xref: /freebsd/sys/netinet/ip_output.c (revision 193d9e768ba63fcfb187cfd17f461f7d41345048)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1990, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 #include "opt_ratelimit.h"
37 #include "opt_ipsec.h"
38 #include "opt_mbuf_stress_test.h"
39 #include "opt_mpath.h"
40 #include "opt_route.h"
41 #include "opt_sctp.h"
42 #include "opt_rss.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/lock.h>
48 #include <sys/malloc.h>
49 #include <sys/mbuf.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/protosw.h>
53 #include <sys/rmlock.h>
54 #include <sys/sdt.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sysctl.h>
58 #include <sys/ucred.h>
59 
60 #include <net/if.h>
61 #include <net/if_var.h>
62 #include <net/if_llatbl.h>
63 #include <net/netisr.h>
64 #include <net/pfil.h>
65 #include <net/route.h>
66 #include <net/flowtable.h>
67 #ifdef RADIX_MPATH
68 #include <net/radix_mpath.h>
69 #endif
70 #include <net/rss_config.h>
71 #include <net/vnet.h>
72 
73 #include <netinet/in.h>
74 #include <netinet/in_kdtrace.h>
75 #include <netinet/in_systm.h>
76 #include <netinet/ip.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/in_rss.h>
79 #include <netinet/in_var.h>
80 #include <netinet/ip_var.h>
81 #include <netinet/ip_options.h>
82 #ifdef SCTP
83 #include <netinet/sctp.h>
84 #include <netinet/sctp_crc32.h>
85 #endif
86 
87 #include <netipsec/ipsec_support.h>
88 
89 #include <machine/in_cksum.h>
90 
91 #include <security/mac/mac_framework.h>
92 
93 #ifdef MBUF_STRESS_TEST
94 static int mbuf_frag_size = 0;
95 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
96 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
97 #endif
98 
99 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
100 
101 
102 extern int in_mcast_loop;
103 extern	struct protosw inetsw[];
104 
105 static inline int
106 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
107     struct sockaddr_in *dst, int *fibnum, int *error)
108 {
109 	struct m_tag *fwd_tag = NULL;
110 	struct mbuf *m;
111 	struct in_addr odst;
112 	struct ip *ip;
113 
114 	m = *mp;
115 	ip = mtod(m, struct ip *);
116 
117 	/* Run through list of hooks for output packets. */
118 	odst.s_addr = ip->ip_dst.s_addr;
119 	*error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, inp);
120 	m = *mp;
121 	if ((*error) != 0 || m == NULL)
122 		return 1; /* Finished */
123 
124 	ip = mtod(m, struct ip *);
125 
126 	/* See if destination IP address was changed by packet filter. */
127 	if (odst.s_addr != ip->ip_dst.s_addr) {
128 		m->m_flags |= M_SKIP_FIREWALL;
129 		/* If destination is now ourself drop to ip_input(). */
130 		if (in_localip(ip->ip_dst)) {
131 			m->m_flags |= M_FASTFWD_OURS;
132 			if (m->m_pkthdr.rcvif == NULL)
133 				m->m_pkthdr.rcvif = V_loif;
134 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
135 				m->m_pkthdr.csum_flags |=
136 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
137 				m->m_pkthdr.csum_data = 0xffff;
138 			}
139 			m->m_pkthdr.csum_flags |=
140 				CSUM_IP_CHECKED | CSUM_IP_VALID;
141 #ifdef SCTP
142 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
143 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
144 #endif
145 			*error = netisr_queue(NETISR_IP, m);
146 			return 1; /* Finished */
147 		}
148 
149 		bzero(dst, sizeof(*dst));
150 		dst->sin_family = AF_INET;
151 		dst->sin_len = sizeof(*dst);
152 		dst->sin_addr = ip->ip_dst;
153 
154 		return -1; /* Reloop */
155 	}
156 	/* See if fib was changed by packet filter. */
157 	if ((*fibnum) != M_GETFIB(m)) {
158 		m->m_flags |= M_SKIP_FIREWALL;
159 		*fibnum = M_GETFIB(m);
160 		return -1; /* Reloop for FIB change */
161 	}
162 
163 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
164 	if (m->m_flags & M_FASTFWD_OURS) {
165 		if (m->m_pkthdr.rcvif == NULL)
166 			m->m_pkthdr.rcvif = V_loif;
167 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
168 			m->m_pkthdr.csum_flags |=
169 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
170 			m->m_pkthdr.csum_data = 0xffff;
171 		}
172 #ifdef SCTP
173 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
174 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
175 #endif
176 		m->m_pkthdr.csum_flags |=
177 			CSUM_IP_CHECKED | CSUM_IP_VALID;
178 
179 		*error = netisr_queue(NETISR_IP, m);
180 		return 1; /* Finished */
181 	}
182 	/* Or forward to some other address? */
183 	if ((m->m_flags & M_IP_NEXTHOP) &&
184 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
185 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
186 		m->m_flags |= M_SKIP_FIREWALL;
187 		m->m_flags &= ~M_IP_NEXTHOP;
188 		m_tag_delete(m, fwd_tag);
189 
190 		return -1; /* Reloop for CHANGE of dst */
191 	}
192 
193 	return 0;
194 }
195 
196 /*
197  * IP output.  The packet in mbuf chain m contains a skeletal IP
198  * header (with len, off, ttl, proto, tos, src, dst).
