xref: /freebsd/sys/netinet/ip_output.c (revision 2ef9ff7dd34a78a7890ba4d6de64da34d9c10942)
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_ratelimit.h"
39 #include "opt_ipsec.h"
40 #include "opt_mbuf_stress_test.h"
41 #include "opt_mpath.h"
42 #include "opt_route.h"
43 #include "opt_sctp.h"
44 #include "opt_rss.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 /*
208  * IP output.  The packet in mbuf chain m contains a skeletal IP
209  * header (with len, off, ttl, proto, tos, src, dst).
210  * The mbuf chain containing the packet will be freed.
211  * The mbuf opt, if present, will not be freed.
212  * If route ro is present and has ro_rt initialized, route lookup would be
213  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
214  * then result of route lookup is stored in ro->ro_rt.
215  *
216  * In the IP forwarding case, the packet will arrive with options already
217  * inserted, so must have a NULL opt pointer.
218  */
219 int
220 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
221     struct ip_moptions *imo, struct inpcb *inp)
222 {
223 	struct rm_priotracker in_ifa_tracker;
224 	struct epoch_tracker et;
225 	struct ip *ip;
226 	struct ifnet *ifp = NULL;	/* keep compiler happy */
227 	struct mbuf *m0;
228 	int hlen = sizeof (struct ip);
229 	int mtu;
230 	int error = 0;
231 	struct sockaddr_in *dst, sin;
232 	const struct sockaddr_in *gw;
233 	struct in_ifaddr *ia;
234 	struct in_addr src;
235 	int isbroadcast;
236 	uint16_t ip_len, ip_off;
237 	uint32_t fibnum;
238 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
239 	int no_route_but_check_spd = 0;
240 #endif
241 	M_ASSERTPKTHDR(m);
242 
243 	if (inp != NULL) {
244 		INP_LOCK_ASSERT(inp);
245 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
246 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
247 			m->m_pkthdr.flowid = inp->inp_flowid;
248 			M_HASHTYPE_SET(m, inp->inp_flowtype);
249 		}
250 #ifdef NUMA
251 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
252 #endif
253 	}
254 
255 	if (opt) {
256 		int len = 0;
257 		m = ip_insertoptions(m, opt, &len);
258 		if (len != 0)
259 			hlen = len; /* ip->ip_hl is updated above */
260 	}
261 	ip = mtod(m, struct ip *);
262 	ip_len = ntohs(ip->ip_len);
263 	ip_off = ntohs(ip->ip_off);
264 
265 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
266 		ip->ip_v = IPVERSION;
267 		ip->ip_hl = hlen >> 2;
268 		ip_fillid(ip);
269 	} else {
270 		/* Header already set, fetch hlen from there */
271 		hlen = ip->ip_hl << 2;
272 	}
273 	if ((flags & IP_FORWARDING) == 0)
274 		IPSTAT_INC(ips_localout);
275 
276 	/*
277 	 * dst/gw handling:
278 	 *
279 	 * gw is readonly but can point either to dst OR rt_gateway,
280 	 * therefore we need restore gw if we're redoing lookup.
281 	 */
282 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
283 	if (ro != NULL)
284 		dst = (struct sockaddr_in *)&ro->ro_dst;
285 	else
286 		dst = &sin;
287 	if (ro == NULL || ro->ro_rt == NULL) {
288 		bzero(dst, sizeof(*dst));
289 		dst->sin_family = AF_INET;
290 		dst->sin_len = sizeof(*dst);
291 		dst->sin_addr = ip->ip_dst;
292 	}
293 	gw = dst;
294 	NET_EPOCH_ENTER(et);
295 again:
296 	/*
297 	 * Validate route against routing table additions;
298 	 * a better/more specific route might have been added.
299 	 */
300 	if (inp != NULL && ro != NULL && ro->ro_rt != NULL)
301 		RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
302 	/*
303 	 * If there is a cached route,
304 	 * check that it is to the same destination
305 	 * and is still up.  If not, free it and try again.
306 	 * The address family should also be checked in case of sharing the
307 	 * cache with IPv6.
308 	 * Also check whether routing cache needs invalidation.
309 	 */
310 	if (ro != NULL && ro->ro_rt != NULL &&
311 	    ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
312 	    ro->ro_rt->rt_ifp == NULL || !RT_LINK_IS_UP(ro->ro_rt->rt_ifp) ||
313 	    dst->sin_family != AF_INET ||
314 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
315 		RO_INVALIDATE_CACHE(ro);
316 	ia = NULL;
317 	/*
318 	 * If routing to interface only, short circuit routing lookup.
319 	 * The use of an all-ones broadcast address implies this; an
320 	 * interface is specified by the broadcast address of an interface,
321 	 * or the destination address of a ptp interface.
322 	 */
323 	if (flags & IP_SENDONES) {
324 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
325 						      M_GETFIB(m)))) == NULL &&
326 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
327 						    M_GETFIB(m)))) == NULL) {
328 			IPSTAT_INC(ips_noroute);
329 			error = ENETUNREACH;
330 			goto bad;
331 		}
332 		ip->ip_dst.s_addr = INADDR_BROADCAST;
333 		dst->sin_addr = ip->ip_dst;
334 		ifp = ia->ia_ifp;
335 		mtu = ifp->if_mtu;
336 		ip->ip_ttl = 1;
337 		isbroadcast = 1;
338 		src = IA_SIN(ia)->sin_addr;
339 	} else if (flags & IP_ROUTETOIF) {
340 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
341 						    M_GETFIB(m)))) == NULL &&
342 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
343 						M_GETFIB(m)))) == NULL) {
344 			IPSTAT_INC(ips_noroute);
345 			error = ENETUNREACH;
346 			goto bad;
347 		}
348 		ifp = ia->ia_ifp;
349 		mtu = ifp->if_mtu;
350 		ip->ip_ttl = 1;
351 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
352 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
353 		src = IA_SIN(ia)->sin_addr;
354 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
355 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
356 		/*
357 		 * Bypass the normal routing lookup for multicast
358 		 * packets if the interface is specified.
