xref: /freebsd/sys/netinet/ip_output.c (revision 1682a3ab4b725e31e9aab4aa4a8aa2ba400463fe)
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_mbuf_stress_test.h"
40 #include "opt_mpath.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/lock.h>
50 #include <sys/malloc.h>
51 #include <sys/mbuf.h>
52 #include <sys/priv.h>
53 #include <sys/proc.h>
54 #include <sys/protosw.h>
55 #include <sys/rmlock.h>
56 #include <sys/sdt.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/sysctl.h>
60 #include <sys/ucred.h>
61 
62 #include <net/if.h>
63 #include <net/if_var.h>
64 #include <net/if_llatbl.h>
65 #include <net/netisr.h>
66 #include <net/pfil.h>
67 #include <net/route.h>
68 #ifdef RADIX_MPATH
69 #include <net/radix_mpath.h>
70 #endif
71 #include <net/rss_config.h>
72 #include <net/vnet.h>
73 
74 #include <netinet/in.h>
75 #include <netinet/in_fib.h>
76 #include <netinet/in_kdtrace.h>
77 #include <netinet/in_systm.h>
78 #include <netinet/ip.h>
79 #include <netinet/in_pcb.h>
80 #include <netinet/in_rss.h>
81 #include <netinet/in_var.h>
82 #include <netinet/ip_var.h>
83 #include <netinet/ip_options.h>
84 
85 #include <netinet/udp.h>
86 #include <netinet/udp_var.h>
87 
88 #ifdef SCTP
89 #include <netinet/sctp.h>
90 #include <netinet/sctp_crc32.h>
91 #endif
92 
93 #include <netipsec/ipsec_support.h>
94 
95 #include <machine/in_cksum.h>
96 
97 #include <security/mac/mac_framework.h>
98 
99 #ifdef MBUF_STRESS_TEST
100 static int mbuf_frag_size = 0;
101 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
102 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
103 #endif
104 
105 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
106 
107 
108 extern int in_mcast_loop;
109 extern	struct protosw inetsw[];
110 
111 static inline int
112 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
113     struct sockaddr_in *dst, int *fibnum, int *error)
114 {
115 	struct m_tag *fwd_tag = NULL;
116 	struct mbuf *m;
117 	struct in_addr odst;
118 	struct ip *ip;
119 
120 	m = *mp;
121 	ip = mtod(m, struct ip *);
122 
123 	/* Run through list of hooks for output packets. */
124 	odst.s_addr = ip->ip_dst.s_addr;
125 	switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, PFIL_OUT, inp)) {
126 	case PFIL_DROPPED:
127 		*error = EPERM;
128 		/* FALLTHROUGH */
129 	case PFIL_CONSUMED:
130 		return 1; /* Finished */
131 	case PFIL_PASS:
132 		*error = 0;
133 	}
134 	m = *mp;
135 	ip = mtod(m, struct ip *);
136 
137 	/* See if destination IP address was changed by packet filter. */
138 	if (odst.s_addr != ip->ip_dst.s_addr) {
139 		m->m_flags |= M_SKIP_FIREWALL;
140 		/* If destination is now ourself drop to ip_input(). */
141 		if (in_localip(ip->ip_dst)) {
142 			m->m_flags |= M_FASTFWD_OURS;
143 			if (m->m_pkthdr.rcvif == NULL)
144 				m->m_pkthdr.rcvif = V_loif;
145 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
146 				m->m_pkthdr.csum_flags |=
147 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
148 				m->m_pkthdr.csum_data = 0xffff;
149 			}
150 			m->m_pkthdr.csum_flags |=
151 				CSUM_IP_CHECKED | CSUM_IP_VALID;
152 #ifdef SCTP
153 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
154 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
155 #endif
156 			*error = netisr_queue(NETISR_IP, m);
157 			return 1; /* Finished */
158 		}
159 
160 		bzero(dst, sizeof(*dst));
161 		dst->sin_family = AF_INET;
162 		dst->sin_len = sizeof(*dst);
163 		dst->sin_addr = ip->ip_dst;
164 
165 		return -1; /* Reloop */
166 	}
167 	/* See if fib was changed by packet filter. */
168 	if ((*fibnum) != M_GETFIB(m)) {
169 		m->m_flags |= M_SKIP_FIREWALL;
170 		*fibnum = M_GETFIB(m);
171 		return -1; /* Reloop for FIB change */
172 	}
173 
174 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
175 	if (m->m_flags & M_FASTFWD_OURS) {
176 		if (m->m_pkthdr.rcvif == NULL)
177 			m->m_pkthdr.rcvif = V_loif;
178 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
179 			m->m_pkthdr.csum_flags |=
180 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
181 			m->m_pkthdr.csum_data = 0xffff;
182 		}
183 #ifdef SCTP
184 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
185 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
186 #endif
187 		m->m_pkthdr.csum_flags |=
188 			CSUM_IP_CHECKED | CSUM_IP_VALID;
189 
190 		*error = netisr_queue(NETISR_IP, m);
191 		return 1; /* Finished */
192 	}
193 	/* Or forward to some other address? */
194 	if ((m->m_flags & M_IP_NEXTHOP) &&
195 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
196 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
197 		m->m_flags |= M_SKIP_FIREWALL;
198 		m->m_flags &= ~M_IP_NEXTHOP;
199 		m_tag_delete(m, fwd_tag);
200 
201 		return -1; /* Reloop for CHANGE of dst */
202 	}
203 
204 	return 0;
205 }
206 
207 static int
208 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
209     const struct sockaddr_in *gw, struct route *ro)
210 {
211 	struct m_snd_tag *mst;
212 	int error;
213 
214 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
215 	mst = NULL;
216 
217 #ifdef RATELIMIT
218 	if (inp != NULL) {
219 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
220 		    (inp->inp_snd_tag != NULL &&
221 		    inp->inp_snd_tag->ifp != ifp))
222 			in_pcboutput_txrtlmt(inp, ifp, m);
223 
224 		if (inp->inp_snd_tag != NULL)
225 			mst = inp->inp_snd_tag;
226 	}
227 #endif
228 	if (mst != NULL) {
229 		KASSERT(m->m_pkthdr.rcvif == NULL,
230 		    ("trying to add a send tag to a forwarded packet"));
231 		if (mst->ifp != ifp) {
232 			error = EAGAIN;
233 			goto done;
234 		}
235 
236 		/* stamp send tag on mbuf */
237 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
238 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
239 	}
240 
241 	error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
242 
243 done:
244 	/* Check for route change invalidating send tags. */
245 #ifdef RATELIMIT
246 	if (error == EAGAIN)
247 		in_pcboutput_eagain(inp);
248 #endif
249 	return (error);
250 }
251 
252 /*
253  * IP output.  The packet in mbuf chain m contains a skeletal IP
254  * header (with len, off, ttl, proto, tos, src, dst).
