDR:
1371 		case IP_RECVDSTADDR:
1372 		case IP_RECVTTL:
1373 		case IP_RECVIF:
1374 		case IP_PORTRANGE:
1375 		case IP_ONESBCAST:
1376 		case IP_DONTFRAG:
1377 		case IP_BINDANY:
1378 		case IP_RECVTOS:
1379 		case IP_FLOWID:
1380 		case IP_FLOWTYPE:
1381 		case IP_RECVFLOWID:
1382 #ifdef	RSS
1383 		case IP_RSSBUCKETID:
1384 		case IP_RECVRSSBUCKETID:
1385 #endif
1386 		case IP_VLAN_PCP:
1387 			switch (sopt->sopt_name) {
1388 			case IP_TOS:
1389 				optval = inp->inp_ip_tos;
1390 				break;
1391 
1392 			case IP_TTL:
1393 				optval = inp->inp_ip_ttl;
1394 				break;
1395 
1396 			case IP_MINTTL:
1397 				optval = inp->inp_ip_minttl;
1398 				break;
1399 
1400 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1401 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1402 
1403 			case IP_RECVOPTS:
1404 				optval = OPTBIT(INP_RECVOPTS);
1405 				break;
1406 
1407 			case IP_RECVRETOPTS:
1408 				optval = OPTBIT(INP_RECVRETOPTS);
1409 				break;
1410 
1411 			case IP_RECVDSTADDR:
1412 				optval = OPTBIT(INP_RECVDSTADDR);
1413 				break;
1414 
1415 			case IP_ORIGDSTADDR:
1416 				optval = OPTBIT2(INP_ORIGDSTADDR);
1417 				break;
1418 
1419 			case IP_RECVTTL:
1420 				optval = OPTBIT(INP_RECVTTL);
1421 				break;
1422 
1423 			case IP_RECVIF:
1424 				optval = OPTBIT(INP_RECVIF);
1425 				break;
1426 
1427 			case IP_PORTRANGE:
1428 				if (inp->inp_flags & INP_HIGHPORT)
1429 					optval = IP_PORTRANGE_HIGH;
1430 				else if (inp->inp_flags & INP_LOWPORT)
1431 					optval = IP_PORTRANGE_LOW;
1432 				else
1433 					optval = 0;
1434 				break;
1435 
1436 			case IP_ONESBCAST:
1437 				optval = OPTBIT(INP_ONESBCAST);
1438 				break;
1439 			case IP_DONTFRAG:
1440 				optval = OPTBIT(INP_DONTFRAG);
1441 				break;
1442 			case IP_BINDANY:
1443 				optval = OPTBIT(INP_BINDANY);
1444 				break;
1445 			case IP_RECVTOS:
1446 				optval = OPTBIT(INP_RECVTOS);
1447 				break;
1448 			case IP_FLOWID:
1449 				optval = inp->inp_flowid;
1450 				break;
1451 			case IP_FLOWTYPE:
1452 				optval = inp->inp_flowtype;
1453 				break;
1454 			case IP_RECVFLOWID:
1455 				optval = OPTBIT2(INP_RECVFLOWID);
1456 				break;
1457 #ifdef	RSS
1458 			case IP_RSSBUCKETID:
1459 				retval = rss_hash2bucket(inp->inp_flowid,
1460 				    inp->inp_flowtype,
1461 				    &rss_bucket);
1462 				if (retval == 0)
1463 					optval = rss_bucket;
1464 				else
1465 					error = EINVAL;
1466 				break;
1467 			case IP_RECVRSSBUCKETID:
1468 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1469 				break;
1470 #endif
1471 			case IP_VLAN_PCP:
1472 				if (OPTBIT2(INP_2PCP_SET)) {
1473 					optval = (inp->inp_flags2 &
1474 					    INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1475 				} else {
1476 					optval = -1;
1477 				}
1478 				break;
1479 			}
1480 			error = sooptcopyout(sopt, &optval, sizeof optval);
1481 			break;
1482 
1483 		/*
1484 		 * Multicast socket options are processed by the in_mcast
1485 		 * module.
1486 		 */
1487 		case IP_MULTICAST_IF:
1488 		case IP_MULTICAST_VIF:
1489 		case IP_MULTICAST_TTL:
1490 		case IP_MULTICAST_LOOP:
1491 		case IP_MSFILTER:
1492 			error = inp_getmoptions(inp, sopt);
1493 			break;
1494 
1495 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1496 		case IP_IPSEC_POLICY:
1497 			if (IPSEC_ENABLED(ipv4)) {
1498 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1499 				break;
1500 			}
1501 			/* FALLTHROUGH */
1502 #endif /* IPSEC */
1503 
1504 		default:
1505 			error = ENOPROTOOPT;
1506 			break;
1507 		}
1508 		break;
1509 	}
1510 	return (error);
1511 }
1512 
1513 /*
1514  * Routine called from ip_output() to loop back a copy of an IP multicast
1515  * packet to the input queue of a specified interface.  Note that this
1516  * calls the output routine of the loopback "driver", but with an interface
1517  * pointer that might NOT be a loopback interface -- evil, but easier than
1518  * replicating that code here.
1519  */
1520 static void
1521 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1522 {
1523 	struct ip *ip;
1524 	struct mbuf *copym;
1525 
1526 	/*
1527 	 * Make a deep copy of the packet because we're going to
1528 	 * modify the pack in order to generate checksums.
1529 	 */
1530 	copym = m_dup(m, M_NOWAIT);
1531 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1532 		copym = m_pullup(copym, hlen);
1533 	if (copym != NULL) {
1534 		/* If needed, compute the checksum and mark it as valid. */
1535 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1536 			in_delayed_cksum(copym);
1537 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1538 			copym->m_pkthdr.csum_flags |=
1539 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1540 			copym->m_pkthdr.csum_data = 0xffff;
1541 		}
1542 		/*
1543 		 * We don't bother to fragment if the IP length is greater
1544 		 * than the interface's MTU.  Can this possibly matter?
1545 		 */
1546 		ip = mtod(copym, struct ip *);
1547 		ip->ip_sum = 0;
1548 		ip->ip_sum = in_cksum(copym, hlen);
1549 		if_simloop(ifp, copym, AF_INET, 0);
1550 	}
1551 }
1552