359 		 */
360 		ifp = imo->imo_multicast_ifp;
361 		mtu = ifp->if_mtu;
362 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
363 		isbroadcast = 0;	/* fool gcc */
364 		/* Interface may have no addresses. */
365 		if (ia != NULL)
366 			src = IA_SIN(ia)->sin_addr;
367 		else
368 			src.s_addr = INADDR_ANY;
369 	} else if (ro != NULL) {
370 		if (ro->ro_rt == NULL) {
371 			/*
372 			 * We want to do any cloning requested by the link
373 			 * layer, as this is probably required in all cases
374 			 * for correct operation (as it is for ARP).
375 			 */
376 #ifdef RADIX_MPATH
377 			rtalloc_mpath_fib(ro,
378 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
379 			    fibnum);
380 #else
381 			in_rtalloc_ign(ro, 0, fibnum);
382 #endif
383 			if (ro->ro_rt == NULL ||
384 			    (ro->ro_rt->rt_flags & RTF_UP) == 0 ||
385 			    ro->ro_rt->rt_ifp == NULL ||
386 			    !RT_LINK_IS_UP(ro->ro_rt->rt_ifp)) {
387 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
388 				/*
389 				 * There is no route for this packet, but it is
390 				 * possible that a matching SPD entry exists.
391 				 */
392 				no_route_but_check_spd = 1;
393 				mtu = 0; /* Silence GCC warning. */
394 				goto sendit;
395 #endif
396 				IPSTAT_INC(ips_noroute);
397 				error = EHOSTUNREACH;
398 				goto bad;
399 			}
400 		}
401 		ia = ifatoia(ro->ro_rt->rt_ifa);
402 		ifp = ro->ro_rt->rt_ifp;
403 		counter_u64_add(ro->ro_rt->rt_pksent, 1);
404 		rt_update_ro_flags(ro);
405 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
406 			gw = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
407 		if (ro->ro_rt->rt_flags & RTF_HOST)
408 			isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
409 		else if (ifp->if_flags & IFF_BROADCAST)
410 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
411 		else
412 			isbroadcast = 0;
413 		if (ro->ro_rt->rt_flags & RTF_HOST)
414 			mtu = ro->ro_rt->rt_mtu;
415 		else
416 			mtu = ifp->if_mtu;
417 		src = IA_SIN(ia)->sin_addr;
418 	} else {
419 		struct nhop4_extended nh;
420 
421 		bzero(&nh, sizeof(nh));
422 		if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh) !=
423 		    0) {
424 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
425 			/*
426 			 * There is no route for this packet, but it is
427 			 * possible that a matching SPD entry exists.
428 			 */
429 			no_route_but_check_spd = 1;
430 			mtu = 0; /* Silence GCC warning. */
431 			goto sendit;
432 #endif
433 			IPSTAT_INC(ips_noroute);
434 			error = EHOSTUNREACH;
435 			goto bad;
436 		}
437 		ifp = nh.nh_ifp;
438 		mtu = nh.nh_mtu;
439 		/*
440 		 * We are rewriting here dst to be gw actually, contradicting
441 		 * comment at the beginning of the function. However, in this
442 		 * case we are always dealing with on stack dst.
443 		 * In case if pfil(9) sends us back to beginning of the
444 		 * function, the dst would be rewritten by ip_output_pfil().
445 		 */
446 		MPASS(dst == &sin);
447 		dst->sin_addr = nh.nh_addr;
448 		ia = nh.nh_ia;
449 		src = nh.nh_src;
450 		isbroadcast = (((nh.nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
451 		    (NHF_HOST | NHF_BROADCAST)) ||
452 		    ((ifp->if_flags & IFF_BROADCAST) &&
453 		    in_ifaddr_broadcast(dst->sin_addr, ia)));
454 	}
455 
456 	/* Catch a possible divide by zero later. */
457 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (rt_flags=0x%08x) ifp=%p",
458 	    __func__, mtu, ro,
459 	    (ro != NULL && ro->ro_rt != NULL) ? ro->ro_rt->rt_flags : 0, ifp));
460 
461 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
462 		m->m_flags |= M_MCAST;
463 		/*
464 		 * IP destination address is multicast.  Make sure "gw"
465 		 * still points to the address in "ro".  (It may have been
466 		 * changed to point to a gateway address, above.)
467 		 */
468 		gw = dst;
469 		/*
470 		 * See if the caller provided any multicast options
471 		 */
472 		if (imo != NULL) {
473 			ip->ip_ttl = imo->imo_multicast_ttl;
474 			if (imo->imo_multicast_vif != -1)
475 				ip->ip_src.s_addr =
476 				    ip_mcast_src ?
477 				    ip_mcast_src(imo->imo_multicast_vif) :
478 				    INADDR_ANY;
479 		} else
480 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
481 		/*
482 		 * Confirm that the outgoing interface supports multicast.
483 		 */
484 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
485 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
486 				IPSTAT_INC(ips_noroute);
487 				error = ENETUNREACH;
488 				goto bad;
489 			}
490 		}
491 		/*
492 		 * If source address not specified yet, use address
493 		 * of outgoing interface.
