xref: /freebsd/sys/netinet/ip_output.c (revision 97cb52fa9aefd90fad38790fded50905aeeb9b9e)
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_kdtrace.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/ip.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/in_rss.h>
80 #include <netinet/in_var.h>
81 #include <netinet/ip_var.h>
82 #include <netinet/ip_options.h>
83 #ifdef SCTP
84 #include <netinet/sctp.h>
85 #include <netinet/sctp_crc32.h>
86 #endif
87 
88 #include <netipsec/ipsec_support.h>
89 
90 #include <machine/in_cksum.h>
91 
92 #include <security/mac/mac_framework.h>
93 
94 #ifdef MBUF_STRESS_TEST
95 static int mbuf_frag_size = 0;
96 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
97 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
98 #endif
99 
100 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
101 
102 
103 extern int in_mcast_loop;
104 extern	struct protosw inetsw[];
105 
106 static inline int
107 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
108     struct sockaddr_in *dst, int *fibnum, int *error)
109 {
110 	struct m_tag *fwd_tag = NULL;
111 	struct mbuf *m;
112 	struct in_addr odst;
113 	struct ip *ip;
114 
115 	m = *mp;
116 	ip = mtod(m, struct ip *);
117 
118 	/* Run through list of hooks for output packets. */
119 	odst.s_addr = ip->ip_dst.s_addr;
120 	*error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, inp);
121 	m = *mp;
122 	if ((*error) != 0 || m == NULL)
123 		return 1; /* Finished */
124 
125 	ip = mtod(m, struct ip *);
126 
127 	/* See if destination IP address was changed by packet filter. */
128 	if (odst.s_addr != ip->ip_dst.s_addr) {
129 		m->m_flags |= M_SKIP_FIREWALL;
130 		/* If destination is now ourself drop to ip_input(). */
131 		if (in_localip(ip->ip_dst)) {
132 			m->m_flags |= M_FASTFWD_OURS;
133 			if (m->m_pkthdr.rcvif == NULL)
134 				m->m_pkthdr.rcvif = V_loif;
135 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
136 				m->m_pkthdr.csum_flags |=
137 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
138 				m->m_pkthdr.csum_data = 0xffff;
139 			}
140 			m->m_pkthdr.csum_flags |=
141 				CSUM_IP_CHECKED | CSUM_IP_VALID;
142 #ifdef SCTP
143 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
144 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
145 #endif
146 			*error = netisr_queue(NETISR_IP, m);
147 			return 1; /* Finished */
148 		}
149 
150 		bzero(dst, sizeof(*dst));
151 		dst->sin_family = AF_INET;
152 		dst->sin_len = sizeof(*dst);
153 		dst->sin_addr = ip->ip_dst;
154 
155 		return -1; /* Reloop */
156 	}
157 	/* See if fib was changed by packet filter. */
158 	if ((*fibnum) != M_GETFIB(m)) {
159 		m->m_flags |= M_SKIP_FIREWALL;
160 		*fibnum = M_GETFIB(m);
161 		return -1; /* Reloop for FIB change */
162 	}
163 
164 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
165 	if (m->m_flags & M_FASTFWD_OURS) {
166 		if (m->m_pkthdr.rcvif == NULL)
167 			m->m_pkthdr.rcvif = V_loif;
168 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
169 			m->m_pkthdr.csum_flags |=
170 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
171 			m->m_pkthdr.csum_data = 0xffff;
172 		}
173 #ifdef SCTP
174 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
175 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
176 #endif
177 		m->m_pkthdr.csum_flags |=
178 			CSUM_IP_CHECKED | CSUM_IP_VALID;
179 
180 		*error = netisr_queue(NETISR_IP, m);
181 		return 1; /* Finished */
182 	}
183 	/* Or forward to some other address? */
184 	if ((m->m_flags & M_IP_NEXTHOP) &&
185 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
186 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
187 		m->m_flags |= M_SKIP_FIREWALL;
188 		m->m_flags &= ~M_IP_NEXTHOP;
189 		m_tag_delete(m, fwd_tag);
190 
191 		return -1; /* Reloop for CHANGE of dst */
192 	}
193 
194 	return 0;
195 }
196 
197 /*
198  * IP output.  The packet in mbuf chain m contains a skeletal IP
199  * header (with len, off, ttl, proto, tos, src, dst).
200  * The mbuf chain containing the packet will be freed.
201  * The mbuf opt, if present, will not be freed.
202  * If route ro is present and has ro_rt initialized, route lookup would be
203  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
204  * then result of route lookup is stored in ro->ro_rt.
205  *
206  * In the IP forwarding case, the packet will arrive with options already
207  * inserted, so must have a NULL opt pointer.
208  */
209 int
210 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
211     struct ip_moptions *imo, struct inpcb *inp)
212 {
213 	struct rm_priotracker in_ifa_tracker;
214 	struct ip *ip;
215 	struct ifnet *ifp = NULL;	/* keep compiler happy */
216 	struct mbuf *m0;
217 	int hlen = sizeof (struct ip);
218 	int mtu;
219 	int error = 0;
220 	struct sockaddr_in *dst;
221 	const struct sockaddr_in *gw;
222 	struct in_ifaddr *ia;
223 	int isbroadcast;
224 	uint16_t ip_len, ip_off;
225 	struct route iproute;
226 	struct rtentry *rte;	/* cache for ro->ro_rt */
227 	uint32_t fibnum;
228 	int have_ia_ref;
229 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
230 	int no_route_but_check_spd = 0;
231 #endif
232 	M_ASSERTPKTHDR(m);
233 
234 	if (inp != NULL) {
235 		INP_LOCK_ASSERT(inp);
236 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
237 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
238 			m->m_pkthdr.flowid = inp->inp_flowid;
239 			M_HASHTYPE_SET(m, inp->inp_flowtype);
240 		}
241 	}
242 
243 	if (ro == NULL) {
244 		ro = &iproute;
245 		bzero(ro, sizeof (*ro));
246 	}
247 
248 	if (opt) {
249 		int len = 0;
250 		m = ip_insertoptions(m, opt, &len);
251 		if (len != 0)
252 			hlen = len; /* ip->ip_hl is updated above */
253 	}
254 	ip = mtod(m, struct ip *);
255 	ip_len = ntohs(ip->ip_len);
256 	ip_off = ntohs(ip->ip_off);
257 
258 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
259 		ip->ip_v = IPVERSION;
260 		ip->ip_hl = hlen >> 2;
261 		ip_fillid(ip);
262 		IPSTAT_INC(ips_localout);
263 	} else {
264 		/* Header already set, fetch hlen from there */
265 		hlen = ip->ip_hl << 2;
266 	}
267 
268 	/*
269 	 * dst/gw handling:
270 	 *
271 	 * dst can be rewritten but always points to &ro->ro_dst.
