xref: /freebsd/sys/netinet/ip_output.c (revision dc318a4ffabcbfa23bb56a33403aad36e6de30af)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1990, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_inet.h"
38 #include "opt_ipsec.h"
39 #include "opt_kern_tls.h"
40 #include "opt_mbuf_stress_test.h"
41 #include "opt_ratelimit.h"
42 #include "opt_route.h"
43 #include "opt_rss.h"
44 #include "opt_sctp.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
49 #include <sys/ktls.h>
50 #include <sys/lock.h>
51 #include <sys/malloc.h>
52 #include <sys/mbuf.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/protosw.h>
56 #include <sys/rmlock.h>
57 #include <sys/sdt.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
60 #include <sys/sysctl.h>
61 #include <sys/ucred.h>
62 
63 #include <net/if.h>
64 #include <net/if_var.h>
65 #include <net/if_vlan_var.h>
66 #include <net/if_llatbl.h>
67 #include <net/ethernet.h>
68 #include <net/netisr.h>
69 #include <net/pfil.h>
70 #include <net/route.h>
71 #include <net/route/nhop.h>
72 #include <net/rss_config.h>
73 #include <net/vnet.h>
74 
75 #include <netinet/in.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_kdtrace.h>
78 #include <netinet/in_systm.h>
79 #include <netinet/ip.h>
80 #include <netinet/in_fib.h>
81 #include <netinet/in_pcb.h>
82 #include <netinet/in_rss.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip_var.h>
85 #include <netinet/ip_options.h>
86 #include <netinet/ip_mroute.h>
87 
88 #include <netinet/udp.h>
89 #include <netinet/udp_var.h>
90 
91 #if defined(SCTP) || defined(SCTP_SUPPORT)
92 #include <netinet/sctp.h>
93 #include <netinet/sctp_crc32.h>
94 #endif
95 
96 #include <netipsec/ipsec_support.h>
97 
98 #include <machine/in_cksum.h>
99 
100 #include <security/mac/mac_framework.h>
101 
102 #ifdef MBUF_STRESS_TEST
103 static int mbuf_frag_size = 0;
104 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
105 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
106 #endif
107 
108 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
109 
110 extern int in_mcast_loop;
111 extern	struct protosw inetsw[];
112 
113 static inline int
114 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
115     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
116 {
117 	struct m_tag *fwd_tag = NULL;
118 	struct mbuf *m;
119 	struct in_addr odst;
120 	struct ip *ip;
121 	int pflags = PFIL_OUT;
122 
123 	if (flags & IP_FORWARDING)
124 		pflags |= PFIL_FWD;
125 
126 	m = *mp;
127 	ip = mtod(m, struct ip *);
128 
129 	/* Run through list of hooks for output packets. */
130 	odst.s_addr = ip->ip_dst.s_addr;
131 	switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
132 	case PFIL_DROPPED:
133 		*error = EACCES;
134 		/* FALLTHROUGH */
135 	case PFIL_CONSUMED:
136 		return 1; /* Finished */
137 	case PFIL_PASS:
138 		*error = 0;
139 	}
140 	m = *mp;
141 	ip = mtod(m, struct ip *);
142 
143 	/* See if destination IP address was changed by packet filter. */
144 	if (odst.s_addr != ip->ip_dst.s_addr) {
145 		m->m_flags |= M_SKIP_FIREWALL;
146 		/* If destination is now ourself drop to ip_input(). */
147 		if (in_localip(ip->ip_dst)) {
148 			m->m_flags |= M_FASTFWD_OURS;
149 			if (m->m_pkthdr.rcvif == NULL)
150 				m->m_pkthdr.rcvif = V_loif;
151 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
152 				m->m_pkthdr.csum_flags |=
153 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
154 				m->m_pkthdr.csum_data = 0xffff;
155 			}
156 			m->m_pkthdr.csum_flags |=
157 				CSUM_IP_CHECKED | CSUM_IP_VALID;
158 #if defined(SCTP) || defined(SCTP_SUPPORT)
159 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
160 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
161 #endif
162 			*error = netisr_queue(NETISR_IP, m);
163 			return 1; /* Finished */
164 		}
165 
166 		bzero(dst, sizeof(*dst));
167 		dst->sin_family = AF_INET;
168 		dst->sin_len = sizeof(*dst);
169 		dst->sin_addr = ip->ip_dst;
170 
171 		return -1; /* Reloop */
172 	}
173 	/* See if fib was changed by packet filter. */
174 	if ((*fibnum) != M_GETFIB(m)) {
175 		m->m_flags |= M_SKIP_FIREWALL;
176 		*fibnum = M_GETFIB(m);
177 		return -1; /* Reloop for FIB change */
178 	}
179 
180 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
181 	if (m->m_flags & M_FASTFWD_OURS) {
182 		if (m->m_pkthdr.rcvif == NULL)
183 			m->m_pkthdr.rcvif = V_loif;
184 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
185 			m->m_pkthdr.csum_flags |=
186 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
187 			m->m_pkthdr.csum_data = 0xffff;
188 		}
189 #if defined(SCTP) || defined(SCTP_SUPPORT)
190 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
191 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
192 #endif
193 		m->m_pkthdr.csum_flags |=
194 			CSUM_IP_CHECKED | CSUM_IP_VALID;
195 
196 		*error = netisr_queue(NETISR_IP, m);
197 		return 1; /* Finished */
198 	}
199 	/* Or forward to some other address? */
200 	if ((m->m_flags & M_IP_NEXTHOP) &&
201 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
202 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
203 		m->m_flags |= M_SKIP_FIREWALL;
204 		m->m_flags &= ~M_IP_NEXTHOP;
205 		m_tag_delete(m, fwd_tag);
206 
207 		return -1; /* Reloop for CHANGE of dst */
208 	}
209 
210 	return 0;
211 }
212 
213 static int
214 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
215     const struct sockaddr_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_epg_tls);
235 		mst = tls->snd_tag;
236 
237 		/*
238 		 * If a TLS session doesn't have a valid tag, it must
239 		 * have had an earlier ifp mismatch, so drop this
240 		 * packet.
241 		 */
242 		if (mst == NULL) {
243 			error = EAGAIN;
244 			goto done;
245 		}
246 		/*
247 		 * Always stamp tags that include NIC ktls.