255  * The mbuf chain containing the packet will be freed.
256  * The mbuf opt, if present, will not be freed.
257  * If route ro is present and has ro_rt initialized, route lookup would be
258  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
259  * then result of route lookup is stored in ro->ro_rt.
260  *
261  * In the IP forwarding case, the packet will arrive with options already
262  * inserted, so must have a NULL opt pointer.
263  */
264 int
265 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
266     struct ip_moptions *imo, struct inpcb *inp)
267 {
268 	struct rm_priotracker in_ifa_tracker;
269 	struct epoch_tracker et;
270 	struct ip *ip;
271 	struct ifnet *ifp = NULL;	/* keep compiler happy */
272 	struct mbuf *m0;
273 	int hlen = sizeof (struct ip);
274 	int mtu;
275 	int error = 0;
276 	struct sockaddr_in *dst, sin;
277 	const struct sockaddr_in *gw;
278 	struct in_ifaddr *ia;
279 	struct in_addr src;
280 	int isbroadcast;
281 	uint16_t ip_len, ip_off;
282 	uint32_t fibnum;
283 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
284 	int no_route_but_check_spd = 0;
285 #endif
286 
287 	M_ASSERTPKTHDR(m);
288 
289 	if (inp != NULL) {
290 		INP_LOCK_ASSERT(inp);
291 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
292 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
293 			m->m_pkthdr.flowid = inp->inp_flowid;
294 			M_HASHTYPE_SET(m, inp->inp_flowtype);
295 		}
296 #ifdef NUMA
297 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
298 #endif
299 	}
300 
301 	if (opt) {
302 		int len = 0;
303 		m = ip_insertoptions(m, opt, &len);
304 		if (len != 0)
305 			hlen = len; /* ip->ip_hl is updated above */
306 	}
307 	ip = mtod(m, struct ip *);
308 	ip_len = ntohs(ip->ip_len);
309 	ip_off = ntohs(ip->ip_off);
310 
311 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
312 		ip->ip_v = IPVERSION;
313 		ip->ip_hl = hlen >> 2;
314 		ip_fillid(ip);
315 	} else {
316 		/* Header already set, fetch hlen from there */
317 		hlen = ip->ip_hl << 2;
318 	}
319 	if ((flags & IP_FORWARDING) == 0)
320 		IPSTAT_INC(ips_localout);
321 
322 	/*
323 	 * dst/gw handling:
324 	 *
325 	 * gw is readonly but can point either to dst OR rt_gateway,
326 	 * therefore we need restore gw if we're redoing lookup.
327 	 */
328 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
329 	if (ro != NULL)
330 		dst = (struct sockaddr_in *)&ro->ro_dst;
331 	else
332 		dst = &sin;
333 	if (ro == NULL || ro->ro_rt == NULL) {
334 		bzero(dst, sizeof(*dst));
335 		dst->sin_family = AF_INET;
336 		dst->sin_len = sizeof(*dst);
337 		dst->sin_addr = ip->ip_dst;
338 	}
339 	gw = dst;
340 	NET_EPOCH_ENTER(et);
341 again:
342 	/*
343 	 * Validate route against routing table additions;
344 	 * a better/more specific route might have been added.
345 	 */
346 	if (inp != NULL && ro != NULL && ro->ro_rt != NULL)
347 		RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
348 	/*
349 	 * If there is a cached route,
350 	 * check that it is to the same destination
351 	 * and is still up.  If not, free it and try again.
352 	 * The address family should also be checked in case of sharing the
353 	 * cache with IPv6.
354 	 * Also check whether routing cache needs invalidation.
355 	 */
356 	if (ro != NULL && ro->ro_rt != NULL &&
357 	    ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
358 	    ro->ro_rt->rt_ifp == NULL || !RT_LINK_IS_UP(ro->ro_rt->rt_ifp) ||
359 	    dst->sin_family != AF_INET ||
360 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
361 		RO_INVALIDATE_CACHE(ro);
362 	ia = NULL;
363 	/*
364 	 * If routing to interface only, short circuit routing lookup.
365 	 * The use of an all-ones broadcast address implies this; an
366 	 * interface is specified by the broadcast address of an interface,
367 	 * or the destination address of a ptp interface.
368 	 */
369 	if (flags & IP_SENDONES) {
370 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
371 						      M_GETFIB(m)))) == NULL &&
372 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
373 						    M_GETFIB(m)))) == NULL) {
374 			IPSTAT_INC(ips_noroute);
375 			error = ENETUNREACH;
376 			goto bad;
377 		}
378 		ip->ip_dst.s_addr = INADDR_BROADCAST;
379 		dst->sin_addr = ip->ip_dst;
380 		ifp = ia->ia_ifp;
381 		mtu = ifp->if_mtu;
382 		ip->ip_ttl = 1;
383 		isbroadcast = 1;
384 		src = IA_SIN(ia)->sin_addr;
385 	} else if (flags & IP_ROUTETOIF) {
386 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
387 						    M_GETFIB(m)))) == NULL &&
388 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
389 						M_GETFIB(m)))) == NULL) {
390 			IPSTAT_INC(ips_noroute);
391 			error = ENETUNREACH;
392 			goto bad;
393 		}
394 		ifp = ia->ia_ifp;
395 		mtu = ifp->if_mtu;
396 		ip->ip_ttl = 1;
397 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
398 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
399 		src = IA_SIN(ia)->sin_addr;
400 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
401 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
402 		/*
403 		 * Bypass the normal routing lookup for multicast
404 		 * packets if the interface is specified.