272 	 * gw is readonly but can point either to dst OR rt_gateway,
273 	 * therefore we need restore gw if we're redoing lookup.
274 	 */
275 	gw = dst = (struct sockaddr_in *)&ro->ro_dst;
276 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
277 	rte = ro->ro_rt;
278 	if (rte == NULL) {
279 		bzero(dst, sizeof(*dst));
280 		dst->sin_family = AF_INET;
281 		dst->sin_len = sizeof(*dst);
282 		dst->sin_addr = ip->ip_dst;
283 	}
284 again:
285 	/*
286 	 * Validate route against routing table additions;
287 	 * a better/more specific route might have been added.
288 	 */
289 	if (inp)
290 		RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
291 	/*
292 	 * If there is a cached route,
293 	 * check that it is to the same destination
294 	 * and is still up.  If not, free it and try again.
295 	 * The address family should also be checked in case of sharing the
296 	 * cache with IPv6.
297 	 * Also check whether routing cache needs invalidation.
298 	 */
299 	rte = ro->ro_rt;
300 	if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
301 		    rte->rt_ifp == NULL ||
302 		    !RT_LINK_IS_UP(rte->rt_ifp) ||
303 			  dst->sin_family != AF_INET ||
304 			  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
305 		RTFREE(rte);
306 		rte = ro->ro_rt = (struct rtentry *)NULL;
307 		if (ro->ro_lle)
308 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
309 		ro->ro_lle = (struct llentry *)NULL;
310 	}
311 	ia = NULL;
312 	have_ia_ref = 0;
313 	/*
314 	 * If routing to interface only, short circuit routing lookup.
315 	 * The use of an all-ones broadcast address implies this; an
316 	 * interface is specified by the broadcast address of an interface,
317 	 * or the destination address of a ptp interface.
318 	 */
319 	if (flags & IP_SENDONES) {
320 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
321 						      M_GETFIB(m)))) == NULL &&
322 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
323 						    M_GETFIB(m)))) == NULL) {
324 			IPSTAT_INC(ips_noroute);
325 			error = ENETUNREACH;
326 			goto bad;
327 		}
328 		have_ia_ref = 1;
329 		ip->ip_dst.s_addr = INADDR_BROADCAST;
330 		dst->sin_addr = ip->ip_dst;
331 		ifp = ia->ia_ifp;
332 		ip->ip_ttl = 1;
333 		isbroadcast = 1;
334 	} else if (flags & IP_ROUTETOIF) {
335 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
336 						    M_GETFIB(m)))) == NULL &&
337 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
338 						M_GETFIB(m)))) == NULL) {
339 			IPSTAT_INC(ips_noroute);
340 			error = ENETUNREACH;
341 			goto bad;
342 		}
343 		have_ia_ref = 1;
344 		ifp = ia->ia_ifp;
345 		ip->ip_ttl = 1;
346 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
347 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
348 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
349 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
350 		/*
351 		 * Bypass the normal routing lookup for multicast
352 		 * packets if the interface is specified.
353 		 */
354 		ifp = imo->imo_multicast_ifp;
355 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
356 		if (ia)
357 			have_ia_ref = 1;
358 		isbroadcast = 0;	/* fool gcc */
359 	} else {
360 		/*
361 		 * We want to do any cloning requested by the link layer,
362 		 * as this is probably required in all cases for correct
363 		 * operation (as it is for ARP).
364 		 */
365 		if (rte == NULL) {
366 #ifdef RADIX_MPATH
367 			rtalloc_mpath_fib(ro,
368 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
369 			    fibnum);
370 #else
371 			in_rtalloc_ign(ro, 0, fibnum);
372 #endif
373 			rte = ro->ro_rt;
374 		}
375 		if (rte == NULL ||
376 		    (rte->rt_flags & RTF_UP) == 0 ||
377 		    rte->rt_ifp == NULL ||
378 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
379 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
380 			/*
381 			 * There is no route for this packet, but it is
382 			 * possible that a matching SPD entry exists.
383 			 */
384 			no_route_but_check_spd = 1;
385 			mtu = 0; /* Silence GCC warning. */
386 			goto sendit;
387 #endif
388 			IPSTAT_INC(ips_noroute);
389 			error = EHOSTUNREACH;
390 			goto bad;
391 		}
392 		ia = ifatoia(rte->rt_ifa);
393 		ifp = rte->rt_ifp;
394 		counter_u64_add(rte->rt_pksent, 1);
395 		rt_update_ro_flags(ro);
396 		if (rte->rt_flags & RTF_GATEWAY)
397 			gw = (struct sockaddr_in *)rte->rt_gateway;
398 		if (rte->rt_flags & RTF_HOST)
399 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
400 		else if (ifp->if_flags & IFF_BROADCAST)
401 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
402 		else
403 			isbroadcast = 0;
404 	}
405 
406 	/*
407 	 * Calculate MTU.  If we have a route that is up, use that,
408 	 * otherwise use the interface's MTU.