248 		 */
249 		stamp_tag = true;
250 	}
251 #endif
252 #ifdef RATELIMIT
253 	if (inp != NULL && mst == NULL) {
254 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
255 		    (inp->inp_snd_tag != NULL &&
256 		    inp->inp_snd_tag->ifp != ifp))
257 			in_pcboutput_txrtlmt(inp, ifp, m);
258 
259 		if (inp->inp_snd_tag != NULL)
260 			mst = inp->inp_snd_tag;
261 	}
262 #endif
263 	if (stamp_tag && mst != NULL) {
264 		KASSERT(m->m_pkthdr.rcvif == NULL,
265 		    ("trying to add a send tag to a forwarded packet"));
266 		if (mst->ifp != ifp) {
267 			error = EAGAIN;
268 			goto done;
269 		}
270 
271 		/* stamp send tag on mbuf */
272 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
273 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
274 	}
275 
276 	error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
277 
278 done:
279 	/* Check for route change invalidating send tags. */
280 #ifdef KERN_TLS
281 	if (tls != NULL) {
282 		if (error == EAGAIN)
283 			error = ktls_output_eagain(inp, tls);
284 		ktls_free(tls);
285 	}
286 #endif
287 #ifdef RATELIMIT
288 	if (error == EAGAIN)
289 		in_pcboutput_eagain(inp);
290 #endif
291 	return (error);
292 }
293 
294 /* rte<>ro_flags translation */
295 static inline void
296 rt_update_ro_flags(struct route *ro)
297 {
298 	int nh_flags = ro->ro_nh->nh_flags;
299 
300 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
301 
302 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
303 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
304 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
305 }
306 
307 /*
308  * IP output.  The packet in mbuf chain m contains a skeletal IP
309  * header (with len, off, ttl, proto, tos, src, dst).
310  * The mbuf chain containing the packet will be freed.
311  * The mbuf opt, if present, will not be freed.
312  * If route ro is present and has ro_rt initialized, route lookup would be
313  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
314  * then result of route lookup is stored in ro->ro_rt.
315  *
316  * In the IP forwarding case, the packet will arrive with options already
317  * inserted, so must have a NULL opt pointer.
318  */
319 int
320 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
321     struct ip_moptions *imo, struct inpcb *inp)
322 {
323 	MROUTER_RLOCK_TRACKER;
324 	struct rm_priotracker in_ifa_tracker;
325 	struct ip *ip;
326 	struct ifnet *ifp = NULL;	/* keep compiler happy */
327 	struct mbuf *m0;
328 	int hlen = sizeof (struct ip);
329 	int mtu = 0;
330 	int error = 0;
331 	int vlan_pcp = -1;
332 	struct sockaddr_in *dst, sin;
333 	const struct sockaddr_in *gw;
334 	struct in_ifaddr *ia = NULL;
335 	struct in_addr src;
336 	int isbroadcast;
337 	uint16_t ip_len, ip_off;
338 	uint32_t fibnum;
339 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
340 	int no_route_but_check_spd = 0;
341 #endif
342 
343 	M_ASSERTPKTHDR(m);
344 	NET_EPOCH_ASSERT();
345 
346 	if (inp != NULL) {
347 		INP_LOCK_ASSERT(inp);
348 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
349 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
350 			m->m_pkthdr.flowid = inp->inp_flowid;
351 			M_HASHTYPE_SET(m, inp->inp_flowtype);
352 		}
353 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
354 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
355 			    INP_2PCP_SHIFT;
356 #ifdef NUMA
357 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
358 #endif
359 	}
360 
361 	if (opt) {
362 		int len = 0;
363 		m = ip_insertoptions(m, opt, &len);
364 		if (len != 0)
365 			hlen = len; /* ip->ip_hl is updated above */
366 	}
367 	ip = mtod(m, struct ip *);
368 	ip_len = ntohs(ip->ip_len);
369 	ip_off = ntohs(ip->ip_off);
370 
371 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
372 		ip->ip_v = IPVERSION;
373 		ip->ip_hl = hlen >> 2;
374 		ip_fillid(ip);
375 	} else {
376 		/* Header already set, fetch hlen from there */
377 		hlen = ip->ip_hl << 2;
378 	}
379 	if ((flags & IP_FORWARDING) == 0)
380 		IPSTAT_INC(ips_localout);
381 
382 	/*
383 	 * dst/gw handling:
384 	 *
385 	 * gw is readonly but can point either to dst OR rt_gateway,
386 	 * therefore we need restore gw if we're redoing lookup.
387 	 */
388 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
389 	if (ro != NULL)
390 		dst = (struct sockaddr_in *)&ro->ro_dst;
391 	else
392 		dst = &sin;
393 	if (ro == NULL || ro->ro_nh == NULL) {
394 		bzero(dst, sizeof(*dst));
395 		dst->sin_family = AF_INET;
396 		dst->sin_len = sizeof(*dst);
397 		dst->sin_addr = ip->ip_dst;
398 	}
399 	gw = dst;
400 again:
401 	/*
402 	 * Validate route against routing table additions;
403 	 * a better/more specific route might have been added.
404 	 */
405 	if (inp != NULL && ro != NULL && ro->ro_nh != NULL)
406 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
407 	/*
408 	 * If there is a cached route,
409 	 * check that it is to the same destination
410 	 * and is still up.  If not, free it and try again.
411 	 * The address family should also be checked in case of sharing the
412 	 * cache with IPv6.
413 	 * Also check whether routing cache needs invalidation.
414 	 */
415 	if (ro != NULL && ro->ro_nh != NULL &&
416 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
417 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
418 		RO_INVALIDATE_CACHE(ro);
419 	ia = NULL;
420 	/*
421 	 * If routing to interface only, short circuit routing lookup.
422 	 * The use of an all-ones broadcast address implies this; an
423 	 * interface is specified by the broadcast address of an interface,
424 	 * or the destination address of a ptp interface.
425 	 */
426 	if (flags & IP_SENDONES) {
427 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
428 						      M_GETFIB(m)))) == NULL &&
429 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
430 						    M_GETFIB(m)))) == NULL) {
431 			IPSTAT_INC(ips_noroute);
432 			error = ENETUNREACH;
433 			goto bad;
434 		}
435 		ip->ip_dst.s_addr = INADDR_BROADCAST;
436 		dst->sin_addr = ip->ip_dst;
437 		ifp = ia->ia_ifp;
438 		mtu = ifp->if_mtu;
439 		ip->ip_ttl = 1;
440 		isbroadcast = 1;
441 		src = IA_SIN(ia)->sin_addr;
442 	} else if (flags & IP_ROUTETOIF) {
443 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
444 						    M_GETFIB(m)))) == NULL &&
445 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
446 						M_GETFIB(m)))) == NULL) {
447 			IPSTAT_INC(ips_noroute);
448 			error = ENETUNREACH;
449 			goto bad;
450 		}
451 		ifp = ia->ia_ifp;
452 		mtu = ifp->if_mtu;
453 		ip->ip_ttl = 1;
454 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
455 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
456 		src = IA_SIN(ia)->sin_addr;
457 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
458 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
459 		/*
460 		 * Bypass the normal routing lookup for multicast
461 		 * packets if the interface is specified.