494 		 */
495 		if (ip->ip_src.s_addr == INADDR_ANY)
496 			ip->ip_src = src;
497 
498 		if ((imo == NULL && in_mcast_loop) ||
499 		    (imo && imo->imo_multicast_loop)) {
500 			/*
501 			 * Loop back multicast datagram if not expressly
502 			 * forbidden to do so, even if we are not a member
503 			 * of the group; ip_input() will filter it later,
504 			 * thus deferring a hash lookup and mutex acquisition
505 			 * at the expense of a cheap copy using m_copym().
506 			 */
507 			ip_mloopback(ifp, m, hlen);
508 		} else {
509 			/*
510 			 * If we are acting as a multicast router, perform
511 			 * multicast forwarding as if the packet had just
512 			 * arrived on the interface to which we are about
513 			 * to send.  The multicast forwarding function
514 			 * recursively calls this function, using the
515 			 * IP_FORWARDING flag to prevent infinite recursion.
516 			 *
517 			 * Multicasts that are looped back by ip_mloopback(),
518 			 * above, will be forwarded by the ip_input() routine,
519 			 * if necessary.
520 			 */
521 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
522 				/*
523 				 * If rsvp daemon is not running, do not
524 				 * set ip_moptions. This ensures that the packet
525 				 * is multicast and not just sent down one link
526 				 * as prescribed by rsvpd.
527 				 */
528 				if (!V_rsvp_on)
529 					imo = NULL;
530 				if (ip_mforward &&
531 				    ip_mforward(ip, ifp, m, imo) != 0) {
532 					m_freem(m);
533 					goto done;
534 				}
535 			}
536 		}
537 
538 		/*
539 		 * Multicasts with a time-to-live of zero may be looped-
540 		 * back, above, but must not be transmitted on a network.
541 		 * Also, multicasts addressed to the loopback interface
542 		 * are not sent -- the above call to ip_mloopback() will
543 		 * loop back a copy. ip_input() will drop the copy if
544 		 * this host does not belong to the destination group on
545 		 * the loopback interface.
546 		 */
547 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
548 			m_freem(m);
549 			goto done;
550 		}
551 
552 		goto sendit;
553 	}
554 
555 	/*
556 	 * If the source address is not specified yet, use the address
557 	 * of the outoing interface.
558 	 */
559 	if (ip->ip_src.s_addr == INADDR_ANY)
560 		ip->ip_src = src;
561 
562 	/*
563 	 * Look for broadcast address and
564 	 * verify user is allowed to send
565 	 * such a packet.
566 	 */
567 	if (isbroadcast) {
568 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
569 			error = EADDRNOTAVAIL;
570 			goto bad;
571 		}
572 		if ((flags & IP_ALLOWBROADCAST) == 0) {
573 			error = EACCES;
574 			goto bad;
575 		}
576 		/* don't allow broadcast messages to be fragmented */
577 		if (ip_len > mtu) {
578 			error = EMSGSIZE;
579 			goto bad;
580 		}
581 		m->m_flags |= M_BCAST;
582 	} else {
583 		m->m_flags &= ~M_BCAST;
584 	}
585 
586 sendit:
587 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
588 	if (IPSEC_ENABLED(ipv4)) {
589 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
590 			if (error == EINPROGRESS)
591 				error = 0;
592 			goto done;
593 		}
594 	}
595 	/*
596 	 * Check if there was a route for this packet; return error if not.
597 	 */
598 	if (no_route_but_check_spd) {
599 		IPSTAT_INC(ips_noroute);
600 		error = EHOSTUNREACH;
601 		goto bad;
602 	}
603 	/* Update variables that are affected by ipsec4_output(). */
604 	ip = mtod(m, struct ip *);
605 	hlen = ip->ip_hl << 2;
606 #endif /* IPSEC */
607 
608 	/* Jump over all PFIL processing if hooks are not active. */
609 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
610 		switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
611 		case 1: /* Finished */
612 			goto done;
613 
614 		case 0: /* Continue normally */
615 			ip = mtod(m, struct ip *);
616 			break;
617 
618 		case -1: /* Need to try again */
619 			/* Reset everything for a new round */
620 			if (ro != NULL) {
621 				RO_RTFREE(ro);
622 				ro->ro_prepend = NULL;
623 			}
624 			gw = dst;
625 			ip = mtod(m, struct ip *);
626 			goto again;
627 
628 		}
629 	}
630 
631 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
632 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
633 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
634 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
635 			IPSTAT_INC(ips_badaddr);
636 			error = EADDRNOTAVAIL;
637 			goto bad;
638 		}
639 	}
640 
641 	m->m_pkthdr.csum_flags |= CSUM_IP;
642 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
643 		in_delayed_cksum(m);
644 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
645 	}
646 #ifdef SCTP
647 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
648 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
649 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
650 	}
651 #endif
652 
653 	/*
654 	 * If small enough for interface, or the interface will take
655 	 * care of the fragmentation for us, we can just send directly.
656 	 */
657 	if (ip_len <= mtu ||
658 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
659 		ip->ip_sum = 0;
660 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
661 			ip->ip_sum = in_cksum(m, hlen);
662 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
663 		}
664 
665 		/*
666 		 * Record statistics for this interface address.
667 		 * With CSUM_TSO the byte/packet count will be slightly
668 		 * incorrect because we count the IP+TCP headers only
669 		 * once instead of for every generated packet.