409 	 */
410 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
411 		mtu = rte->rt_mtu;
412 	else
413 		mtu = ifp->if_mtu;
414 	/* Catch a possible divide by zero later. */
415 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
416 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
417 
418 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
419 		m->m_flags |= M_MCAST;
420 		/*
421 		 * IP destination address is multicast.  Make sure "gw"
422 		 * still points to the address in "ro".  (It may have been
423 		 * changed to point to a gateway address, above.)
424 		 */
425 		gw = dst;
426 		/*
427 		 * See if the caller provided any multicast options
428 		 */
429 		if (imo != NULL) {
430 			ip->ip_ttl = imo->imo_multicast_ttl;
431 			if (imo->imo_multicast_vif != -1)
432 				ip->ip_src.s_addr =
433 				    ip_mcast_src ?
434 				    ip_mcast_src(imo->imo_multicast_vif) :
435 				    INADDR_ANY;
436 		} else
437 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
438 		/*
439 		 * Confirm that the outgoing interface supports multicast.
440 		 */
441 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
442 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
443 				IPSTAT_INC(ips_noroute);
444 				error = ENETUNREACH;
445 				goto bad;
446 			}
447 		}
448 		/*
449 		 * If source address not specified yet, use address
450 		 * of outgoing interface.
451 		 */
452 		if (ip->ip_src.s_addr == INADDR_ANY) {
453 			/* Interface may have no addresses. */
454 			if (ia != NULL)
455 				ip->ip_src = IA_SIN(ia)->sin_addr;
456 		}
457 
458 		if ((imo == NULL && in_mcast_loop) ||
459 		    (imo && imo->imo_multicast_loop)) {
460 			/*
461 			 * Loop back multicast datagram if not expressly
462 			 * forbidden to do so, even if we are not a member
463 			 * of the group; ip_input() will filter it later,
464 			 * thus deferring a hash lookup and mutex acquisition
465 			 * at the expense of a cheap copy using m_copym().
466 			 */
467 			ip_mloopback(ifp, m, hlen);
468 		} else {
469 			/*
470 			 * If we are acting as a multicast router, perform
471 			 * multicast forwarding as if the packet had just
472 			 * arrived on the interface to which we are about
473 			 * to send.  The multicast forwarding function
474 			 * recursively calls this function, using the
475 			 * IP_FORWARDING flag to prevent infinite recursion.
476 			 *
477 			 * Multicasts that are looped back by ip_mloopback(),
478 			 * above, will be forwarded by the ip_input() routine,
479 			 * if necessary.
480 			 */
481 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
482 				/*
483 				 * If rsvp daemon is not running, do not
484 				 * set ip_moptions. This ensures that the packet
485 				 * is multicast and not just sent down one link
486 				 * as prescribed by rsvpd.
487 				 */
488 				if (!V_rsvp_on)
489 					imo = NULL;
490 				if (ip_mforward &&
491 				    ip_mforward(ip, ifp, m, imo) != 0) {
492 					m_freem(m);
493 					goto done;
494 				}
495 			}
496 		}
497 
498 		/*
499 		 * Multicasts with a time-to-live of zero may be looped-
500 		 * back, above, but must not be transmitted on a network.
501 		 * Also, multicasts addressed to the loopback interface
502 		 * are not sent -- the above call to ip_mloopback() will
503 		 * loop back a copy. ip_input() will drop the copy if
504 		 * this host does not belong to the destination group on
505 		 * the loopback interface.
506 		 */
507 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
508 			m_freem(m);
509 			goto done;
510 		}
511 
512 		goto sendit;
513 	}
514 
515 	/*
516 	 * If the source address is not specified yet, use the address
517 	 * of the outoing interface.
518 	 */
519 	if (ip->ip_src.s_addr == INADDR_ANY) {
520 		/* Interface may have no addresses. */
521 		if (ia != NULL) {
522 			ip->ip_src = IA_SIN(ia)->sin_addr;
523 		}
524 	}
525 
526 	/*
527 	 * Look for broadcast address and
528 	 * verify user is allowed to send
529 	 * such a packet.
530 	 */
531 	if (isbroadcast) {
532 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
533 			error = EADDRNOTAVAIL;
534 			goto bad;
535 		}
536 		if ((flags & IP_ALLOWBROADCAST) == 0) {
537 			error = EACCES;
538 			goto bad;
539 		}
540 		/* don't allow broadcast messages to be fragmented */
541 		if (ip_len > mtu) {
542 			error = EMSGSIZE;
543 			goto bad;
544 		}
545 		m->m_flags |= M_BCAST;
546 	} else {
547 		m->m_flags &= ~M_BCAST;
548 	}
549 
550 sendit:
551 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
552 	if (IPSEC_ENABLED(ipv4)) {
553 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
554 			if (error == EINPROGRESS)
555 				error = 0;
556 			goto done;
557 		}
558 	}
559 	/*
560 	 * Check if there was a route for this packet; return error if not.