405 		 */
406 		ifp = imo->imo_multicast_ifp;
407 		mtu = ifp->if_mtu;
408 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
409 		isbroadcast = 0;	/* fool gcc */
410 		/* Interface may have no addresses. */
411 		if (ia != NULL)
412 			src = IA_SIN(ia)->sin_addr;
413 		else
414 			src.s_addr = INADDR_ANY;
415 	} else if (ro != NULL) {
416 		if (ro->ro_rt == NULL) {
417 			/*
418 			 * We want to do any cloning requested by the link
419 			 * layer, as this is probably required in all cases
420 			 * for correct operation (as it is for ARP).
421 			 */
422 #ifdef RADIX_MPATH
423 			rtalloc_mpath_fib(ro,
424 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
425 			    fibnum);
426 #else
427 			in_rtalloc_ign(ro, 0, fibnum);
428 #endif
429 			if (ro->ro_rt == NULL ||
430 			    (ro->ro_rt->rt_flags & RTF_UP) == 0 ||
431 			    ro->ro_rt->rt_ifp == NULL ||
432 			    !RT_LINK_IS_UP(ro->ro_rt->rt_ifp)) {
433 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
434 				/*
435 				 * There is no route for this packet, but it is
436 				 * possible that a matching SPD entry exists.
437 				 */
438 				no_route_but_check_spd = 1;
439 				mtu = 0; /* Silence GCC warning. */
440 				goto sendit;
441 #endif
442 				IPSTAT_INC(ips_noroute);
443 				error = EHOSTUNREACH;
444 				goto bad;
445 			}
446 		}
447 		ia = ifatoia(ro->ro_rt->rt_ifa);
448 		ifp = ro->ro_rt->rt_ifp;
449 		counter_u64_add(ro->ro_rt->rt_pksent, 1);
450 		rt_update_ro_flags(ro);
451 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
452 			gw = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
453 		if (ro->ro_rt->rt_flags & RTF_HOST)
454 			isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
455 		else if (ifp->if_flags & IFF_BROADCAST)
456 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
457 		else
458 			isbroadcast = 0;
459 		if (ro->ro_rt->rt_flags & RTF_HOST)
460 			mtu = ro->ro_rt->rt_mtu;
461 		else
462 			mtu = ifp->if_mtu;
463 		src = IA_SIN(ia)->sin_addr;
464 	} else {
465 		struct nhop4_extended nh;
466 
467 		bzero(&nh, sizeof(nh));
468 		if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh) !=
469 		    0) {
470 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
471 			/*
472 			 * There is no route for this packet, but it is
473 			 * possible that a matching SPD entry exists.
474 			 */
475 			no_route_but_check_spd = 1;
476 			mtu = 0; /* Silence GCC warning. */
477 			goto sendit;
478 #endif
479 			IPSTAT_INC(ips_noroute);
480 			error = EHOSTUNREACH;
481 			goto bad;
482 		}
483 		ifp = nh.nh_ifp;
484 		mtu = nh.nh_mtu;
485 		/*
486 		 * We are rewriting here dst to be gw actually, contradicting
487 		 * comment at the beginning of the function. However, in this
488 		 * case we are always dealing with on stack dst.
489 		 * In case if pfil(9) sends us back to beginning of the
490 		 * function, the dst would be rewritten by ip_output_pfil().
491 		 */
492 		MPASS(dst == &sin);
493 		dst->sin_addr = nh.nh_addr;
494 		ia = nh.nh_ia;
495 		src = nh.nh_src;
496 		isbroadcast = (((nh.nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
497 		    (NHF_HOST | NHF_BROADCAST)) ||
498 		    ((ifp->if_flags & IFF_BROADCAST) &&
499 		    in_ifaddr_broadcast(dst->sin_addr, ia)));
500 	}
501 
502 	/* Catch a possible divide by zero later. */
503 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (rt_flags=0x%08x) ifp=%p",
504 	    __func__, mtu, ro,
505 	    (ro != NULL && ro->ro_rt != NULL) ? ro->ro_rt->rt_flags : 0, ifp));
506 
507 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
508 		m->m_flags |= M_MCAST;
509 		/*
510 		 * IP destination address is multicast.  Make sure "gw"
511 		 * still points to the address in "ro".  (It may have been
512 		 * changed to point to a gateway address, above.)
513 		 */
514 		gw = dst;
515 		/*
516 		 * See if the caller provided any multicast options
517 		 */
518 		if (imo != NULL) {
519 			ip->ip_ttl = imo->imo_multicast_ttl;
520 			if (imo->imo_multicast_vif != -1)
521 				ip->ip_src.s_addr =
522 				    ip_mcast_src ?
523 				    ip_mcast_src(imo->imo_multicast_vif) :
524 				    INADDR_ANY;
525 		} else
526 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
527 		/*
528 		 * Confirm that the outgoing interface supports multicast.
529 		 */
530 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
531 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
532 				IPSTAT_INC(ips_noroute);
533 				error = ENETUNREACH;
534 				goto bad;
535 			}
536 		}
537 		/*
538 		 * If source address not specified yet, use address
539 		 * of outgoing interface.
540 		 */
541 		if (ip->ip_src.s_addr == INADDR_ANY)
542 			ip->ip_src = src;
543 
544 		if ((imo == NULL && in_mcast_loop) ||
545 		    (imo && imo->imo_multicast_loop)) {
546 			/*
547 			 * Loop back multicast datagram if not expressly
548 			 * forbidden to do so, even if we are not a member
549 			 * of the group; ip_input() will filter it later,
550 			 * thus deferring a hash lookup and mutex acquisition
551 			 * at the expense of a cheap copy using m_copym().
552 			 */
553 			ip_mloopback(ifp, m, hlen);
554 		} else {
555 			/*
556 			 * If we are acting as a multicast router, perform
557 			 * multicast forwarding as if the packet had just
558 			 * arrived on the interface to which we are about
559 			 * to send.  The multicast forwarding function
560 			 * recursively calls this function, using the
561 			 * IP_FORWARDING flag to prevent infinite recursion.
562 			 *
563 			 * Multicasts that are looped back by ip_mloopback(),
564 			 * above, will be forwarded by the ip_input() routine,
565 			 * if necessary.