670 		 */
671 		if (!(flags & IP_FORWARDING) && ia) {
672 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
673 				counter_u64_add(ia->ia_ifa.ifa_opackets,
674 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
675 			else
676 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
677 
678 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
679 		}
680 #ifdef MBUF_STRESS_TEST
681 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
682 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
683 #endif
684 		/*
685 		 * Reset layer specific mbuf flags
686 		 * to avoid confusing lower layers.
687 		 */
688 		m_clrprotoflags(m);
689 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
690 #ifdef RATELIMIT
691 		if (inp != NULL) {
692 			if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
693 				in_pcboutput_txrtlmt(inp, ifp, m);
694 			/* stamp send tag on mbuf */
695 			m->m_pkthdr.snd_tag = inp->inp_snd_tag;
696 		} else {
697 			m->m_pkthdr.snd_tag = NULL;
698 		}
699 #endif
700 		error = (*ifp->if_output)(ifp, m,
701 		    (const struct sockaddr *)gw, ro);
702 #ifdef RATELIMIT
703 		/* check for route change */
704 		if (error == EAGAIN)
705 			in_pcboutput_eagain(inp);
706 #endif
707 		goto done;
708 	}
709 
710 	/* Balk when DF bit is set or the interface didn't support TSO. */
711 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
712 		error = EMSGSIZE;
713 		IPSTAT_INC(ips_cantfrag);
714 		goto bad;
715 	}
716 
717 	/*
718 	 * Too large for interface; fragment if possible. If successful,
719 	 * on return, m will point to a list of packets to be sent.
720 	 */
721 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
722 	if (error)
723 		goto bad;
724 	for (; m; m = m0) {
725 		m0 = m->m_nextpkt;
726 		m->m_nextpkt = 0;
727 		if (error == 0) {
728 			/* Record statistics for this interface address. */
729 			if (ia != NULL) {
730 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
731 				counter_u64_add(ia->ia_ifa.ifa_obytes,
732 				    m->m_pkthdr.len);
733 			}
734 			/*
735 			 * Reset layer specific mbuf flags
736 			 * to avoid confusing upper layers.
737 			 */
738 			m_clrprotoflags(m);
739 
740 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
741 			    mtod(m, struct ip *), NULL);
742 #ifdef RATELIMIT
743 			if (inp != NULL) {
744 				if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
745 					in_pcboutput_txrtlmt(inp, ifp, m);
746 				/* stamp send tag on mbuf */
747 				m->m_pkthdr.snd_tag = inp->inp_snd_tag;
748 			} else {
749 				m->m_pkthdr.snd_tag = NULL;
750 			}
751 #endif
752 			error = (*ifp->if_output)(ifp, m,
753 			    (const struct sockaddr *)gw, ro);
754 #ifdef RATELIMIT
755 			/* check for route change */
756 			if (error == EAGAIN)
757 				in_pcboutput_eagain(inp);
758 #endif
759 		} else
760 			m_freem(m);
761 	}
762 
763 	if (error == 0)
764 		IPSTAT_INC(ips_fragmented);
765 
766 done:
767 	NET_EPOCH_EXIT(et);
768 	return (error);
769  bad:
770 	m_freem(m);
771 	goto done;
772 }
773 
774 /*
775  * Create a chain of fragments which fit the given mtu. m_frag points to the
776  * mbuf to be fragmented; on return it points to the chain with the fragments.
777  * Return 0 if no error. If error, m_frag may contain a partially built
778  * chain of fragments that should be freed by the caller.
779  *
780  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
781  */
782 int
783 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
784     u_long if_hwassist_flags)
785 {
786 	int error = 0;
787 	int hlen = ip->ip_hl << 2;
788 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
789 	int off;
790 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
791 	int firstlen;
792 	struct mbuf **mnext;
793 	int nfrags;
794 	uint16_t ip_len, ip_off;
795 
796 	ip_len = ntohs(ip->ip_len);
797 	ip_off = ntohs(ip->ip_off);
798 
799 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
800 		IPSTAT_INC(ips_cantfrag);
801 		return EMSGSIZE;
802 	}
803 
804 	/*
805 	 * Must be able to put at least 8 bytes per fragment.
806 	 */
807 	if (len < 8)
808 		return EMSGSIZE;
809 
810 	/*
811 	 * If the interface will not calculate checksums on
812 	 * fragmented packets, then do it here.
813 	 */
814 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
815 		in_delayed_cksum(m0);
816 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
817 	}
818 #ifdef SCTP
819 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
820 		sctp_delayed_cksum(m0, hlen);
821 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
822 	}
823 #endif
824 	if (len > PAGE_SIZE) {
825 		/*
826 		 * Fragment large datagrams such that each segment
827 		 * contains a multiple of PAGE_SIZE amount of data,
828 		 * plus headers. This enables a receiver to perform
829 		 * page-flipping zero-copy optimizations.
830 		 *
831 		 * XXX When does this help given that sender and receiver
832 		 * could have different page sizes, and also mtu could
833 		 * be less than the receiver's page size ?
834 		 */
835 		int newlen;
836 
837 		off = MIN(mtu, m0->m_pkthdr.len);
838 
839 		/*
840 		 * firstlen (off - hlen) must be aligned on an
841 		 * 8-byte boundary
842 		 */
843 		if (off < hlen)
844 			goto smart_frag_failure;
845 		off = ((off - hlen) & ~7) + hlen;
846 		newlen = (~PAGE_MASK) & mtu;
847 		if ((newlen + sizeof (struct ip)) > mtu) {
848 			/* we failed, go back the default */
849 smart_frag_failure:
850 			newlen = len;
851 			off = hlen + len;
852 		}
853 		len = newlen;
854 
855 	} else {
856 		off = hlen + len;
857 	}
858 
859 	firstlen = off - hlen;
860 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
861 
862 	/*
863 	 * Loop through length of segment after first fragment,
864 	 * make new header and copy data of each part and link onto chain.