561 	 */
562 	if (no_route_but_check_spd) {
563 		IPSTAT_INC(ips_noroute);
564 		error = EHOSTUNREACH;
565 		goto bad;
566 	}
567 	/* Update variables that are affected by ipsec4_output(). */
568 	ip = mtod(m, struct ip *);
569 	hlen = ip->ip_hl << 2;
570 #endif /* IPSEC */
571 
572 	/* Jump over all PFIL processing if hooks are not active. */
573 	if (PFIL_HOOKED(&V_inet_pfil_hook)) {
574 		switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
575 		case 1: /* Finished */
576 			goto done;
577 
578 		case 0: /* Continue normally */
579 			ip = mtod(m, struct ip *);
580 			break;
581 
582 		case -1: /* Need to try again */
583 			/* Reset everything for a new round */
584 			RO_RTFREE(ro);
585 			if (have_ia_ref)
586 				ifa_free(&ia->ia_ifa);
587 			ro->ro_prepend = NULL;
588 			rte = NULL;
589 			gw = dst;
590 			ip = mtod(m, struct ip *);
591 			goto again;
592 
593 		}
594 	}
595 
596 	/* 127/8 must not appear on wire - RFC1122. */
597 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
598 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
599 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
600 			IPSTAT_INC(ips_badaddr);
601 			error = EADDRNOTAVAIL;
602 			goto bad;
603 		}
604 	}
605 
606 	m->m_pkthdr.csum_flags |= CSUM_IP;
607 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
608 		in_delayed_cksum(m);
609 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
610 	}
611 #ifdef SCTP
612 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
613 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
614 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
615 	}
616 #endif
617 
618 	/*
619 	 * If small enough for interface, or the interface will take
620 	 * care of the fragmentation for us, we can just send directly.
621 	 */
622 	if (ip_len <= mtu ||
623 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
624 		ip->ip_sum = 0;
625 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
626 			ip->ip_sum = in_cksum(m, hlen);
627 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
628 		}
629 
630 		/*
631 		 * Record statistics for this interface address.
632 		 * With CSUM_TSO the byte/packet count will be slightly
633 		 * incorrect because we count the IP+TCP headers only
634 		 * once instead of for every generated packet.
635 		 */
636 		if (!(flags & IP_FORWARDING) && ia) {
637 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
638 				counter_u64_add(ia->ia_ifa.ifa_opackets,
639 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
640 			else
641 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
642 
643 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
644 		}
645 #ifdef MBUF_STRESS_TEST
646 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
647 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
648 #endif
649 		/*
650 		 * Reset layer specific mbuf flags
651 		 * to avoid confusing lower layers.
652 		 */
653 		m_clrprotoflags(m);
654 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
655 #ifdef RATELIMIT
656 		if (inp != NULL) {
657 			if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
658 				in_pcboutput_txrtlmt(inp, ifp, m);
659 			/* stamp send tag on mbuf */
660 			m->m_pkthdr.snd_tag = inp->inp_snd_tag;
661 		} else {
662 			m->m_pkthdr.snd_tag = NULL;
663 		}
664 #endif
665 		error = (*ifp->if_output)(ifp, m,
666 		    (const struct sockaddr *)gw, ro);
667 #ifdef RATELIMIT
668 		/* check for route change */
669 		if (error == EAGAIN)
670 			in_pcboutput_eagain(inp);
671 #endif
672 		goto done;
673 	}
674 
675 	/* Balk when DF bit is set or the interface didn't support TSO. */
676 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
677 		error = EMSGSIZE;
678 		IPSTAT_INC(ips_cantfrag);
679 		goto bad;
680 	}
681 
682 	/*
683 	 * Too large for interface; fragment if possible. If successful,
684 	 * on return, m will point to a list of packets to be sent.
685 	 */
686 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
687 	if (error)
688 		goto bad;
689 	for (; m; m = m0) {
690 		m0 = m->m_nextpkt;
691 		m->m_nextpkt = 0;
692 		if (error == 0) {
693 			/* Record statistics for this interface address. */
694 			if (ia != NULL) {
695 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
696 				counter_u64_add(ia->ia_ifa.ifa_obytes,
697 				    m->m_pkthdr.len);
698 			}
699 			/*
700 			 * Reset layer specific mbuf flags
701 			 * to avoid confusing upper layers.
702 			 */
703 			m_clrprotoflags(m);
704 
705 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
706 			    mtod(m, struct ip *), NULL);
707 #ifdef RATELIMIT
708 			if (inp != NULL) {
709 				if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
710 					in_pcboutput_txrtlmt(inp, ifp, m);
711 				/* stamp send tag on mbuf */
712 				m->m_pkthdr.snd_tag = inp->inp_snd_tag;
713 			} else {
714 				m->m_pkthdr.snd_tag = NULL;
715 			}
716 #endif
717 			error = (*ifp->if_output)(ifp, m,
718 			    (const struct sockaddr *)gw, ro);
719 #ifdef RATELIMIT
720 			/* check for route change */
721 			if (error == EAGAIN)
722 				in_pcboutput_eagain(inp);
723 #endif
724 		} else
725 			m_freem(m);
726 	}
727 
728 	if (error == 0)
729 		IPSTAT_INC(ips_fragmented);
730 
731 done:
732 	if (ro == &iproute)
733 		RO_RTFREE(ro);
734 	else if (rte == NULL)
735 		/*
736 		 * If the caller supplied a route but somehow the reference
737 		 * to it has been released need to prevent the caller
738 		 * calling RTFREE on it again.
739 		 */
740 		ro->ro_rt = NULL;
741 	if (have_ia_ref)
742 		ifa_free(&ia->ia_ifa);
743 	return (error);
744 bad:
745 	m_freem(m);
746 	goto done;
747 }
748 
749 /*
750  * Create a chain of fragments which fit the given mtu. m_frag points to the
751  * mbuf to be fragmented; on return it points to the chain with the fragments.
752  * Return 0 if no error. If error, m_frag may contain a partially built
753  * chain of fragments that should be freed by the caller.