566 			 */
567 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
568 				/*
569 				 * If rsvp daemon is not running, do not
570 				 * set ip_moptions. This ensures that the packet
571 				 * is multicast and not just sent down one link
572 				 * as prescribed by rsvpd.
573 				 */
574 				if (!V_rsvp_on)
575 					imo = NULL;
576 				if (ip_mforward &&
577 				    ip_mforward(ip, ifp, m, imo) != 0) {
578 					m_freem(m);
579 					goto done;
580 				}
581 			}
582 		}
583 
584 		/*
585 		 * Multicasts with a time-to-live of zero may be looped-
586 		 * back, above, but must not be transmitted on a network.
587 		 * Also, multicasts addressed to the loopback interface
588 		 * are not sent -- the above call to ip_mloopback() will
589 		 * loop back a copy. ip_input() will drop the copy if
590 		 * this host does not belong to the destination group on
591 		 * the loopback interface.
592 		 */
593 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
594 			m_freem(m);
595 			goto done;
596 		}
597 
598 		goto sendit;
599 	}
600 
601 	/*
602 	 * If the source address is not specified yet, use the address
603 	 * of the outoing interface.
604 	 */
605 	if (ip->ip_src.s_addr == INADDR_ANY)
606 		ip->ip_src = src;
607 
608 	/*
609 	 * Look for broadcast address and
610 	 * verify user is allowed to send
611 	 * such a packet.
612 	 */
613 	if (isbroadcast) {
614 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
615 			error = EADDRNOTAVAIL;
616 			goto bad;
617 		}
618 		if ((flags & IP_ALLOWBROADCAST) == 0) {
619 			error = EACCES;
620 			goto bad;
621 		}
622 		/* don't allow broadcast messages to be fragmented */
623 		if (ip_len > mtu) {
624 			error = EMSGSIZE;
625 			goto bad;
626 		}
627 		m->m_flags |= M_BCAST;
628 	} else {
629 		m->m_flags &= ~M_BCAST;
630 	}
631 
632 sendit:
633 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
634 	if (IPSEC_ENABLED(ipv4)) {
635 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
636 			if (error == EINPROGRESS)
637 				error = 0;
638 			goto done;
639 		}
640 	}
641 	/*
642 	 * Check if there was a route for this packet; return error if not.
643 	 */
644 	if (no_route_but_check_spd) {
645 		IPSTAT_INC(ips_noroute);
646 		error = EHOSTUNREACH;
647 		goto bad;
648 	}
649 	/* Update variables that are affected by ipsec4_output(). */
650 	ip = mtod(m, struct ip *);
651 	hlen = ip->ip_hl << 2;
652 #endif /* IPSEC */
653 
654 	/* Jump over all PFIL processing if hooks are not active. */
655 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
656 		switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
657 		case 1: /* Finished */
658 			goto done;
659 
660 		case 0: /* Continue normally */
661 			ip = mtod(m, struct ip *);
662 			break;
663 
664 		case -1: /* Need to try again */
665 			/* Reset everything for a new round */
666 			if (ro != NULL) {
667 				RO_RTFREE(ro);
668 				ro->ro_prepend = NULL;
669 			}
670 			gw = dst;
671 			ip = mtod(m, struct ip *);
672 			goto again;
673 
674 		}
675 	}
676 
677 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
678 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
679 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
680 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
681 			IPSTAT_INC(ips_badaddr);
682 			error = EADDRNOTAVAIL;
683 			goto bad;
684 		}
685 	}
686 
687 	m->m_pkthdr.csum_flags |= CSUM_IP;
688 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
689 		in_delayed_cksum(m);
690 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
691 	}
692 #ifdef SCTP
693 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
694 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
695 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
696 	}
697 #endif
698 
699 	/*
700 	 * If small enough for interface, or the interface will take
701 	 * care of the fragmentation for us, we can just send directly.
702 	 */
703 	if (ip_len <= mtu ||
704 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
705 		ip->ip_sum = 0;
706 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
707 			ip->ip_sum = in_cksum(m, hlen);
708 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
709 		}
710 
711 		/*
712 		 * Record statistics for this interface address.
713 		 * With CSUM_TSO the byte/packet count will be slightly
714 		 * incorrect because we count the IP+TCP headers only
715 		 * once instead of for every generated packet.
716 		 */
717 		if (!(flags & IP_FORWARDING) && ia) {
718 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
719 				counter_u64_add(ia->ia_ifa.ifa_opackets,
720 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
721 			else
722 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
723 
724 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
725 		}
726 #ifdef MBUF_STRESS_TEST
727 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
728 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
729 #endif
730 		/*
731 		 * Reset layer specific mbuf flags
732 		 * to avoid confusing lower layers.
733 		 */
734 		m_clrprotoflags(m);
735 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
736 		error = ip_output_send(inp, ifp, m, gw, ro);
737 		goto done;
738 	}
739 
740 	/* Balk when DF bit is set or the interface didn't support TSO. */
741 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
742 		error = EMSGSIZE;
743 		IPSTAT_INC(ips_cantfrag);
744 		goto bad;
745 	}
746 
747 	/*
748 	 * Too large for interface; fragment if possible. If successful,
749 	 * on return, m will point to a list of packets to be sent.
750 	 */
751 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
752 	if (error)
753 		goto bad;
754 	for (; m; m = m0) {
755 		m0 = m->m_nextpkt;
756 		m->m_nextpkt = 0;
757 		if (error == 0) {
758 			/* Record statistics for this interface address. */
759 			if (ia != NULL) {
760 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
761 				counter_u64_add(ia->ia_ifa.ifa_obytes,
762 				    m->m_pkthdr.len);
763 			}
764 			/*
765 			 * Reset layer specific mbuf flags
766 			 * to avoid confusing upper layers.
767 			 */
768 			m_clrprotoflags(m);
769 
770 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
771 			    mtod(m, struct ip *), NULL);
772 			error = ip_output_send(inp, ifp, m, gw, ro);
773 		} else
774 			m_freem(m);
775 	}
776 
777 	if (error == 0)
778 		IPSTAT_INC(ips_fragmented);
779 
780 done:
781 	NET_EPOCH_EXIT(et);
782 	return (error);
783  bad:
784 	m_freem(m);
785 	goto done;
786 }
787 
788 /*
789  * Create a chain of fragments which fit the given mtu. m_frag points to the
790  * mbuf to be fragmented; on return it points to the chain with the fragments.