865 	 * Here, m0 is the original packet, m is the fragment being created.
866 	 * The fragments are linked off the m_nextpkt of the original
867 	 * packet, which after processing serves as the first fragment.
868 	 */
869 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
870 		struct ip *mhip;	/* ip header on the fragment */
871 		struct mbuf *m;
872 		int mhlen = sizeof (struct ip);
873 
874 		m = m_gethdr(M_NOWAIT, MT_DATA);
875 		if (m == NULL) {
876 			error = ENOBUFS;
877 			IPSTAT_INC(ips_odropped);
878 			goto done;
879 		}
880 		/*
881 		 * Make sure the complete packet header gets copied
882 		 * from the originating mbuf to the newly created
883 		 * mbuf. This also ensures that existing firewall
884 		 * classification(s), VLAN tags and so on get copied
885 		 * to the resulting fragmented packet(s):
886 		 */
887 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
888 			m_free(m);
889 			error = ENOBUFS;
890 			IPSTAT_INC(ips_odropped);
891 			goto done;
892 		}
893 		/*
894 		 * In the first mbuf, leave room for the link header, then
895 		 * copy the original IP header including options. The payload
896 		 * goes into an additional mbuf chain returned by m_copym().
897 		 */
898 		m->m_data += max_linkhdr;
899 		mhip = mtod(m, struct ip *);
900 		*mhip = *ip;
901 		if (hlen > sizeof (struct ip)) {
902 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
903 			mhip->ip_v = IPVERSION;
904 			mhip->ip_hl = mhlen >> 2;
905 		}
906 		m->m_len = mhlen;
907 		/* XXX do we need to add ip_off below ? */
908 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
909 		if (off + len >= ip_len)
910 			len = ip_len - off;
911 		else
912 			mhip->ip_off |= IP_MF;
913 		mhip->ip_len = htons((u_short)(len + mhlen));
914 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
915 		if (m->m_next == NULL) {	/* copy failed */
916 			m_free(m);
917 			error = ENOBUFS;	/* ??? */
918 			IPSTAT_INC(ips_odropped);
919 			goto done;
920 		}
921 		m->m_pkthdr.len = mhlen + len;
922 #ifdef MAC
923 		mac_netinet_fragment(m0, m);
924 #endif
925 		mhip->ip_off = htons(mhip->ip_off);
926 		mhip->ip_sum = 0;
927 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
928 			mhip->ip_sum = in_cksum(m, mhlen);
929 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
930 		}
931 		*mnext = m;
932 		mnext = &m->m_nextpkt;
933 	}
934 	IPSTAT_ADD(ips_ofragments, nfrags);
935 
936 	/*
937 	 * Update first fragment by trimming what's been copied out
938 	 * and updating header.
939 	 */
940 	m_adj(m0, hlen + firstlen - ip_len);
941 	m0->m_pkthdr.len = hlen + firstlen;
942 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
943 	ip->ip_off = htons(ip_off | IP_MF);
944 	ip->ip_sum = 0;
945 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
946 		ip->ip_sum = in_cksum(m0, hlen);
947 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
948 	}
949 
950 done:
951 	*m_frag = m0;
952 	return error;
953 }
954 
955 void
956 in_delayed_cksum(struct mbuf *m)
957 {
958 	struct ip *ip;
959 	struct udphdr *uh;
960 	uint16_t cklen, csum, offset;
961 
962 	ip = mtod(m, struct ip *);
963 	offset = ip->ip_hl << 2 ;
964 
965 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
966 		/* if udp header is not in the first mbuf copy udplen */
967 		if (offset + sizeof(struct udphdr) > m->m_len) {
968 			m_copydata(m, offset + offsetof(struct udphdr,
969 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
970 			cklen = ntohs(cklen);
971 		} else {
972 			uh = (struct udphdr *)mtodo(m, offset);
973 			cklen = ntohs(uh->uh_ulen);
974 		}
975 		csum = in_cksum_skip(m, cklen + offset, offset);
976 		if (csum == 0)
977 			csum = 0xffff;
978 	} else {
979 		cklen = ntohs(ip->ip_len);
980 		csum = in_cksum_skip(m, cklen, offset);
981 	}
982 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
983 
984 	if (offset + sizeof(csum) > m->m_len)
985 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
986 	else
987 		*(u_short *)mtodo(m, offset) = csum;
988 }
989 
990 /*
991  * IP socket option processing.