754  *
755  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
756  */
757 int
758 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
759     u_long if_hwassist_flags)
760 {
761 	int error = 0;
762 	int hlen = ip->ip_hl << 2;
763 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
764 	int off;
765 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
766 	int firstlen;
767 	struct mbuf **mnext;
768 	int nfrags;
769 	uint16_t ip_len, ip_off;
770 
771 	ip_len = ntohs(ip->ip_len);
772 	ip_off = ntohs(ip->ip_off);
773 
774 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
775 		IPSTAT_INC(ips_cantfrag);
776 		return EMSGSIZE;
777 	}
778 
779 	/*
780 	 * Must be able to put at least 8 bytes per fragment.
781 	 */
782 	if (len < 8)
783 		return EMSGSIZE;
784 
785 	/*
786 	 * If the interface will not calculate checksums on
787 	 * fragmented packets, then do it here.
788 	 */
789 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
790 		in_delayed_cksum(m0);
791 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
792 	}
793 #ifdef SCTP
794 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
795 		sctp_delayed_cksum(m0, hlen);
796 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
797 	}
798 #endif
799 	if (len > PAGE_SIZE) {
800 		/*
801 		 * Fragment large datagrams such that each segment
802 		 * contains a multiple of PAGE_SIZE amount of data,
803 		 * plus headers. This enables a receiver to perform
804 		 * page-flipping zero-copy optimizations.
805 		 *
806 		 * XXX When does this help given that sender and receiver
807 		 * could have different page sizes, and also mtu could
808 		 * be less than the receiver's page size ?
809 		 */
810 		int newlen;
811 
812 		off = MIN(mtu, m0->m_pkthdr.len);
813 
814 		/*
815 		 * firstlen (off - hlen) must be aligned on an
816 		 * 8-byte boundary
817 		 */
818 		if (off < hlen)
819 			goto smart_frag_failure;
820 		off = ((off - hlen) & ~7) + hlen;
821 		newlen = (~PAGE_MASK) & mtu;
822 		if ((newlen + sizeof (struct ip)) > mtu) {
823 			/* we failed, go back the default */
824 smart_frag_failure:
825 			newlen = len;
826 			off = hlen + len;
827 		}
828 		len = newlen;
829 
830 	} else {
831 		off = hlen + len;
832 	}
833 
834 	firstlen = off - hlen;
835 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
836 
837 	/*
838 	 * Loop through length of segment after first fragment,
839 	 * make new header and copy data of each part and link onto chain.
840 	 * Here, m0 is the original packet, m is the fragment being created.
841 	 * The fragments are linked off the m_nextpkt of the original
842 	 * packet, which after processing serves as the first fragment.
843 	 */
844 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
845 		struct ip *mhip;	/* ip header on the fragment */
846 		struct mbuf *m;
847 		int mhlen = sizeof (struct ip);
848 
849 		m = m_gethdr(M_NOWAIT, MT_DATA);
850 		if (m == NULL) {
851 			error = ENOBUFS;
852 			IPSTAT_INC(ips_odropped);
853 			goto done;
854 		}
855 		/*
856 		 * Make sure the complete packet header gets copied
857 		 * from the originating mbuf to the newly created
858 		 * mbuf. This also ensures that existing firewall
859 		 * classification(s), VLAN tags and so on get copied
860 		 * to the resulting fragmented packet(s):
861 		 */
862 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
863 			m_free(m);
864 			error = ENOBUFS;
865 			IPSTAT_INC(ips_odropped);
866 			goto done;
867 		}
868 		/*
869 		 * In the first mbuf, leave room for the link header, then
870 		 * copy the original IP header including options. The payload
871 		 * goes into an additional mbuf chain returned by m_copym().
872 		 */
873 		m->m_data += max_linkhdr;
874 		mhip = mtod(m, struct ip *);
875 		*mhip = *ip;
876 		if (hlen > sizeof (struct ip)) {
877 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
878 			mhip->ip_v = IPVERSION;
879 			mhip->ip_hl = mhlen >> 2;
880 		}
881 		m->m_len = mhlen;
882 		/* XXX do we need to add ip_off below ? */
883 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
884 		if (off + len >= ip_len)
885 			len = ip_len - off;
886 		else
887 			mhip->ip_off |= IP_MF;
888 		mhip->ip_len = htons((u_short)(len + mhlen));
889 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
890 		if (m->m_next == NULL) {	/* copy failed */
891 			m_free(m);
892 			error = ENOBUFS;	/* ??? */
893 			IPSTAT_INC(ips_odropped);
894 			goto done;
895 		}
896 		m->m_pkthdr.len = mhlen + len;
897 #ifdef MAC
898 		mac_netinet_fragment(m0, m);
899 #endif
900 		mhip->ip_off = htons(mhip->ip_off);
901 		mhip->ip_sum = 0;
902 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
903 			mhip->ip_sum = in_cksum(m, mhlen);
904 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
905 		}
906 		*mnext = m;
907 		mnext = &m->m_nextpkt;
908 	}
909 	IPSTAT_ADD(ips_ofragments, nfrags);
910 
911 	/*
912 	 * Update first fragment by trimming what's been copied out
913 	 * and updating header.
914 	 */
915 	m_adj(m0, hlen + firstlen - ip_len);
916 	m0->m_pkthdr.len = hlen + firstlen;
917 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
918 	ip->ip_off = htons(ip_off | IP_MF);
919 	ip->ip_sum = 0;
920 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
921 		ip->ip_sum = in_cksum(m0, hlen);
922 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
923 	}
924 
925 done:
926 	*m_frag = m0;
927 	return error;
928 }
929 
930 void
931 in_delayed_cksum(struct mbuf *m)
932 {
933 	struct ip *ip;
934 	uint16_t csum, offset, ip_len;
935 
936 	ip = mtod(m, struct ip *);
937 	offset = ip->ip_hl << 2 ;
938 	ip_len = ntohs(ip->ip_len);
939 	csum = in_cksum_skip(m, ip_len, offset);
940 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
941 		csum = 0xffff;
942 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
943 
944 	/* find the mbuf in the chain where the checksum starts*/
945 	while ((m != NULL) && (offset >= m->m_len)) {
946 		offset -= m->m_len;
947 		m = m->m_next;
948 	}
949 	KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
950 	KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
951 	*(u_short *)(m->m_data + offset) = csum;
952 }
953 
954 /*
955  * IP socket option processing.