462 		 */
463 		ifp = imo->imo_multicast_ifp;
464 		mtu = ifp->if_mtu;
465 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
466 		isbroadcast = 0;	/* fool gcc */
467 		/* Interface may have no addresses. */
468 		if (ia != NULL)
469 			src = IA_SIN(ia)->sin_addr;
470 		else
471 			src.s_addr = INADDR_ANY;
472 	} else if (ro != NULL) {
473 		if (ro->ro_nh == NULL) {
474 			/*
475 			 * We want to do any cloning requested by the link
476 			 * layer, as this is probably required in all cases
477 			 * for correct operation (as it is for ARP).
478 			 */
479 			uint32_t flowid;
480 			flowid = m->m_pkthdr.flowid;
481 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
482 			    NHR_REF, flowid);
483 
484 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
485 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
486 				/*
487 				 * There is no route for this packet, but it is
488 				 * possible that a matching SPD entry exists.
489 				 */
490 				no_route_but_check_spd = 1;
491 				goto sendit;
492 #endif
493 				IPSTAT_INC(ips_noroute);
494 				error = EHOSTUNREACH;
495 				goto bad;
496 			}
497 		}
498 		ia = ifatoia(ro->ro_nh->nh_ifa);
499 		ifp = ro->ro_nh->nh_ifp;
500 		counter_u64_add(ro->ro_nh->nh_pksent, 1);
501 		rt_update_ro_flags(ro);
502 		if (ro->ro_nh->nh_flags & NHF_GATEWAY)
503 			gw = &ro->ro_nh->gw4_sa;
504 		if (ro->ro_nh->nh_flags & NHF_HOST)
505 			isbroadcast = (ro->ro_nh->nh_flags & NHF_BROADCAST);
506 		else if (ifp->if_flags & IFF_BROADCAST)
507 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
508 		else
509 			isbroadcast = 0;
510 		if (ro->ro_nh->nh_flags & NHF_HOST)
511 			mtu = ro->ro_nh->nh_mtu;
512 		else
513 			mtu = ifp->if_mtu;
514 		src = IA_SIN(ia)->sin_addr;
515 	} else {
516 		struct nhop_object *nh;
517 
518 		nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE,
519 		    m->m_pkthdr.flowid);
520 		if (nh == NULL) {
521 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
522 			/*
523 			 * There is no route for this packet, but it is
524 			 * possible that a matching SPD entry exists.
525 			 */
526 			no_route_but_check_spd = 1;
527 			goto sendit;
528 #endif
529 			IPSTAT_INC(ips_noroute);
530 			error = EHOSTUNREACH;
531 			goto bad;
532 		}
533 		ifp = nh->nh_ifp;
534 		mtu = nh->nh_mtu;
535 		/*
536 		 * We are rewriting here dst to be gw actually, contradicting
537 		 * comment at the beginning of the function. However, in this
538 		 * case we are always dealing with on stack dst.
539 		 * In case if pfil(9) sends us back to beginning of the
540 		 * function, the dst would be rewritten by ip_output_pfil().
541 		 */
542 		MPASS(dst == &sin);
543 		if (nh->nh_flags & NHF_GATEWAY)
544 			dst->sin_addr = nh->gw4_sa.sin_addr;
545 		ia = ifatoia(nh->nh_ifa);
546 		src = IA_SIN(ia)->sin_addr;
547 		isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
548 		    (NHF_HOST | NHF_BROADCAST)) ||
549 		    ((ifp->if_flags & IFF_BROADCAST) &&
550 		    in_ifaddr_broadcast(dst->sin_addr, ia)));
551 	}
552 
553 	/* Catch a possible divide by zero later. */
554 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
555 	    __func__, mtu, ro,
556 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
557 
558 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
559 		m->m_flags |= M_MCAST;
560 		/*
561 		 * IP destination address is multicast.  Make sure "gw"
562 		 * still points to the address in "ro".  (It may have been
563 		 * changed to point to a gateway address, above.)
564 		 */
565 		gw = dst;
566 		/*
567 		 * See if the caller provided any multicast options
568 		 */
569 		if (imo != NULL) {
570 			ip->ip_ttl = imo->imo_multicast_ttl;
571 			if (imo->imo_multicast_vif != -1)
572 				ip->ip_src.s_addr =
573 				    ip_mcast_src ?
574 				    ip_mcast_src(imo->imo_multicast_vif) :
575 				    INADDR_ANY;
576 		} else
577 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
578 		/*
579 		 * Confirm that the outgoing interface supports multicast.
580 		 */
581 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
582 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
583 				IPSTAT_INC(ips_noroute);
584 				error = ENETUNREACH;
585 				goto bad;
586 			}
587 		}
588 		/*
589 		 * If source address not specified yet, use address
590 		 * of outgoing interface.
591 		 */
592 		if (ip->ip_src.s_addr == INADDR_ANY)
593 			ip->ip_src = src;
594 
595 		if ((imo == NULL && in_mcast_loop) ||
596 		    (imo && imo->imo_multicast_loop)) {
597 			/*
598 			 * Loop back multicast datagram if not expressly
599 			 * forbidden to do so, even if we are not a member
600 			 * of the group; ip_input() will filter it later,
601 			 * thus deferring a hash lookup and mutex acquisition
602 			 * at the expense of a cheap copy using m_copym().
603 			 */
604 			ip_mloopback(ifp, m, hlen);
605 		} else {
606 			/*
607 			 * If we are acting as a multicast router, perform
608 			 * multicast forwarding as if the packet had just
609 			 * arrived on the interface to which we are about
610 			 * to send.  The multicast forwarding function
611 			 * recursively calls this function, using the
612 			 * IP_FORWARDING flag to prevent infinite recursion.
613 			 *
614 			 * Multicasts that are looped back by ip_mloopback(),
615 			 * above, will be forwarded by the ip_input() routine,
616 			 * if necessary.
617 			 */
618 			MROUTER_RLOCK();
619 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
620 				/*
621 				 * If rsvp daemon is not running, do not
622 				 * set ip_moptions. This ensures that the packet
623 				 * is multicast and not just sent down one link
624 				 * as prescribed by rsvpd.
625 				 */
626 				if (!V_rsvp_on)
627 					imo = NULL;
628 				if (ip_mforward &&
629 				    ip_mforward(ip, ifp, m, imo) != 0) {
630 					MROUTER_RUNLOCK();
631 					m_freem(m);
632 					goto done;
633 				}
634 			}
635 			MROUTER_RUNLOCK();
636 		}
637 
638 		/*
639 		 * Multicasts with a time-to-live of zero may be looped-
640 		 * back, above, but must not be transmitted on a network.