992  */
993 int
994 ip_ctloutput(struct socket *so, struct sockopt *sopt)
995 {
996 	struct	inpcb *inp = sotoinpcb(so);
997 	int	error, optval;
998 #ifdef	RSS
999 	uint32_t rss_bucket;
1000 	int retval;
1001 #endif
1002 
1003 	error = optval = 0;
1004 	if (sopt->sopt_level != IPPROTO_IP) {
1005 		error = EINVAL;
1006 
1007 		if (sopt->sopt_level == SOL_SOCKET &&
1008 		    sopt->sopt_dir == SOPT_SET) {
1009 			switch (sopt->sopt_name) {
1010 			case SO_REUSEADDR:
1011 				INP_WLOCK(inp);
1012 				if ((so->so_options & SO_REUSEADDR) != 0)
1013 					inp->inp_flags2 |= INP_REUSEADDR;
1014 				else
1015 					inp->inp_flags2 &= ~INP_REUSEADDR;
1016 				INP_WUNLOCK(inp);
1017 				error = 0;
1018 				break;
1019 			case SO_REUSEPORT:
1020 				INP_WLOCK(inp);
1021 				if ((so->so_options & SO_REUSEPORT) != 0)
1022 					inp->inp_flags2 |= INP_REUSEPORT;
1023 				else
1024 					inp->inp_flags2 &= ~INP_REUSEPORT;
1025 				INP_WUNLOCK(inp);
1026 				error = 0;
1027 				break;
1028 			case SO_REUSEPORT_LB:
1029 				INP_WLOCK(inp);
1030 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1031 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1032 				else
1033 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1034 				INP_WUNLOCK(inp);
1035 				error = 0;
1036 				break;
1037 			case SO_SETFIB:
1038 				INP_WLOCK(inp);
1039 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1040 				INP_WUNLOCK(inp);
1041 				error = 0;
1042 				break;
1043 			case SO_MAX_PACING_RATE:
1044 #ifdef RATELIMIT
1045 				INP_WLOCK(inp);
1046 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1047 				INP_WUNLOCK(inp);
1048 				error = 0;
1049 #else
1050 				error = EOPNOTSUPP;
1051 #endif
1052 				break;
1053 			default:
1054 				break;
1055 			}
1056 		}
1057 		return (error);
1058 	}
1059 
1060 	switch (sopt->sopt_dir) {
1061 	case SOPT_SET:
1062 		switch (sopt->sopt_name) {
1063 		case IP_OPTIONS:
1064 #ifdef notyet
1065 		case IP_RETOPTS:
1066 #endif
1067 		{
1068 			struct mbuf *m;
1069 			if (sopt->sopt_valsize > MLEN) {
1070 				error = EMSGSIZE;
1071 				break;
1072 			}
1073 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1074 			if (m == NULL) {
1075 				error = ENOBUFS;
1076 				break;
1077 			}
1078 			m->m_len = sopt->sopt_valsize;
1079 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1080 					    m->m_len);
1081 			if (error) {
1082 				m_free(m);
1083 				break;
1084 			}
1085 			INP_WLOCK(inp);
1086 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1087 			INP_WUNLOCK(inp);
1088 			return (error);
1089 		}
1090 
1091 		case IP_BINDANY:
1092 			if (sopt->sopt_td != NULL) {
1093 				error = priv_check(sopt->sopt_td,
1094 				    PRIV_NETINET_BINDANY);
1095 				if (error)
1096 					break;
1097 			}
1098 			/* FALLTHROUGH */
1099 		case IP_BINDMULTI:
1100 #ifdef	RSS
1101 		case IP_RSS_LISTEN_BUCKET:
1102 #endif
1103 		case IP_TOS:
1104 		case IP_TTL:
1105 		case IP_MINTTL:
1106 		case IP_RECVOPTS:
1107 		case IP_RECVRETOPTS:
1108 		case IP_ORIGDSTADDR:
1109 		case IP_RECVDSTADDR:
1110 		case IP_RECVTTL:
1111 		case IP_RECVIF:
1112 		case IP_ONESBCAST:
1113 		case IP_DONTFRAG:
1114 		case IP_RECVTOS:
1115 		case IP_RECVFLOWID:
1116 #ifdef	RSS
1117 		case IP_RECVRSSBUCKETID:
1118 #endif
1119 			error = sooptcopyin(sopt, &optval, sizeof optval,
1120 					    sizeof optval);
1121 			if (error)
1122 				break;
1123 
1124 			switch (sopt->sopt_name) {
1125 			case IP_TOS:
1126 				inp->inp_ip_tos = optval;
1127 				break;
1128 
1129 			case IP_TTL:
1130 				inp->inp_ip_ttl = optval;
1131 				break;
1132 
1133 			case IP_MINTTL:
1134 				if (optval >= 0 && optval <= MAXTTL)
1135 					inp->inp_ip_minttl = optval;
1136 				else
1137 					error = EINVAL;
1138 				break;
1139 
1140 #define	OPTSET(bit) do {						\
1141 	INP_WLOCK(inp);							\
1142 	if (optval)							\
1143 		inp->inp_flags |= bit;					\
1144 	else								\
1145 		inp->inp_flags &= ~bit;					\
1146 	INP_WUNLOCK(inp);						\
1147 } while (0)
1148 
1149 #define	OPTSET2(bit, val) do {						\
1150 	INP_WLOCK(inp);							\
1151 	if (val)							\
1152 		inp->inp_flags2 |= bit;					\
1153 	else								\
1154 		inp->inp_flags2 &= ~bit;				\
1155 	INP_WUNLOCK(inp);						\
1156 } while (0)
1157 
1158 			case IP_RECVOPTS:
1159 				OPTSET(INP_RECVOPTS);
1160 				break;
1161 
1162 			case IP_RECVRETOPTS:
1163 				OPTSET(INP_RECVRETOPTS);
1164 				break;
1165 
1166 			case IP_RECVDSTADDR:
1167 				OPTSET(INP_RECVDSTADDR);
1168 				break;
1169 
1170 			case IP_ORIGDSTADDR:
1171 				OPTSET2(INP_ORIGDSTADDR, optval);
1172 				break;
1173 
1174 			case IP_RECVTTL:
1175 				OPTSET(INP_RECVTTL);
1176 				break;
1177 
1178 			case IP_RECVIF:
1179 				OPTSET(INP_RECVIF);
1180 				break;
1181 
1182 			case IP_ONESBCAST:
1183 				OPTSET(INP_ONESBCAST);
1184 				break;
1185 			case IP_DONTFRAG:
1186 				OPTSET(INP_DONTFRAG);
1187 				break;
1188 			case IP_BINDANY:
1189 				OPTSET(INP_BINDANY);
1190 				break;
1191 			case IP_RECVTOS:
1192 				OPTSET(INP_RECVTOS);
1193 				break;
1194 			case IP_BINDMULTI:
1195 				OPTSET2(INP_BINDMULTI, optval);
1196 				break;
1197 			case IP_RECVFLOWID:
1198 				OPTSET2(INP_RECVFLOWID, optval);
1199 				break;
1200 #ifdef	RSS
1201 			case IP_RSS_LISTEN_BUCKET:
1202 				if ((optval >= 0) &&
1203 				    (optval < rss_getnumbuckets())) {
1204 					inp->inp_rss_listen_bucket = optval;
1205 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1206 				} else {
1207 					error = EINVAL;
1208 				}
1209 				break;
1210 			case IP_RECVRSSBUCKETID:
1211 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1212 				break;
1213 #endif
1214 			}
1215 			break;
1216 #undef OPTSET
1217 #undef OPTSET2
1218 
1219 		/*
1220 		 * Multicast socket options are processed by the in_mcast
1221 		 * module.