956  */
957 int
958 ip_ctloutput(struct socket *so, struct sockopt *sopt)
959 {
960 	struct	inpcb *inp = sotoinpcb(so);
961 	int	error, optval;
962 #ifdef	RSS
963 	uint32_t rss_bucket;
964 	int retval;
965 #endif
966 
967 	error = optval = 0;
968 	if (sopt->sopt_level != IPPROTO_IP) {
969 		error = EINVAL;
970 
971 		if (sopt->sopt_level == SOL_SOCKET &&
972 		    sopt->sopt_dir == SOPT_SET) {
973 			switch (sopt->sopt_name) {
974 			case SO_REUSEADDR:
975 				INP_WLOCK(inp);
976 				if ((so->so_options & SO_REUSEADDR) != 0)
977 					inp->inp_flags2 |= INP_REUSEADDR;
978 				else
979 					inp->inp_flags2 &= ~INP_REUSEADDR;
980 				INP_WUNLOCK(inp);
981 				error = 0;
982 				break;
983 			case SO_REUSEPORT:
984 				INP_WLOCK(inp);
985 				if ((so->so_options & SO_REUSEPORT) != 0)
986 					inp->inp_flags2 |= INP_REUSEPORT;
987 				else
988 					inp->inp_flags2 &= ~INP_REUSEPORT;
989 				INP_WUNLOCK(inp);
990 				error = 0;
991 				break;
992 			case SO_SETFIB:
993 				INP_WLOCK(inp);
994 				inp->inp_inc.inc_fibnum = so->so_fibnum;
995 				INP_WUNLOCK(inp);
996 				error = 0;
997 				break;
998 			case SO_MAX_PACING_RATE:
999 #ifdef RATELIMIT
1000 				INP_WLOCK(inp);
1001 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1002 				INP_WUNLOCK(inp);
1003 				error = 0;
1004 #else
1005 				error = EOPNOTSUPP;
1006 #endif
1007 				break;
1008 			default:
1009 				break;
1010 			}
1011 		}
1012 		return (error);
1013 	}
1014 
1015 	switch (sopt->sopt_dir) {
1016 	case SOPT_SET:
1017 		switch (sopt->sopt_name) {
1018 		case IP_OPTIONS:
1019 #ifdef notyet
1020 		case IP_RETOPTS:
1021 #endif
1022 		{
1023 			struct mbuf *m;
1024 			if (sopt->sopt_valsize > MLEN) {
1025 				error = EMSGSIZE;
1026 				break;
1027 			}
1028 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1029 			if (m == NULL) {
1030 				error = ENOBUFS;
1031 				break;
1032 			}
1033 			m->m_len = sopt->sopt_valsize;
1034 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1035 					    m->m_len);
1036 			if (error) {
1037 				m_free(m);
1038 				break;
1039 			}
1040 			INP_WLOCK(inp);
1041 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1042 			INP_WUNLOCK(inp);
1043 			return (error);
1044 		}
1045 
1046 		case IP_BINDANY:
1047 			if (sopt->sopt_td != NULL) {
1048 				error = priv_check(sopt->sopt_td,
1049 				    PRIV_NETINET_BINDANY);
1050 				if (error)
1051 					break;
1052 			}
1053 			/* FALLTHROUGH */
1054 		case IP_BINDMULTI:
1055 #ifdef	RSS
1056 		case IP_RSS_LISTEN_BUCKET:
1057 #endif
1058 		case IP_TOS:
1059 		case IP_TTL:
1060 		case IP_MINTTL:
1061 		case IP_RECVOPTS:
1062 		case IP_RECVRETOPTS:
1063 		case IP_ORIGDSTADDR:
1064 		case IP_RECVDSTADDR:
1065 		case IP_RECVTTL:
1066 		case IP_RECVIF:
1067 		case IP_ONESBCAST:
1068 		case IP_DONTFRAG:
1069 		case IP_RECVTOS:
1070 		case IP_RECVFLOWID:
1071 #ifdef	RSS
1072 		case IP_RECVRSSBUCKETID:
1073 #endif
1074 			error = sooptcopyin(sopt, &optval, sizeof optval,
1075 					    sizeof optval);
1076 			if (error)
1077 				break;
1078 
1079 			switch (sopt->sopt_name) {
1080 			case IP_TOS:
1081 				inp->inp_ip_tos = optval;
1082 				break;
1083 
1084 			case IP_TTL:
1085 				inp->inp_ip_ttl = optval;
1086 				break;
1087 
1088 			case IP_MINTTL:
1089 				if (optval >= 0 && optval <= MAXTTL)
1090 					inp->inp_ip_minttl = optval;
1091 				else
1092 					error = EINVAL;
1093 				break;
1094 
1095 #define	OPTSET(bit) do {						\
1096 	INP_WLOCK(inp);							\
1097 	if (optval)							\
1098 		inp->inp_flags |= bit;					\
1099 	else								\
1100 		inp->inp_flags &= ~bit;					\
1101 	INP_WUNLOCK(inp);						\
1102 } while (0)
1103 
1104 #define	OPTSET2(bit, val) do {						\
1105 	INP_WLOCK(inp);							\
1106 	if (val)							\
1107 		inp->inp_flags2 |= bit;					\
1108 	else								\
1109 		inp->inp_flags2 &= ~bit;				\
1110 	INP_WUNLOCK(inp);						\
1111 } while (0)
1112 
1113 			case IP_RECVOPTS:
1114 				OPTSET(INP_RECVOPTS);
1115 				break;
1116 
1117 			case IP_RECVRETOPTS:
1118 				OPTSET(INP_RECVRETOPTS);
1119 				break;
1120 
1121 			case IP_RECVDSTADDR:
1122 				OPTSET(INP_RECVDSTADDR);
1123 				break;
1124 
1125 			case IP_ORIGDSTADDR:
1126 				OPTSET2(INP_ORIGDSTADDR, optval);