641 		 * Also, multicasts addressed to the loopback interface
642 		 * are not sent -- the above call to ip_mloopback() will
643 		 * loop back a copy. ip_input() will drop the copy if
644 		 * this host does not belong to the destination group on
645 		 * the loopback interface.
646 		 */
647 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
648 			m_freem(m);
649 			goto done;
650 		}
651 
652 		goto sendit;
653 	}
654 
655 	/*
656 	 * If the source address is not specified yet, use the address
657 	 * of the outoing interface.
658 	 */
659 	if (ip->ip_src.s_addr == INADDR_ANY)
660 		ip->ip_src = src;
661 
662 	/*
663 	 * Look for broadcast address and
664 	 * verify user is allowed to send
665 	 * such a packet.
666 	 */
667 	if (isbroadcast) {
668 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
669 			error = EADDRNOTAVAIL;
670 			goto bad;
671 		}
672 		if ((flags & IP_ALLOWBROADCAST) == 0) {
673 			error = EACCES;
674 			goto bad;
675 		}
676 		/* don't allow broadcast messages to be fragmented */
677 		if (ip_len > mtu) {
678 			error = EMSGSIZE;
679 			goto bad;
680 		}
681 		m->m_flags |= M_BCAST;
682 	} else {
683 		m->m_flags &= ~M_BCAST;
684 	}
685 
686 sendit:
687 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
688 	if (IPSEC_ENABLED(ipv4)) {
689 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
690 			if (error == EINPROGRESS)
691 				error = 0;
692 			goto done;
693 		}
694 	}
695 	/*
696 	 * Check if there was a route for this packet; return error if not.
697 	 */
698 	if (no_route_but_check_spd) {
699 		IPSTAT_INC(ips_noroute);
700 		error = EHOSTUNREACH;
701 		goto bad;
702 	}
703 	/* Update variables that are affected by ipsec4_output(). */
704 	ip = mtod(m, struct ip *);
705 	hlen = ip->ip_hl << 2;
706 #endif /* IPSEC */
707 
708 	/* Jump over all PFIL processing if hooks are not active. */
709 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
710 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
711 		    &error)) {
712 		case 1: /* Finished */
713 			goto done;
714 
715 		case 0: /* Continue normally */
716 			ip = mtod(m, struct ip *);
717 			break;
718 
719 		case -1: /* Need to try again */
720 			/* Reset everything for a new round */
721 			if (ro != NULL) {
722 				RO_NHFREE(ro);
723 				ro->ro_prepend = NULL;
724 			}
725 			gw = dst;
726 			ip = mtod(m, struct ip *);
727 			goto again;
728 		}
729 	}
730 
731 	if (vlan_pcp > -1)
732 		EVL_APPLY_PRI(m, vlan_pcp);
733 
734 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
735 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
736 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
737 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
738 			IPSTAT_INC(ips_badaddr);
739 			error = EADDRNOTAVAIL;
740 			goto bad;
741 		}
742 	}
743 
744 	m->m_pkthdr.csum_flags |= CSUM_IP;
745 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~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 		in_delayed_cksum(m);
753 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
754 	} else if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
755 		m = mb_unmapped_to_ext(m);
756 		if (m == NULL) {
757 			IPSTAT_INC(ips_odropped);
758 			error = ENOBUFS;
759 			goto bad;
760 		}
761 	}
762 #if defined(SCTP) || defined(SCTP_SUPPORT)
763 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
764 		m = mb_unmapped_to_ext(m);
765 		if (m == NULL) {
766 			IPSTAT_INC(ips_odropped);
767 			error = ENOBUFS;
768 			goto bad;
769 		}
770 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
771 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
772 	}
773 #endif
774 
775 	/*
776 	 * If small enough for interface, or the interface will take
777 	 * care of the fragmentation for us, we can just send directly.
778 	 * Note that if_vxlan could have requested TSO even though the outer
779 	 * frame is UDP.  It is correct to not fragment such datagrams and
780 	 * instead just pass them on to the driver.
781 	 */
782 	if (ip_len <= mtu ||
783 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
784 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
785 		ip->ip_sum = 0;
786 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
787 			ip->ip_sum = in_cksum(m, hlen);
788 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
789 		}
790 
791 		/*
792 		 * Record statistics for this interface address.
793 		 * With CSUM_TSO the byte/packet count will be slightly
794 		 * incorrect because we count the IP+TCP headers only
795 		 * once instead of for every generated packet.
796 		 */
797 		if (!(flags & IP_FORWARDING) && ia) {
798 			if (m->m_pkthdr.csum_flags &
799 			    (CSUM_TSO | CSUM_INNER_TSO))
800 				counter_u64_add(ia->ia_ifa.ifa_opackets,
801 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
802 			else
803 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
804 
805 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
806 		}
807 #ifdef MBUF_STRESS_TEST
808 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
809 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
810 #endif
811 		/*
812 		 * Reset layer specific mbuf flags
813 		 * to avoid confusing lower layers.
814 		 */
815 		m_clrprotoflags(m);
816 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
817 		error = ip_output_send(inp, ifp, m, gw, ro,
818 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
819 		goto done;
820 	}
821 
822 	/* Balk when DF bit is set or the interface didn't support TSO. */
823 	if ((ip_off & IP_DF) ||
824 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
825 		error = EMSGSIZE;
826 		IPSTAT_INC(ips_cantfrag);
827 		goto bad;
828 	}
829 
830 	/*
831 	 * Too large for interface; fragment if possible. If successful,
832 	 * on return, m will point to a list of packets to be sent.
833 	 */
834 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
835 	if (error)
836 		goto bad;
837 	for (; m; m = m0) {
838 		m0 = m->m_nextpkt;
839 		m->m_nextpkt = 0;
840 		if (error == 0) {
841 			/* Record statistics for this interface address. */
842 			if (ia != NULL) {
843 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
844 				counter_u64_add(ia->ia_ifa.ifa_obytes,
845 				    m->m_pkthdr.len);
846 			}
847 			/*
848 			 * Reset layer specific mbuf flags
849 			 * to avoid confusing upper layers.
850 			 */
851 			m_clrprotoflags(m);
852 
853 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
854 			    mtod(m, struct ip *), NULL);
855 			error = ip_output_send(inp, ifp, m, gw, ro, true);
856 		} else
857 			m_freem(m);
858 	}
859 
860 	if (error == 0)
861 		IPSTAT_INC(ips_fragmented);
862 
863 done:
864 	return (error);
865  bad:
866 	m_freem(m);
867 	goto done;
868 }
869 
870 /*
871  * Create a chain of fragments which fit the given mtu. m_frag points to the
872  * mbuf to be fragmented; on return it points to the chain with the fragments.