791  * Return 0 if no error. If error, m_frag may contain a partially built
792  * chain of fragments that should be freed by the caller.
793  *
794  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
795  */
796 int
797 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
798     u_long if_hwassist_flags)
799 {
800 	int error = 0;
801 	int hlen = ip->ip_hl << 2;
802 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
803 	int off;
804 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
805 	int firstlen;
806 	struct mbuf **mnext;
807 	int nfrags;
808 	uint16_t ip_len, ip_off;
809 
810 	ip_len = ntohs(ip->ip_len);
811 	ip_off = ntohs(ip->ip_off);
812 
813 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
814 		IPSTAT_INC(ips_cantfrag);
815 		return EMSGSIZE;
816 	}
817 
818 	/*
819 	 * Must be able to put at least 8 bytes per fragment.
820 	 */
821 	if (len < 8)
822 		return EMSGSIZE;
823 
824 	/*
825 	 * If the interface will not calculate checksums on
826 	 * fragmented packets, then do it here.
827 	 */
828 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
829 		in_delayed_cksum(m0);
830 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
831 	}
832 #ifdef SCTP
833 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
834 		sctp_delayed_cksum(m0, hlen);
835 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
836 	}
837 #endif
838 	if (len > PAGE_SIZE) {
839 		/*
840 		 * Fragment large datagrams such that each segment
841 		 * contains a multiple of PAGE_SIZE amount of data,
842 		 * plus headers. This enables a receiver to perform
843 		 * page-flipping zero-copy optimizations.
844 		 *
845 		 * XXX When does this help given that sender and receiver
846 		 * could have different page sizes, and also mtu could
847 		 * be less than the receiver's page size ?
848 		 */
849 		int newlen;
850 
851 		off = MIN(mtu, m0->m_pkthdr.len);
852 
853 		/*
854 		 * firstlen (off - hlen) must be aligned on an
855 		 * 8-byte boundary
856 		 */
857 		if (off < hlen)
858 			goto smart_frag_failure;
859 		off = ((off - hlen) & ~7) + hlen;
860 		newlen = (~PAGE_MASK) & mtu;
861 		if ((newlen + sizeof (struct ip)) > mtu) {
862 			/* we failed, go back the default */
863 smart_frag_failure:
864 			newlen = len;
865 			off = hlen + len;
866 		}
867 		len = newlen;
868 
869 	} else {
870 		off = hlen + len;
871 	}
872 
873 	firstlen = off - hlen;
874 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
875 
876 	/*
877 	 * Loop through length of segment after first fragment,
878 	 * make new header and copy data of each part and link onto chain.
879 	 * Here, m0 is the original packet, m is the fragment being created.
880 	 * The fragments are linked off the m_nextpkt of the original
881 	 * packet, which after processing serves as the first fragment.
882 	 */
883 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
884 		struct ip *mhip;	/* ip header on the fragment */
885 		struct mbuf *m;
886 		int mhlen = sizeof (struct ip);
887 
888 		m = m_gethdr(M_NOWAIT, MT_DATA);
889 		if (m == NULL) {
890 			error = ENOBUFS;
891 			IPSTAT_INC(ips_odropped);
892 			goto done;
893 		}
894 		/*
895 		 * Make sure the complete packet header gets copied
896 		 * from the originating mbuf to the newly created
897 		 * mbuf. This also ensures that existing firewall
898 		 * classification(s), VLAN tags and so on get copied
899 		 * to the resulting fragmented packet(s):
900 		 */
901 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
902 			m_free(m);
903 			error = ENOBUFS;
904 			IPSTAT_INC(ips_odropped);
905 			goto done;
906 		}
907 		/*
908 		 * In the first mbuf, leave room for the link header, then
909 		 * copy the original IP header including options. The payload
910 		 * goes into an additional mbuf chain returned by m_copym().
911 		 */
912 		m->m_data += max_linkhdr;
913 		mhip = mtod(m, struct ip *);
914 		*mhip = *ip;
915 		if (hlen > sizeof (struct ip)) {
916 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
917 			mhip->ip_v = IPVERSION;
918 			mhip->ip_hl = mhlen >> 2;
919 		}
920 		m->m_len = mhlen;
921 		/* XXX do we need to add ip_off below ? */
922 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
923 		if (off + len >= ip_len)
924 			len = ip_len - off;
925 		else
926 			mhip->ip_off |= IP_MF;
927 		mhip->ip_len = htons((u_short)(len + mhlen));
928 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
929 		if (m->m_next == NULL) {	/* copy failed */
930 			m_free(m);
931 			error = ENOBUFS;	/* ??? */
932 			IPSTAT_INC(ips_odropped);
933 			goto done;
934 		}
935 		m->m_pkthdr.len = mhlen + len;
936 #ifdef MAC
937 		mac_netinet_fragment(m0, m);
938 #endif
939 		mhip->ip_off = htons(mhip->ip_off);
940 		mhip->ip_sum = 0;
941 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
942 			mhip->ip_sum = in_cksum(m, mhlen);
943 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
944 		}
945 		*mnext = m;
946 		mnext = &m->m_nextpkt;
947 	}
948 	IPSTAT_ADD(ips_ofragments, nfrags);
949 
950 	/*
951 	 * Update first fragment by trimming what's been copied out
952 	 * and updating header.