1127 				break;
1128 
1129 			case IP_RECVTTL:
1130 				OPTSET(INP_RECVTTL);
1131 				break;
1132 
1133 			case IP_RECVIF:
1134 				OPTSET(INP_RECVIF);
1135 				break;
1136 
1137 			case IP_ONESBCAST:
1138 				OPTSET(INP_ONESBCAST);
1139 				break;
1140 			case IP_DONTFRAG:
1141 				OPTSET(INP_DONTFRAG);
1142 				break;
1143 			case IP_BINDANY:
1144 				OPTSET(INP_BINDANY);
1145 				break;
1146 			case IP_RECVTOS:
1147 				OPTSET(INP_RECVTOS);
1148 				break;
1149 			case IP_BINDMULTI:
1150 				OPTSET2(INP_BINDMULTI, optval);
1151 				break;
1152 			case IP_RECVFLOWID:
1153 				OPTSET2(INP_RECVFLOWID, optval);
1154 				break;
1155 #ifdef	RSS
1156 			case IP_RSS_LISTEN_BUCKET:
1157 				if ((optval >= 0) &&
1158 				    (optval < rss_getnumbuckets())) {
1159 					inp->inp_rss_listen_bucket = optval;
1160 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1161 				} else {
1162 					error = EINVAL;
1163 				}
1164 				break;
1165 			case IP_RECVRSSBUCKETID:
1166 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1167 				break;
1168 #endif
1169 			}
1170 			break;
1171 #undef OPTSET
1172 #undef OPTSET2
1173 
1174 		/*
1175 		 * Multicast socket options are processed by the in_mcast
1176 		 * module.
1177 		 */
1178 		case IP_MULTICAST_IF:
1179 		case IP_MULTICAST_VIF:
1180 		case IP_MULTICAST_TTL:
1181 		case IP_MULTICAST_LOOP:
1182 		case IP_ADD_MEMBERSHIP:
1183 		case IP_DROP_MEMBERSHIP:
1184 		case IP_ADD_SOURCE_MEMBERSHIP:
1185 		case IP_DROP_SOURCE_MEMBERSHIP:
1186 		case IP_BLOCK_SOURCE:
1187 		case IP_UNBLOCK_SOURCE:
1188 		case IP_MSFILTER:
1189 		case MCAST_JOIN_GROUP:
1190 		case MCAST_LEAVE_GROUP:
1191 		case MCAST_JOIN_SOURCE_GROUP:
1192 		case MCAST_LEAVE_SOURCE_GROUP:
1193 		case MCAST_BLOCK_SOURCE:
1194 		case MCAST_UNBLOCK_SOURCE:
1195 			error = inp_setmoptions(inp, sopt);
1196 			break;
1197 
1198 		case IP_PORTRANGE:
1199 			error = sooptcopyin(sopt, &optval, sizeof optval,
1200 					    sizeof optval);
1201 			if (error)
1202 				break;
1203 
1204 			INP_WLOCK(inp);
1205 			switch (optval) {
1206 			case IP_PORTRANGE_DEFAULT:
1207 				inp->inp_flags &= ~(INP_LOWPORT);
1208 				inp->inp_flags &= ~(INP_HIGHPORT);
1209 				break;
1210 
1211 			case IP_PORTRANGE_HIGH:
1212 				inp->inp_flags &= ~(INP_LOWPORT);
1213 				inp->inp_flags |= INP_HIGHPORT;
1214 				break;
1215 
1216 			case IP_PORTRANGE_LOW:
1217 				inp->inp_flags &= ~(INP_HIGHPORT);
1218 				inp->inp_flags |= INP_LOWPORT;
1219 				break;
1220 
1221 			default:
1222 				error = EINVAL;
1223 				break;
1224 			}
1225 			INP_WUNLOCK(inp);
1226 			break;
1227 
1228 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1229 		case IP_IPSEC_POLICY:
1230 			if (IPSEC_ENABLED(ipv4)) {
1231 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1232 				break;
1233 			}
1234 			/* FALLTHROUGH */
1235 #endif /* IPSEC */
1236 
1237 		default:
1238 			error = ENOPROTOOPT;
1239 			break;
1240 		}
1241 		break;
1242 
1243 	case SOPT_GET:
1244 		switch (sopt->sopt_name) {
1245 		case IP_OPTIONS:
1246 		case IP_RETOPTS:
1247 			if (inp->inp_options)
1248 				error = sooptcopyout(sopt,
1249 						     mtod(inp->inp_options,
1250 							  char *),
1251 						     inp->inp_options->m_len);
1252 			else
1253 				sopt->sopt_valsize = 0;
1254 			break;
1255 
1256 		case IP_TOS:
1257 		case IP_TTL:
1258 		case IP_MINTTL:
1259 		case IP_RECVOPTS:
1260 		case IP_RECVRETOPTS:
1261 		case IP_ORIGDSTADDR:
1262 		case IP_RECVDSTADDR:
1263 		case IP_RECVTTL:
1264 		case IP_RECVIF:
1265 		case IP_PORTRANGE:
1266 		case IP_ONESBCAST:
1267 		case IP_DONTFRAG:
1268 		case IP_BINDANY:
1269 		case IP_RECVTOS:
1270 		case IP_BINDMULTI:
1271 		case IP_FLOWID:
1272 		case IP_FLOWTYPE:
1273 		case IP_RECVFLOWID:
1274 #ifdef	RSS
1275 		case IP_RSSBUCKETID:
1276 		case IP_RECVRSSBUCKETID:
1277 #endif
1278 			switch (sopt->sopt_name) {
1279 
1280 			case IP_TOS:
1281 				optval = inp->inp_ip_tos;
1282 				break;
1283 
1284 			case IP_TTL:
1285 				optval = inp->inp_ip_ttl;
1286 				break;
1287 
1288 			case IP_MINTTL:
1289 				optval = inp->inp_ip_minttl;
1290 				break;
1291 