873  * Return 0 if no error. If error, m_frag may contain a partially built
874  * chain of fragments that should be freed by the caller.
875  *
876  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
877  */
878 int
879 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
880     u_long if_hwassist_flags)
881 {
882 	int error = 0;
883 	int hlen = ip->ip_hl << 2;
884 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
885 	int off;
886 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
887 	int firstlen;
888 	struct mbuf **mnext;
889 	int nfrags;
890 	uint16_t ip_len, ip_off;
891 
892 	ip_len = ntohs(ip->ip_len);
893 	ip_off = ntohs(ip->ip_off);
894 
895 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
896 		IPSTAT_INC(ips_cantfrag);
897 		return EMSGSIZE;
898 	}
899 
900 	/*
901 	 * Must be able to put at least 8 bytes per fragment.
902 	 */
903 	if (len < 8)
904 		return EMSGSIZE;
905 
906 	/*
907 	 * If the interface will not calculate checksums on
908 	 * fragmented packets, then do it here.
909 	 */
910 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
911 		m0 = mb_unmapped_to_ext(m0);
912 		if (m0 == NULL) {
913 			error = ENOBUFS;
914 			IPSTAT_INC(ips_odropped);
915 			goto done;
916 		}
917 		in_delayed_cksum(m0);
918 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
919 	}
920 #if defined(SCTP) || defined(SCTP_SUPPORT)
921 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
922 		m0 = mb_unmapped_to_ext(m0);
923 		if (m0 == NULL) {
924 			error = ENOBUFS;
925 			IPSTAT_INC(ips_odropped);
926 			goto done;
927 		}
928 		sctp_delayed_cksum(m0, hlen);
929 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
930 	}
931 #endif
932 	if (len > PAGE_SIZE) {
933 		/*
934 		 * Fragment large datagrams such that each segment
935 		 * contains a multiple of PAGE_SIZE amount of data,
936 		 * plus headers. This enables a receiver to perform
937 		 * page-flipping zero-copy optimizations.
938 		 *
939 		 * XXX When does this help given that sender and receiver
940 		 * could have different page sizes, and also mtu could
941 		 * be less than the receiver's page size ?
942 		 */
943 		int newlen;
944 
945 		off = MIN(mtu, m0->m_pkthdr.len);
946 
947 		/*
948 		 * firstlen (off - hlen) must be aligned on an
949 		 * 8-byte boundary
950 		 */
951 		if (off < hlen)
952 			goto smart_frag_failure;
953 		off = ((off - hlen) & ~7) + hlen;
954 		newlen = (~PAGE_MASK) & mtu;
955 		if ((newlen + sizeof (struct ip)) > mtu) {
956 			/* we failed, go back the default */
957 smart_frag_failure:
958 			newlen = len;
959 			off = hlen + len;
960 		}
961 		len = newlen;
962 
963 	} else {
964 		off = hlen + len;
965 	}
966 
967 	firstlen = off - hlen;
968 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
969 
970 	/*
971 	 * Loop through length of segment after first fragment,
972 	 * make new header and copy data of each part and link onto chain.
973 	 * Here, m0 is the original packet, m is the fragment being created.
974 	 * The fragments are linked off the m_nextpkt of the original
975 	 * packet, which after processing serves as the first fragment.
976 	 */
977 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
978 		struct ip *mhip;	/* ip header on the fragment */
979 		struct mbuf *m;
980 		int mhlen = sizeof (struct ip);
981 
982 		m = m_gethdr(M_NOWAIT, MT_DATA);
983 		if (m == NULL) {
984 			error = ENOBUFS;
985 			IPSTAT_INC(ips_odropped);
986 			goto done;
987 		}
988 		/*
989 		 * Make sure the complete packet header gets copied
990 		 * from the originating mbuf to the newly created
991 		 * mbuf. This also ensures that existing firewall
992 		 * classification(s), VLAN tags and so on get copied
993 		 * to the resulting fragmented packet(s):
994 		 */
995 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
996 			m_free(m);
997 			error = ENOBUFS;
998 			IPSTAT_INC(ips_odropped);
999 			goto done;
1000 		}
1001 		/*
1002 		 * In the first mbuf, leave room for the link header, then
1003 		 * copy the original IP header including options. The payload
1004 		 * goes into an additional mbuf chain returned by m_copym().
1005 		 */
1006 		m->m_data += max_linkhdr;
1007 		mhip = mtod(m, struct ip *);
1008 		*mhip = *ip;
1009 		if (hlen > sizeof (struct ip)) {
1010 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1011 			mhip->ip_v = IPVERSION;
1012 			mhip->ip_hl = mhlen >> 2;
1013 		}
1014 		m->m_len = mhlen;
1015 		/* XXX do we need to add ip_off below ? */
1016 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1017 		if (off + len >= ip_len)
1018 			len = ip_len - off;
1019 		else
1020 			mhip->ip_off |= IP_MF;
1021 		mhip->ip_len = htons((u_short)(len + mhlen));
1022 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1023 		if (m->m_next == NULL) {	/* copy failed */
1024 			m_free(m);
1025 			error = ENOBUFS;	/* ??? */
1026 			IPSTAT_INC(ips_odropped);
1027 			goto done;
1028 		}
1029 		m->m_pkthdr.len = mhlen + len;
1030 #ifdef MAC
1031 		mac_netinet_fragment(m0, m);
1032 #endif
1033 		mhip->ip_off = htons(mhip->ip_off);
1034 		mhip->ip_sum = 0;
1035 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1036 			mhip->ip_sum = in_cksum(m, mhlen);
1037 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1038 		}
1039 		*mnext = m;
1040 		mnext = &m->m_nextpkt;
1041 	}
1042 	IPSTAT_ADD(ips_ofragments, nfrags);
1043 
1044 	/*
1045 	 * Update first fragment by trimming what's been copied out
1046 	 * and updating header.