1222 		 */
1223 		case IP_MULTICAST_IF:
1224 		case IP_MULTICAST_VIF:
1225 		case IP_MULTICAST_TTL:
1226 		case IP_MULTICAST_LOOP:
1227 		case IP_ADD_MEMBERSHIP:
1228 		case IP_DROP_MEMBERSHIP:
1229 		case IP_ADD_SOURCE_MEMBERSHIP:
1230 		case IP_DROP_SOURCE_MEMBERSHIP:
1231 		case IP_BLOCK_SOURCE:
1232 		case IP_UNBLOCK_SOURCE:
1233 		case IP_MSFILTER:
1234 		case MCAST_JOIN_GROUP:
1235 		case MCAST_LEAVE_GROUP:
1236 		case MCAST_JOIN_SOURCE_GROUP:
1237 		case MCAST_LEAVE_SOURCE_GROUP:
1238 		case MCAST_BLOCK_SOURCE:
1239 		case MCAST_UNBLOCK_SOURCE:
1240 			error = inp_setmoptions(inp, sopt);
1241 			break;
1242 
1243 		case IP_PORTRANGE:
1244 			error = sooptcopyin(sopt, &optval, sizeof optval,
1245 					    sizeof optval);
1246 			if (error)
1247 				break;
1248 
1249 			INP_WLOCK(inp);
1250 			switch (optval) {
1251 			case IP_PORTRANGE_DEFAULT:
1252 				inp->inp_flags &= ~(INP_LOWPORT);
1253 				inp->inp_flags &= ~(INP_HIGHPORT);
1254 				break;
1255 
1256 			case IP_PORTRANGE_HIGH:
1257 				inp->inp_flags &= ~(INP_LOWPORT);
1258 				inp->inp_flags |= INP_HIGHPORT;
1259 				break;
1260 
1261 			case IP_PORTRANGE_LOW:
1262 				inp->inp_flags &= ~(INP_HIGHPORT);
1263 				inp->inp_flags |= INP_LOWPORT;
1264 				break;
1265 
1266 			default:
1267 				error = EINVAL;
1268 				break;
1269 			}
1270 			INP_WUNLOCK(inp);
1271 			break;
1272 
1273 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1274 		case IP_IPSEC_POLICY:
1275 			if (IPSEC_ENABLED(ipv4)) {
1276 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1277 				break;
1278 			}
1279 			/* FALLTHROUGH */
1280 #endif /* IPSEC */
1281 
1282 		default:
1283 			error = ENOPROTOOPT;
1284 			break;
1285 		}
1286 		break;
1287 
1288 	case SOPT_GET:
1289 		switch (sopt->sopt_name) {
1290 		case IP_OPTIONS:
1291 		case IP_RETOPTS:
1292 			INP_RLOCK(inp);
1293 			if (inp->inp_options) {
1294 				struct mbuf *options;
1295 
1296 				options = m_copym(inp->inp_options, 0,
1297 				    M_COPYALL, M_NOWAIT);
1298 				INP_RUNLOCK(inp);
1299 				if (options != NULL) {
1300 					error = sooptcopyout(sopt,
1301 							     mtod(options, char *),
1302 							     options->m_len);
1303 					m_freem(options);
1304 				} else
1305 					error = ENOMEM;
1306 			} else {
1307 				INP_RUNLOCK(inp);
1308 				sopt->sopt_valsize = 0;
1309 			}
1310 			break;
1311 
1312 		case IP_TOS:
1313 		case IP_TTL:
1314 		case IP_MINTTL:
1315 		case IP_RECVOPTS:
1316 		case IP_RECVRETOPTS:
1317 		case IP_ORIGDSTADDR:
1318 		case IP_RECVDSTADDR:
1319 		case IP_RECVTTL:
1320 		case IP_RECVIF:
1321 		case IP_PORTRANGE:
1322 		case IP_ONESBCAST:
1323 		case IP_DONTFRAG:
1324 		case IP_BINDANY:
1325 		case IP_RECVTOS:
1326 		case IP_BINDMULTI:
1327 		case IP_FLOWID:
1328 		case IP_FLOWTYPE:
1329 		case IP_RECVFLOWID:
1330 #ifdef	RSS
1331 		case IP_RSSBUCKETID:
1332 		case IP_RECVRSSBUCKETID:
1333 #endif
1334 			switch (sopt->sopt_name) {
1335 
1336 			case IP_TOS:
1337 				optval = inp->inp_ip_tos;
1338 				break;
1339 
1340 			case IP_TTL:
1341 				optval = inp->inp_ip_ttl;
1342 				break;
1343 
1344 			case IP_MINTTL:
1345 				optval = inp->inp_ip_minttl;
1346 				break;
1347 
1348 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1349 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1350 
1351 			case IP_RECVOPTS:
1352 				optval = OPTBIT(INP_RECVOPTS);
1353 				break;
1354 
1355 			case IP_RECVRETOPTS:
1356 				optval = OPTBIT(INP_RECVRETOPTS);
1357 				break;
1358 
1359 			case IP_RECVDSTADDR:
1360 				optval = OPTBIT(INP_RECVDSTADDR);
1361 				break;
1362 
1363 			case IP_ORIGDSTADDR:
1364 				optval = OPTBIT2(INP_ORIGDSTADDR);
1365 				break;
1366 
1367 			case IP_RECVTTL:
1368 				optval = OPTBIT(INP_RECVTTL);
1369 				break;
1370 
1371 			case IP_RECVIF:
1372 				optval = OPTBIT(INP_RECVIF);
1373 				break;
1374 
1375 			case IP_PORTRANGE:
1376 				if (inp->inp_flags & INP_HIGHPORT)
1377 					optval = IP_PORTRANGE_HIGH;
1378 				else if (inp->inp_flags & INP_LOWPORT)
1379 					optval = IP_PORTRANGE_LOW;
1380 				else
1381 					optval = 0;
1382 				break;
1383 
1384 			case IP_ONESBCAST:
1385 				optval = OPTBIT(INP_ONESBCAST);
1386 				break;
1387 			case IP_DONTFRAG:
1388 				optval = OPTBIT(INP_DONTFRAG);
1389 				break;
1390 			case IP_BINDANY:
1391 				optval = OPTBIT(INP_BINDANY);
1392 				break;
1393 			case IP_RECVTOS:
1394 				optval = OPTBIT(INP_RECVTOS);
1395 				break;
1396 			case IP_FLOWID:
1397 				optval = inp->inp_flowid;
1398 				break;
1399 			case IP_FLOWTYPE:
1400 				optval = inp->inp_flowtype;
1401 				break;
1402 			case IP_RECVFLOWID:
1403 				optval = OPTBIT2(INP_RECVFLOWID);
1404 				break;
1405 #ifdef	RSS
1406 			case IP_RSSBUCKETID:
1407 				retval = rss_hash2bucket(inp->inp_flowid,
1408 				    inp->inp_flowtype,
1409 				    &rss_bucket);
1410 				if (retval == 0)
1411 					optval = rss_bucket;
1412 				else
1413 					error = EINVAL;
1414 				break;
1415 			case IP_RECVRSSBUCKETID:
1416 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1417 				break;
1418 #endif
1419 			case IP_BINDMULTI:
1420 				optval = OPTBIT2(INP_BINDMULTI);
1421 				break;
1422 			}
1423 			error = sooptcopyout(sopt, &optval, sizeof optval);
1424 			break;
1425 
1426 		/*
1427 		 * Multicast socket options are processed by the in_mcast
1428 		 * module.
1429 		 */
1430 		case IP_MULTICAST_IF:
1431 		case IP_MULTICAST_VIF:
1432 		case IP_MULTICAST_TTL:
1433 		case IP_MULTICAST_LOOP:
1434 		case IP_MSFILTER:
1435 			error = inp_getmoptions(inp, sopt);
1436 			break;
1437 
1438 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1439 		case IP_IPSEC_POLICY:
1440 			if (IPSEC_ENABLED(ipv4)) {
1441 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1442 				break;
1443 			}
1444 			/* FALLTHROUGH */
1445 #endif /* IPSEC */
1446 
1447 		default:
1448 			error = ENOPROTOOPT;
1449 			break;
1450 		}
1451 		break;
1452 	}
1453 	return (error);
1454 }
1455 
1456 /*
1457  * Routine called from ip_output() to loop back a copy of an IP multicast
1458  * packet to the input queue of a specified interface.  Note that this
1459  * calls the output routine of the loopback "driver", but with an interface
1460  * pointer that might NOT be a loopback interface -- evil, but easier than
1461  * replicating that code here.
1462  */
1463 static void
1464 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1465 {
1466 	struct ip *ip;
1467 	struct mbuf *copym;
1468 
1469 	/*
1470 	 * Make a deep copy of the packet because we're going to
1471 	 * modify the pack in order to generate checksums.
1472 	 */
1473 	copym = m_dup(m, M_NOWAIT);
1474 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1475 		copym = m_pullup(copym, hlen);
1476 	if (copym != NULL) {
1477 		/* If needed, compute the checksum and mark it as valid. */
1478 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1479 			in_delayed_cksum(copym);
1480 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1481 			copym->m_pkthdr.csum_flags |=
1482 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1483 			copym->m_pkthdr.csum_data = 0xffff;
1484 		}
1485 		/*
1486 		 * We don't bother to fragment if the IP length is greater
1487 		 * than the interface's MTU.  Can this possibly matter?
1488 		 */
1489 		ip = mtod(copym, struct ip *);
1490 		ip->ip_sum = 0;
1491 		ip->ip_sum = in_cksum(copym, hlen);
1492 		if_simloop(ifp, copym, AF_INET, 0);
1493 	}
1494 }
1495