1292 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1293 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1294 
1295 			case IP_RECVOPTS:
1296 				optval = OPTBIT(INP_RECVOPTS);
1297 				break;
1298 
1299 			case IP_RECVRETOPTS:
1300 				optval = OPTBIT(INP_RECVRETOPTS);
1301 				break;
1302 
1303 			case IP_RECVDSTADDR:
1304 				optval = OPTBIT(INP_RECVDSTADDR);
1305 				break;
1306 
1307 			case IP_ORIGDSTADDR:
1308 				optval = OPTBIT2(INP_ORIGDSTADDR);
1309 				break;
1310 
1311 			case IP_RECVTTL:
1312 				optval = OPTBIT(INP_RECVTTL);
1313 				break;
1314 
1315 			case IP_RECVIF:
1316 				optval = OPTBIT(INP_RECVIF);
1317 				break;
1318 
1319 			case IP_PORTRANGE:
1320 				if (inp->inp_flags & INP_HIGHPORT)
1321 					optval = IP_PORTRANGE_HIGH;
1322 				else if (inp->inp_flags & INP_LOWPORT)
1323 					optval = IP_PORTRANGE_LOW;
1324 				else
1325 					optval = 0;
1326 				break;
1327 
1328 			case IP_ONESBCAST:
1329 				optval = OPTBIT(INP_ONESBCAST);
1330 				break;
1331 			case IP_DONTFRAG:
1332 				optval = OPTBIT(INP_DONTFRAG);
1333 				break;
1334 			case IP_BINDANY:
1335 				optval = OPTBIT(INP_BINDANY);
1336 				break;
1337 			case IP_RECVTOS:
1338 				optval = OPTBIT(INP_RECVTOS);
1339 				break;
1340 			case IP_FLOWID:
1341 				optval = inp->inp_flowid;
1342 				break;
1343 			case IP_FLOWTYPE:
1344 				optval = inp->inp_flowtype;
1345 				break;
1346 			case IP_RECVFLOWID:
1347 				optval = OPTBIT2(INP_RECVFLOWID);
1348 				break;
1349 #ifdef	RSS
1350 			case IP_RSSBUCKETID:
1351 				retval = rss_hash2bucket(inp->inp_flowid,
1352 				    inp->inp_flowtype,
1353 				    &rss_bucket);
1354 				if (retval == 0)
1355 					optval = rss_bucket;
1356 				else
1357 					error = EINVAL;
1358 				break;
1359 			case IP_RECVRSSBUCKETID:
1360 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1361 				break;
1362 #endif
1363 			case IP_BINDMULTI:
1364 				optval = OPTBIT2(INP_BINDMULTI);
1365 				break;
1366 			}
1367 			error = sooptcopyout(sopt, &optval, sizeof optval);
1368 			break;
1369 
1370 		/*
1371 		 * Multicast socket options are processed by the in_mcast
1372 		 * module.
1373 		 */
1374 		case IP_MULTICAST_IF:
1375 		case IP_MULTICAST_VIF:
1376 		case IP_MULTICAST_TTL:
1377 		case IP_MULTICAST_LOOP:
1378 		case IP_MSFILTER:
1379 			error = inp_getmoptions(inp, sopt);
1380 			break;
1381 
1382 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1383 		case IP_IPSEC_POLICY:
1384 			if (IPSEC_ENABLED(ipv4)) {
1385 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1386 				break;
1387 			}
1388 			/* FALLTHROUGH */
1389 #endif /* IPSEC */
1390 
1391 		default:
1392 			error = ENOPROTOOPT;
1393 			break;
1394 		}
1395 		break;
1396 	}
1397 	return (error);
1398 }
1399 
1400 /*
1401  * Routine called from ip_output() to loop back a copy of an IP multicast
1402  * packet to the input queue of a specified interface.  Note that this
1403  * calls the output routine of the loopback "driver", but with an interface
1404  * pointer that might NOT be a loopback interface -- evil, but easier than
1405  * replicating that code here.
1406  */
1407 static void
1408 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1409 {
1410 	struct ip *ip;
1411 	struct mbuf *copym;
1412 
1413 	/*
1414 	 * Make a deep copy of the packet because we're going to
1415 	 * modify the pack in order to generate checksums.
1416 	 */
1417 	copym = m_dup(m, M_NOWAIT);
1418 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1419 		copym = m_pullup(copym, hlen);
1420 	if (copym != NULL) {
1421 		/* If needed, compute the checksum and mark it as valid. */
1422 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1423 			in_delayed_cksum(copym);
1424 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1425 			copym->m_pkthdr.csum_flags |=
1426 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1427 			copym->m_pkthdr.csum_data = 0xffff;
1428 		}
1429 		/*
1430 		 * We don't bother to fragment if the IP length is greater
1431 		 * than the interface's MTU.  Can this possibly matter?
1432 		 */
1433 		ip = mtod(copym, struct ip *);
1434 		ip->ip_sum = 0;
1435 		ip->ip_sum = in_cksum(copym, hlen);
1436 		if_simloop(ifp, copym, AF_INET, 0);
1437 	}
1438 }
1439