1047 	 */
1048 	m_adj(m0, hlen + firstlen - ip_len);
1049 	m0->m_pkthdr.len = hlen + firstlen;
1050 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1051 	ip->ip_off = htons(ip_off | IP_MF);
1052 	ip->ip_sum = 0;
1053 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1054 		ip->ip_sum = in_cksum(m0, hlen);
1055 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1056 	}
1057 
1058 done:
1059 	*m_frag = m0;
1060 	return error;
1061 }
1062 
1063 void
1064 in_delayed_cksum(struct mbuf *m)
1065 {
1066 	struct ip *ip;
1067 	struct udphdr *uh;
1068 	uint16_t cklen, csum, offset;
1069 
1070 	ip = mtod(m, struct ip *);
1071 	offset = ip->ip_hl << 2 ;
1072 
1073 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1074 		/* if udp header is not in the first mbuf copy udplen */
1075 		if (offset + sizeof(struct udphdr) > m->m_len) {
1076 			m_copydata(m, offset + offsetof(struct udphdr,
1077 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1078 			cklen = ntohs(cklen);
1079 		} else {
1080 			uh = (struct udphdr *)mtodo(m, offset);
1081 			cklen = ntohs(uh->uh_ulen);
1082 		}
1083 		csum = in_cksum_skip(m, cklen + offset, offset);
1084 		if (csum == 0)
1085 			csum = 0xffff;
1086 	} else {
1087 		cklen = ntohs(ip->ip_len);
1088 		csum = in_cksum_skip(m, cklen, offset);
1089 	}
1090 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1091 
1092 	if (offset + sizeof(csum) > m->m_len)
1093 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1094 	else
1095 		*(u_short *)mtodo(m, offset) = csum;
1096 }
1097 
1098 /*
1099  * IP socket option processing.
1100  */
1101 int
1102 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1103 {
1104 	struct inpcb *inp = sotoinpcb(so);
1105 	int	error, optval;
1106 #ifdef	RSS
1107 	uint32_t rss_bucket;
1108 	int retval;
1109 #endif
1110 
1111 	error = optval = 0;
1112 	if (sopt->sopt_level != IPPROTO_IP) {
1113 		error = EINVAL;
1114 
1115 		if (sopt->sopt_level == SOL_SOCKET &&
1116 		    sopt->sopt_dir == SOPT_SET) {
1117 			switch (sopt->sopt_name) {
1118 			case SO_REUSEADDR:
1119 				INP_WLOCK(inp);
1120 				if ((so->so_options & SO_REUSEADDR) != 0)
1121 					inp->inp_flags2 |= INP_REUSEADDR;
1122 				else
1123 					inp->inp_flags2 &= ~INP_REUSEADDR;
1124 				INP_WUNLOCK(inp);
1125 				error = 0;
1126 				break;
1127 			case SO_REUSEPORT:
1128 				INP_WLOCK(inp);
1129 				if ((so->so_options & SO_REUSEPORT) != 0)
1130 					inp->inp_flags2 |= INP_REUSEPORT;
1131 				else
1132 					inp->inp_flags2 &= ~INP_REUSEPORT;
1133 				INP_WUNLOCK(inp);
1134 				error = 0;
1135 				break;
1136 			case SO_REUSEPORT_LB:
1137 				INP_WLOCK(inp);
1138 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1139 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1140 				else
1141 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1142 				INP_WUNLOCK(inp);
1143 				error = 0;
1144 				break;
1145 			case SO_SETFIB:
1146 				INP_WLOCK(inp);
1147 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1148 				INP_WUNLOCK(inp);
1149 				error = 0;
1150 				break;
1151 			case SO_MAX_PACING_RATE:
1152 #ifdef RATELIMIT
1153 				INP_WLOCK(inp);
1154 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1155 				INP_WUNLOCK(inp);
1156 				error = 0;
1157 #else
1158 				error = EOPNOTSUPP;
1159 #endif
1160 				break;
1161 			default:
1162 				break;
1163 			}
1164 		}
1165 		return (error);
1166 	}
1167 
1168 	switch (sopt->sopt_dir) {
1169 	case SOPT_SET:
1170 		switch (sopt->sopt_name) {
1171 		case IP_OPTIONS:
1172 #ifdef notyet
1173 		case IP_RETOPTS:
1174 #endif
1175 		{
1176 			struct mbuf *m;
1177 			if (sopt->sopt_valsize > MLEN) {
1178 				error = EMSGSIZE;
1179 				break;
1180 			}
1181 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1182 			if (m == NULL) {
1183 				error = ENOBUFS;
1184 				break;
1185 			}
1186 			m->m_len = sopt->sopt_valsize;
1187 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1188 					    m->m_len);
1189 			if (error) {
1190 				m_free(m);
1191 				break;
1192 			}
1193 			INP_WLOCK(inp);
1194 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1195 			INP_WUNLOCK(inp);
1196 			return (error);
1197 		}
1198 
1199 		case IP_BINDANY:
1200 			if (sopt->sopt_td != NULL) {
1201 				error = priv_check(sopt->sopt_td,
1202 				    PRIV_NETINET_BINDANY);
1203 				if (error)
1204 					break;
1205 			}
1206 			/* FALLTHROUGH */
1207 		case IP_BINDMULTI:
1208 #ifdef	RSS
1209 		case IP_RSS_LISTEN_BUCKET:
1210 #endif
1211 		case IP_TOS:
1212 		case IP_TTL:
1213 		case IP_MINTTL:
1214 		case IP_RECVOPTS:
1215 		case IP_RECVRETOPTS:
1216 		case IP_ORIGDSTADDR:
1217 		case IP_RECVDSTADDR:
1218 		case IP_RECVTTL:
1219 		case IP_RECVIF:
1220 		case IP_ONESBCAST:
1221 		case IP_DONTFRAG:
1222 		case IP_RECVTOS:
1223 		case IP_RECVFLOWID:
1224 #ifdef	RSS
1225 		case IP_RECVRSSBUCKETID:
1226 #endif
1227 		case IP_VLAN_PCP:
1228 			error = sooptcopyin(sopt, &optval, sizeof optval,
1229 					    sizeof optval);
1230 			if (error)
1231 				break;
1232 
1233 			switch (sopt->sopt_name) {
1234 			case IP_TOS:
1235 				inp->inp_ip_tos = optval;
1236 				break;
1237 
1238 			case IP_TTL:
1239 				inp->inp_ip_ttl = optval;
1240 				break;
1241 
1242 			case IP_MINTTL:
1243 				if (optval >= 0 && optval <= MAXTTL)
1244 					inp->inp_ip_minttl = optval;
1245 				else
1246 					error = EINVAL;
1247 				break;
1248 
1249 #define	OPTSET(bit) do {						\
1250 	INP_WLOCK(inp);							\
1251 	if (optval)							\
1252 		inp->inp_flags |= bit;					\
1253 	else								\
1254 		inp->inp_flags &= ~bit;					\
1255 	INP_WUNLOCK(inp);						\
1256 } while (0)
1257 
1258 #define	OPTSET2(bit, val) do {						\
1259 	INP_WLOCK(inp);							\
1260 	if (val)							\
1261 		inp->inp_flags2 |= bit;					\
1262 	else								\
1263 		inp->inp_flags2 &= ~bit;				\
1264 	INP_WUNLOCK(inp);						\
1265 } while (0)
1266 
1267 			case IP_RECVOPTS:
1268 				OPTSET(INP_RECVOPTS);
1269 				break;
1270 
1271 			case IP_RECVRETOPTS:
1272 				OPTSET(INP_RECVRETOPTS);
1273 				break;
1274 
1275 			case IP_RECVDSTADDR:
1276 				OPTSET(INP_RECVDSTADDR);
1277 				break;
1278 
1279 			case IP_ORIGDSTADDR:
1280 				OPTSET2(INP_ORIGDSTADDR, optval);
1281 				break;
1282 
1283 			case IP_RECVTTL:
1284 				OPTSET(INP_RECVTTL);
1285 				break;
1286 
1287 			case IP_RECVIF:
1288 				OPTSET(INP_RECVIF);
1289 				break;
1290 
1291 			case IP_ONESBCAST:
1292 				OPTSET(INP_ONESBCAST);
1293 				break;
1294 			case IP_DONTFRAG:
1295 				OPTSET(INP_DONTFRAG);
1296 				break;
1297 			case IP_BINDANY:
1298 				OPTSET(INP_BINDANY);
1299 				break;
1300 			case IP_RECVTOS:
1301 				OPTSET(INP_RECVTOS);
1302 				break;
1303 			case IP_BINDMULTI:
1304 				OPTSET2(INP_BINDMULTI, optval);
1305 				break;
1306 			case IP_RECVFLOWID:
1307 				OPTSET2(INP_RECVFLOWID, optval);
1308 				break;
1309 #ifdef	RSS
1310 			case IP_RSS_LISTEN_BUCKET:
1311 				if ((optval >= 0) &&
1312 				    (optval < rss_getnumbuckets())) {
1313 					inp->inp_rss_listen_bucket = optval;
1314 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1315 				} else {
1316 					error = EINVAL;
1317 				}
1318 				break;
1319 			case IP_RECVRSSBUCKETID:
1320 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1321 				break;
1322 #endif
1323 			case IP_VLAN_PCP:
1324 				if ((optval >= -1) && (optval <=
1325 				    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1326 					if (optval == -1) {
1327 						INP_WLOCK(inp);
1328 						inp->inp_flags2 &=
1329 						    ~(INP_2PCP_SET |
1330 						      INP_2PCP_MASK);
1331 						INP_WUNLOCK(inp);
1332 					} else {
1333 						INP_WLOCK(inp);
1334 						inp->inp_flags2 |=
1335 						    INP_2PCP_SET;
1336 						inp->inp_flags2 &=
1337 						    ~INP_2PCP_MASK;
1338 						inp->inp_flags2 |=
1339 						    optval << INP_2PCP_SHIFT;
1340 						INP_WUNLOCK(inp);
1341 					}
1342 				} else
1343 					error = EINVAL;
1344 				break;
1345 			}
1346 			break;
1347 #undef OPTSET
1348 #undef OPTSET2
1349 
1350 		/*
1351 		 * Multicast socket options are processed by the in_mcast
1352 		 * module.
1353 		 */
1354 		case IP_MULTICAST_IF:
1355 		case IP_MULTICAST_VIF:
1356 		case IP_MULTICAST_TTL:
1357 		case IP_MULTICAST_LOOP:
1358 		case IP_ADD_MEMBERSHIP:
1359 		case IP_DROP_MEMBERSHIP:
1360 		case IP_ADD_SOURCE_MEMBERSHIP:
1361 		case IP_DROP_SOURCE_MEMBERSHIP:
1362 		case IP_BLOCK_SOURCE:
1363 		case IP_UNBLOCK_SOURCE:
1364 		case IP_MSFILTER:
1365 		case MCAST_JOIN_GROUP:
1366 		case MCAST_LEAVE_GROUP:
1367 		case MCAST_JOIN_SOURCE_GROUP:
1368 		case MCAST_LEAVE_SOURCE_GROUP:
1369 		case MCAST_BLOCK_SOURCE:
1370 		case MCAST_UNBLOCK_SOURCE:
1371 			error = inp_setmoptions(inp, sopt);
1372 			break;
1373 
1374 		case IP_PORTRANGE:
1375 			error = sooptcopyin(sopt, &optval, sizeof optval,
1376 					    sizeof optval);
1377 			if (error)
1378 				break;
1379 
1380 			INP_WLOCK(inp);
1381 			switch (optval) {
1382 			case IP_PORTRANGE_DEFAULT:
1383 				inp->inp_flags &= ~(INP_LOWPORT);
1384 				inp->inp_flags &= ~(INP_HIGHPORT);
1385 				break;
1386 
1387 			case IP_PORTRANGE_HIGH:
1388 				inp->inp_flags &= ~(INP_LOWPORT);
1389 				inp->inp_flags |= INP_HIGHPORT;
1390 				break;
1391 
1392 			case IP_PORTRANGE_LOW:
1393 				inp->inp_flags &= ~(INP_HIGHPORT);
1394 				inp->inp_flags |= INP_LOWPORT;
1395 				break;
1396 
1397 			default:
1398 				error = EINVAL;
1399 				break;
1400 			}
1401 			INP_WUNLOCK(inp);
1402 			break;
1403 
1404 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1405 		case IP_IPSEC_POLICY:
1406 			if (IPSEC_ENABLED(ipv4)) {
1407 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1408 				break;
1409 			}
1410 			/* FALLTHROUGH */
1411 #endif /* IPSEC */
1412 
1413 		default:
1414 			error = ENOPROTOOPT;
1415 			break;
1416 		}
1417 		break;
1418 
1419 	case SOPT_GET:
1420 		switch (sopt->sopt_name) {
1421 		case IP_OPTIONS:
1422 		case IP_RETOPTS:
1423 			INP_RLOCK(inp);
1424 			if (inp->inp_options) {
1425 				struct mbuf *options;
1426 
1427 				options = m_copym(inp->inp_options, 0,
1428 				    M_COPYALL, M_NOWAIT);
1429 				INP_RUNLOCK(inp);
1430 				if (options != NULL) {
1431 					error = sooptcopyout(sopt,
1432 							     mtod(options, char *),
1433 							     options->m_len);
1434 					m_freem(options);
1435 				} else
1436 					error = ENOMEM;
1437 			} else {
1438 				INP_RUNLOCK(inp);
1439 				sopt->sopt_valsize = 0;
1440 			}
1441 			break;
1442 
1443 		case IP_TOS:
1444 		case IP_TTL:
1445 		case IP_MINTTL:
1446 		case IP_RECVOPTS:
1447 		case IP_RECVRETOPTS:
1448 		case IP_ORIGDSTADDR:
1449 		case IP_RECVDSTADDR:
1450 		case IP_RECVTTL:
1451 		case IP_RECVIF:
1452 		case IP_PORTRANGE:
1453 		case IP_ONESBCAST:
1454 		case IP_DONTFRAG:
1455 		case IP_BINDANY:
1456 		case IP_RECVTOS:
1457 		case IP_BINDMULTI:
1458 		case IP_FLOWID:
1459 		case IP_FLOWTYPE:
1460 		case IP_RECVFLOWID:
1461 #ifdef	RSS
1462 		case IP_RSSBUCKETID:
1463 		case IP_RECVRSSBUCKETID:
1464 #endif
1465 		case IP_VLAN_PCP:
1466 			switch (sopt->sopt_name) {
1467 			case IP_TOS:
1468 				optval = inp->inp_ip_tos;
1469 				break;
1470 
1471 			case IP_TTL:
1472 				optval = inp->inp_ip_ttl;
1473 				break;
1474 
1475 			case IP_MINTTL:
1476 				optval = inp->inp_ip_minttl;
1477 				break;
1478 
1479 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1480 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1481 
1482 			case IP_RECVOPTS:
1483 				optval = OPTBIT(INP_RECVOPTS);
1484 				break;
1485 
1486 			case IP_RECVRETOPTS:
1487 				optval = OPTBIT(INP_RECVRETOPTS);
1488 				break;
1489 
1490 			case IP_RECVDSTADDR:
1491 				optval = OPTBIT(INP_RECVDSTADDR);
1492 				break;
1493 
1494 			case IP_ORIGDSTADDR:
1495 				optval = OPTBIT2(INP_ORIGDSTADDR);
1496 				break;
1497 
1498 			case IP_RECVTTL:
1499 				optval = OPTBIT(INP_RECVTTL);
1500 				break;
1501 
1502 			case IP_RECVIF:
1503 				optval = OPTBIT(INP_RECVIF);
1504 				break;
1505 
1506 			case IP_PORTRANGE:
1507 				if (inp->inp_flags & INP_HIGHPORT)
1508 					optval = IP_PORTRANGE_HIGH;
1509 				else if (inp->inp_flags & INP_LOWPORT)
1510 					optval = IP_PORTRANGE_LOW;
1511 				else
1512 					optval = 0;
1513 				break;
1514 
1515 			case IP_ONESBCAST:
1516 				optval = OPTBIT(INP_ONESBCAST);
1517 				break;
1518 			case IP_DONTFRAG:
1519 				optval = OPTBIT(INP_DONTFRAG);
1520 				break;
1521 			case IP_BINDANY:
1522 				optval = OPTBIT(INP_BINDANY);
1523 				break;
1524 			case IP_RECVTOS:
1525 				optval = OPTBIT(INP_RECVTOS);
1526 				break;
1527 			case IP_FLOWID:
1528 				optval = inp->inp_flowid;
1529 				break;
1530 			case IP_FLOWTYPE:
1531 				optval = inp->inp_flowtype;
1532 				break;
1533 			case IP_RECVFLOWID:
1534 				optval = OPTBIT2(INP_RECVFLOWID);
1535 				break;
1536 #ifdef	RSS
1537 			case IP_RSSBUCKETID:
1538 				retval = rss_hash2bucket(inp->inp_flowid,
1539 				    inp->inp_flowtype,
1540 				    &rss_bucket);
1541 				if (retval == 0)
1542 					optval = rss_bucket;
1543 				else
1544 					error = EINVAL;
1545 				break;
1546 			case IP_RECVRSSBUCKETID:
1547 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1548 				break;
1549 #endif
1550 			case IP_BINDMULTI:
1551 				optval = OPTBIT2(INP_BINDMULTI);
1552 				break;
1553 			case IP_VLAN_PCP:
1554 				if (OPTBIT2(INP_2PCP_SET)) {
1555 					optval = (inp->inp_flags2 &
1556 					    INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1557 				} else {
1558 					optval = -1;
1559 				}
1560 				break;
1561 			}
1562 			error = sooptcopyout(sopt, &optval, sizeof optval);
1563 			break;
1564 
1565 		/*
1566 		 * Multicast socket options are processed by the in_mcast
1567 		 * module.
1568 		 */
1569 		case IP_MULTICAST_IF:
1570 		case IP_MULTICAST_VIF:
1571 		case IP_MULTICAST_TTL:
1572 		case IP_MULTICAST_LOOP:
1573 		case IP_MSFILTER:
1574 			error = inp_getmoptions(inp, sopt);
1575 			break;
1576 
1577 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1578 		case IP_IPSEC_POLICY:
1579 			if (IPSEC_ENABLED(ipv4)) {
1580 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1581 				break;
1582 			}
1583 			/* FALLTHROUGH */
1584 #endif /* IPSEC */
1585 
1586 		default:
1587 			error = ENOPROTOOPT;
1588 			break;
1589 		}
1590 		break;
1591 	}
1592 	return (error);
1593 }
1594 
1595 /*
1596  * Routine called from ip_output() to loop back a copy of an IP multicast
1597  * packet to the input queue of a specified interface.  Note that this
1598  * calls the output routine of the loopback "driver", but with an interface
1599  * pointer that might NOT be a loopback interface -- evil, but easier than
1600  * replicating that code here.
1601  */
1602 static void
1603 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1604 {
1605 	struct ip *ip;
1606 	struct mbuf *copym;
1607 
1608 	/*
1609 	 * Make a deep copy of the packet because we're going to
1610 	 * modify the pack in order to generate checksums.
1611 	 */
1612 	copym = m_dup(m, M_NOWAIT);
1613 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1614 		copym = m_pullup(copym, hlen);
1615 	if (copym != NULL) {
1616 		/* If needed, compute the checksum and mark it as valid. */
1617 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1618 			in_delayed_cksum(copym);
1619 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1620 			copym->m_pkthdr.csum_flags |=
1621 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1622 			copym->m_pkthdr.csum_data = 0xffff;
1623 		}
1624 		/*
1625 		 * We don't bother to fragment if the IP length is greater
1626 		 * than the interface's MTU.  Can this possibly matter?
1627 		 */
1628 		ip = mtod(copym, struct ip *);
1629 		ip->ip_sum = 0;
1630 		ip->ip_sum = in_cksum(copym, hlen);
1631 		if_simloop(ifp, copym, AF_INET, 0);
1632 	}
1633 }
1634