953 	 */
954 	m_adj(m0, hlen + firstlen - ip_len);
955 	m0->m_pkthdr.len = hlen + firstlen;
956 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
957 	ip->ip_off = htons(ip_off | IP_MF);
958 	ip->ip_sum = 0;
959 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
960 		ip->ip_sum = in_cksum(m0, hlen);
961 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
962 	}
963 
964 done:
965 	*m_frag = m0;
966 	return error;
967 }
968 
969 void
970 in_delayed_cksum(struct mbuf *m)
971 {
972 	struct ip *ip;
973 	struct udphdr *uh;
974 	uint16_t cklen, csum, offset;
975 
976 	ip = mtod(m, struct ip *);
977 	offset = ip->ip_hl << 2 ;
978 
979 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
980 		/* if udp header is not in the first mbuf copy udplen */
981 		if (offset + sizeof(struct udphdr) > m->m_len) {
982 			m_copydata(m, offset + offsetof(struct udphdr,
983 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
984 			cklen = ntohs(cklen);
985 		} else {
986 			uh = (struct udphdr *)mtodo(m, offset);
987 			cklen = ntohs(uh->uh_ulen);
988 		}
989 		csum = in_cksum_skip(m, cklen + offset, offset);
990 		if (csum == 0)
991 			csum = 0xffff;
992 	} else {
993 		cklen = ntohs(ip->ip_len);
994 		csum = in_cksum_skip(m, cklen, offset);
995 	}
996 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
997 
998 	if (offset + sizeof(csum) > m->m_len)
999 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1000 	else
1001 		*(u_short *)mtodo(m, offset) = csum;
1002 }
1003 
1004 /*
1005  * IP socket option processing.
1006  */
1007 int
1008 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1009 {
1010 	struct	inpcb *inp = sotoinpcb(so);
1011 	int	error, optval;
1012 #ifdef	RSS
1013 	uint32_t rss_bucket;
1014 	int retval;
1015 #endif
1016 
1017 	error = optval = 0;
1018 	if (sopt->sopt_level != IPPROTO_IP) {
1019 		error = EINVAL;
1020 
1021 		if (sopt->sopt_level == SOL_SOCKET &&
1022 		    sopt->sopt_dir == SOPT_SET) {
1023 			switch (sopt->sopt_name) {
1024 			case SO_REUSEADDR:
1025 				INP_WLOCK(inp);
1026 				if ((so->so_options & SO_REUSEADDR) != 0)
1027 					inp->inp_flags2 |= INP_REUSEADDR;
1028 				else
1029 					inp->inp_flags2 &= ~INP_REUSEADDR;
1030 				INP_WUNLOCK(inp);
1031 				error = 0;
1032 				break;
1033 			case SO_REUSEPORT:
1034 				INP_WLOCK(inp);
1035 				if ((so->so_options & SO_REUSEPORT) != 0)
1036 					inp->inp_flags2 |= INP_REUSEPORT;
1037 				else
1038 					inp->inp_flags2 &= ~INP_REUSEPORT;
1039 				INP_WUNLOCK(inp);
1040 				error = 0;
1041 				break;
1042 			case SO_REUSEPORT_LB:
1043 				INP_WLOCK(inp);
1044 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1045 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1046 				else
1047 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1048 				INP_WUNLOCK(inp);
1049 				error = 0;
1050 				break;
1051 			case SO_SETFIB:
1052 				INP_WLOCK(inp);
1053 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1054 				INP_WUNLOCK(inp);
1055 				error = 0;
1056 				break;
1057 			case SO_MAX_PACING_RATE:
1058 #ifdef RATELIMIT
1059 				INP_WLOCK(inp);
1060 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1061 				INP_WUNLOCK(inp);
1062 				error = 0;
1063 #else
1064 				error = EOPNOTSUPP;
1065 #endif
1066 				break;
1067 			default:
1068 				break;
1069 			}
1070 		}
1071 		return (error);
1072 	}
1073 
1074 	switch (sopt->sopt_dir) {
1075 	case SOPT_SET:
1076 		switch (sopt->sopt_name) {
1077 		case IP_OPTIONS:
1078 #ifdef notyet
1079 		case IP_RETOPTS:
1080 #endif
1081 		{
1082 			struct mbuf *m;
1083 			if (sopt->sopt_valsize > MLEN) {
1084 				error = EMSGSIZE;
1085 				break;
1086 			}
1087 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1088 			if (m == NULL) {
1089 				error = ENOBUFS;
1090 				break;
1091 			}
1092 			m->m_len = sopt->sopt_valsize;
1093 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1094 					    m->m_len);
1095 			if (error) {
1096 				m_free(m);
1097 				break;
1098 			}
1099 			INP_WLOCK(inp);
1100 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1101 			INP_WUNLOCK(inp);
1102 			return (error);
1103 		}
1104 
1105 		case IP_BINDANY:
1106 			if (sopt->sopt_td != NULL) {
1107 				error = priv_check(sopt->sopt_td,
1108 				    PRIV_NETINET_BINDANY);
1109 				if (error)
1110 					break;
1111 			}
1112 			/* FALLTHROUGH */
1113 		case IP_BINDMULTI:
1114 #ifdef	RSS
1115 		case IP_RSS_LISTEN_BUCKET:
1116 #endif
1117 		case IP_TOS:
1118 		case IP_TTL:
1119 		case IP_MINTTL:
1120 		case IP_RECVOPTS:
1121 		case IP_RECVRETOPTS:
1122 		case IP_ORIGDSTADDR:
1123 		case IP_RECVDSTADDR:
1124 		case IP_RECVTTL:
1125 		case IP_RECVIF:
1126 		case IP_ONESBCAST:
1127 		case IP_DONTFRAG:
1128 		case IP_RECVTOS:
1129 		case IP_RECVFLOWID:
1130 #ifdef	RSS
1131 		case IP_RECVRSSBUCKETID:
1132 #endif
1133 			error = sooptcopyin(sopt, &optval, sizeof optval,
1134 					    sizeof optval);
1135 			if (error)
1136 				break;
1137 
1138 			switch (sopt->sopt_name) {
1139 			case IP_TOS:
1140 				inp->inp_ip_tos = optval;
1141 				break;
1142 
1143 			case IP_TTL:
1144 				inp->inp_ip_ttl = optval;
1145 				break;
1146 
1147 			case IP_MINTTL:
1148 				if (optval >= 0 && optval <= MAXTTL)
1149 					inp->inp_ip_minttl = optval;
1150 				else
1151 					error = EINVAL;
1152 				break;
1153 
1154 #define	OPTSET(bit) do {						\
1155 	INP_WLOCK(inp);							\
1156 	if (optval)							\
1157 		inp->inp_flags |= bit;					\
1158 	else								\
1159 		inp->inp_flags &= ~bit;					\
1160 	INP_WUNLOCK(inp);						\
1161 } while (0)
1162 
1163 #define	OPTSET2(bit, val) do {						\
1164 	INP_WLOCK(inp);							\
1165 	if (val)							\
1166 		inp->inp_flags2 |= bit;					\
1167 	else								\
1168 		inp->inp_flags2 &= ~bit;				\
1169 	INP_WUNLOCK(inp);						\
1170 } while (0)
1171 
1172 			case IP_RECVOPTS:
1173 				OPTSET(INP_RECVOPTS);
1174 				break;
1175 
1176 			case IP_RECVRETOPTS:
1177 				OPTSET(INP_RECVRETOPTS);
1178 				break;
1179 
1180 			case IP_RECVDSTADDR:
1181 				OPTSET(INP_RECVDSTADDR);
1182 				break;
1183 
1184 			case IP_ORIGDSTADDR:
1185 				OPTSET2(INP_ORIGDSTADDR, optval);
1186 				break;
1187 
1188 			case IP_RECVTTL:
1189 				OPTSET(INP_RECVTTL);
1190 				break;
1191 
1192 			case IP_RECVIF:
1193 				OPTSET(INP_RECVIF);
1194 				break;
1195 
1196 			case IP_ONESBCAST:
1197 				OPTSET(INP_ONESBCAST);
1198 				break;
1199 			case IP_DONTFRAG:
1200 				OPTSET(INP_DONTFRAG);
1201 				break;
1202 			case IP_BINDANY:
1203 				OPTSET(INP_BINDANY);
1204 				break;
1205 			case IP_RECVTOS:
1206 				OPTSET(INP_RECVTOS);
1207 				break;
1208 			case IP_BINDMULTI:
1209 				OPTSET2(INP_BINDMULTI, optval);
1210 				break;
1211 			case IP_RECVFLOWID:
1212 				OPTSET2(INP_RECVFLOWID, optval);
1213 				break;
1214 #ifdef	RSS
1215 			case IP_RSS_LISTEN_BUCKET:
1216 				if ((optval >= 0) &&
1217 				    (optval < rss_getnumbuckets())) {
1218 					inp->inp_rss_listen_bucket = optval;
1219 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1220 				} else {
1221 					error = EINVAL;
1222 				}
1223 				break;
1224 			case IP_RECVRSSBUCKETID:
1225 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1226 				break;
1227 #endif
1228 			}
1229 			break;
1230 #undef OPTSET
1231 #undef OPTSET2
1232 
1233 		/*
1234 		 * Multicast socket options are processed by the in_mcast
1235 		 * module.
1236 		 */
1237 		case IP_MULTICAST_IF:
1238 		case IP_MULTICAST_VIF:
1239 		case IP_MULTICAST_TTL:
1240 		case IP_MULTICAST_LOOP:
1241 		case IP_ADD_MEMBERSHIP:
1242 		case IP_DROP_MEMBERSHIP:
1243 		case IP_ADD_SOURCE_MEMBERSHIP:
1244 		case IP_DROP_SOURCE_MEMBERSHIP:
1245 		case IP_BLOCK_SOURCE:
1246 		case IP_UNBLOCK_SOURCE:
1247 		case IP_MSFILTER:
1248 		case MCAST_JOIN_GROUP:
1249 		case MCAST_LEAVE_GROUP:
1250 		case MCAST_JOIN_SOURCE_GROUP:
1251 		case MCAST_LEAVE_SOURCE_GROUP:
1252 		case MCAST_BLOCK_SOURCE:
1253 		case MCAST_UNBLOCK_SOURCE:
1254 			error = inp_setmoptions(inp, sopt);
1255 			break;
1256 
1257 		case IP_PORTRANGE:
1258 			error = sooptcopyin(sopt, &optval, sizeof optval,
1259 					    sizeof optval);
1260 			if (error)
1261 				break;
1262 
1263 			INP_WLOCK(inp);
1264 			switch (optval) {
1265 			case IP_PORTRANGE_DEFAULT:
1266 				inp->inp_flags &= ~(INP_LOWPORT);
1267 				inp->inp_flags &= ~(INP_HIGHPORT);
1268 				break;
1269 
1270 			case IP_PORTRANGE_HIGH:
1271 				inp->inp_flags &= ~(INP_LOWPORT);
1272 				inp->inp_flags |= INP_HIGHPORT;
1273 				break;
1274 
1275 			case IP_PORTRANGE_LOW:
1276 				inp->inp_flags &= ~(INP_HIGHPORT);
1277 				inp->inp_flags |= INP_LOWPORT;
1278 				break;
1279 
1280 			default:
1281 				error = EINVAL;
1282 				break;
1283 			}
1284 			INP_WUNLOCK(inp);
1285 			break;
1286 
1287 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1288 		case IP_IPSEC_POLICY:
1289 			if (IPSEC_ENABLED(ipv4)) {
1290 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1291 				break;
1292 			}
1293 			/* FALLTHROUGH */
1294 #endif /* IPSEC */
1295 
1296 		default:
1297 			error = ENOPROTOOPT;
1298 			break;
1299 		}
1300 		break;
1301 
1302 	case SOPT_GET:
1303 		switch (sopt->sopt_name) {
1304 		case IP_OPTIONS:
1305 		case IP_RETOPTS:
1306 			INP_RLOCK(inp);
1307 			if (inp->inp_options) {
1308 				struct mbuf *options;
1309 
1310 				options = m_copym(inp->inp_options, 0,
1311 				    M_COPYALL, M_NOWAIT);
1312 				INP_RUNLOCK(inp);
1313 				if (options != NULL) {
1314 					error = sooptcopyout(sopt,
1315 							     mtod(options, char *),
1316 							     options->m_len);
1317 					m_freem(options);
1318 				} else
1319 					error = ENOMEM;
1320 			} else {
1321 				INP_RUNLOCK(inp);
1322 				sopt->sopt_valsize = 0;
1323 			}
1324 			break;
1325 
1326 		case IP_TOS:
1327 		case IP_TTL:
1328 		case IP_MINTTL:
1329 		case IP_RECVOPTS:
1330 		case IP_RECVRETOPTS:
1331 		case IP_ORIGDSTADDR:
1332 		case IP_RECVDSTADDR:
1333 		case IP_RECVTTL:
1334 		case IP_RECVIF:
1335 		case IP_PORTRANGE:
1336 		case IP_ONESBCAST:
1337 		case IP_DONTFRAG:
1338 		case IP_BINDANY:
1339 		case IP_RECVTOS:
1340 		case IP_BINDMULTI:
1341 		case IP_FLOWID:
1342 		case IP_FLOWTYPE:
1343 		case IP_RECVFLOWID:
1344 #ifdef	RSS
1345 		case IP_RSSBUCKETID:
1346 		case IP_RECVRSSBUCKETID:
1347 #endif
1348 			switch (sopt->sopt_name) {
1349 
1350 			case IP_TOS:
1351 				optval = inp->inp_ip_tos;
1352 				break;
1353 
1354 			case IP_TTL:
1355 				optval = inp->inp_ip_ttl;
1356 				break;
1357 
1358 			case IP_MINTTL:
1359 				optval = inp->inp_ip_minttl;
1360 				break;
1361 
1362 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1363 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1364 
1365 			case IP_RECVOPTS:
1366 				optval = OPTBIT(INP_RECVOPTS);
1367 				break;
1368 
1369 			case IP_RECVRETOPTS:
1370 				optval = OPTBIT(INP_RECVRETOPTS);
1371 				break;
1372 
1373 			case IP_RECVDSTADDR:
1374 				optval = OPTBIT(INP_RECVDSTADDR);
1375 				break;
1376 
1377 			case IP_ORIGDSTADDR:
1378 				optval = OPTBIT2(INP_ORIGDSTADDR);
1379 				break;
1380 
1381 			case IP_RECVTTL:
1382 				optval = OPTBIT(INP_RECVTTL);
1383 				break;
1384 
1385 			case IP_RECVIF:
1386 				optval = OPTBIT(INP_RECVIF);
1387 				break;
1388 
1389 			case IP_PORTRANGE:
1390 				if (inp->inp_flags & INP_HIGHPORT)
1391 					optval = IP_PORTRANGE_HIGH;
1392 				else if (inp->inp_flags & INP_LOWPORT)
1393 					optval = IP_PORTRANGE_LOW;
1394 				else
1395 					optval = 0;
1396 				break;
1397 
1398 			case IP_ONESBCAST:
1399 				optval = OPTBIT(INP_ONESBCAST);
1400 				break;
1401 			case IP_DONTFRAG:
1402 				optval = OPTBIT(INP_DONTFRAG);
1403 				break;
1404 			case IP_BINDANY:
1405 				optval = OPTBIT(INP_BINDANY);
1406 				break;
1407 			case IP_RECVTOS:
1408 				optval = OPTBIT(INP_RECVTOS);
1409 				break;
1410 			case IP_FLOWID:
1411 				optval = inp->inp_flowid;
1412 				break;
1413 			case IP_FLOWTYPE:
1414 				optval = inp->inp_flowtype;
1415 				break;
1416 			case IP_RECVFLOWID:
1417 				optval = OPTBIT2(INP_RECVFLOWID);
1418 				break;
1419 #ifdef	RSS
1420 			case IP_RSSBUCKETID:
1421 				retval = rss_hash2bucket(inp->inp_flowid,
1422 				    inp->inp_flowtype,
1423 				    &rss_bucket);
1424 				if (retval == 0)
1425 					optval = rss_bucket;
1426 				else
1427 					error = EINVAL;
1428 				break;
1429 			case IP_RECVRSSBUCKETID:
1430 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1431 				break;
1432 #endif
1433 			case IP_BINDMULTI:
1434 				optval = OPTBIT2(INP_BINDMULTI);
1435 				break;
1436 			}
1437 			error = sooptcopyout(sopt, &optval, sizeof optval);
1438 			break;
1439 
1440 		/*
1441 		 * Multicast socket options are processed by the in_mcast
1442 		 * module.
1443 		 */
1444 		case IP_MULTICAST_IF:
1445 		case IP_MULTICAST_VIF:
1446 		case IP_MULTICAST_TTL:
1447 		case IP_MULTICAST_LOOP:
1448 		case IP_MSFILTER:
1449 			error = inp_getmoptions(inp, sopt);
1450 			break;
1451 
1452 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1453 		case IP_IPSEC_POLICY:
1454 			if (IPSEC_ENABLED(ipv4)) {
1455 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1456 				break;
1457 			}
1458 			/* FALLTHROUGH */
1459 #endif /* IPSEC */
1460 
1461 		default:
1462 			error = ENOPROTOOPT;
1463 			break;
1464 		}
1465 		break;
1466 	}
1467 	return (error);
1468 }
1469 
1470 /*
1471  * Routine called from ip_output() to loop back a copy of an IP multicast
1472  * packet to the input queue of a specified interface.  Note that this
1473  * calls the output routine of the loopback "driver", but with an interface
1474  * pointer that might NOT be a loopback interface -- evil, but easier than
1475  * replicating that code here.
1476  */
1477 static void
1478 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1479 {
1480 	struct ip *ip;
1481 	struct mbuf *copym;
1482 
1483 	/*
1484 	 * Make a deep copy of the packet because we're going to
1485 	 * modify the pack in order to generate checksums.
1486 	 */
1487 	copym = m_dup(m, M_NOWAIT);
1488 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1489 		copym = m_pullup(copym, hlen);
1490 	if (copym != NULL) {
1491 		/* If needed, compute the checksum and mark it as valid. */
1492 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1493 			in_delayed_cksum(copym);
1494 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1495 			copym->m_pkthdr.csum_flags |=
1496 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1497 			copym->m_pkthdr.csum_data = 0xffff;
1498 		}
1499 		/*
1500 		 * We don't bother to fragment if the IP length is greater
1501 		 * than the interface's MTU.  Can this possibly matter?
1502 		 */
1503 		ip = mtod(copym, struct ip *);
1504 		ip->ip_sum = 0;
1505 		ip->ip_sum = in_cksum(copym, hlen);
1506 		if_simloop(ifp, copym, AF_INET, 0);
1507 	}
1508 }
1509