199  * The mbuf chain containing the packet will be freed.
200  * The mbuf opt, if present, will not be freed.
201  * If route ro is present and has ro_rt initialized, route lookup would be
202  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
203  * then result of route lookup is stored in ro->ro_rt.
204  *
205  * In the IP forwarding case, the packet will arrive with options already
206  * inserted, so must have a NULL opt pointer.
207  */
208 int
209 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
210     struct ip_moptions *imo, struct inpcb *inp)
211 {
212 	struct rm_priotracker in_ifa_tracker;
213 	struct ip *ip;
214 	struct ifnet *ifp = NULL;	/* keep compiler happy */
215 	struct mbuf *m0;
216 	int hlen = sizeof (struct ip);
217 	int mtu;
218 	int error = 0;
219 	struct sockaddr_in *dst;
220 	const struct sockaddr_in *gw;
221 	struct in_ifaddr *ia;
222 	int isbroadcast;
223 	uint16_t ip_len, ip_off;
224 	struct route iproute;
225 	struct rtentry *rte;	/* cache for ro->ro_rt */
226 	uint32_t fibnum;
227 	int have_ia_ref;
228 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
229 	int no_route_but_check_spd = 0;
230 #endif
231 	M_ASSERTPKTHDR(m);
232 
233 	if (inp != NULL) {
234 		INP_LOCK_ASSERT(inp);
235 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
236 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
237 			m->m_pkthdr.flowid = inp->inp_flowid;
238 			M_HASHTYPE_SET(m, inp->inp_flowtype);
239 		}
240 	}
241 
242 	if (ro == NULL) {
243 		ro = &iproute;
244 		bzero(ro, sizeof (*ro));
245 	} else
246 		ro->ro_flags |= RT_LLE_CACHE;
247 
248 #ifdef FLOWTABLE
249 	if (ro->ro_rt == NULL)
250 		(void )flowtable_lookup(AF_INET, m, ro);
251 #endif
252 
253 	if (opt) {
254 		int len = 0;
255 		m = ip_insertoptions(m, opt, &len);
256 		if (len != 0)
257 			hlen = len; /* ip->ip_hl is updated above */
258 	}
259 	ip = mtod(m, struct ip *);
260 	ip_len = ntohs(ip->ip_len);
261 	ip_off = ntohs(ip->ip_off);
262 
263 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
264 		ip->ip_v = IPVERSION;
265 		ip->ip_hl = hlen >> 2;
266 		ip_fillid(ip);
267 		IPSTAT_INC(ips_localout);
268 	} else {
269 		/* Header already set, fetch hlen from there */
270 		hlen = ip->ip_hl << 2;
271 	}
272 
273 	/*
274 	 * dst/gw handling:
275 	 *
276 	 * dst can be rewritten but always points to &ro->ro_dst.
277 	 * gw is readonly but can point either to dst OR rt_gateway,
278 	 * therefore we need restore gw if we're redoing lookup.
279 	 */
280 	gw = dst = (struct sockaddr_in *)&ro->ro_dst;
281 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
282 	rte = ro->ro_rt;
283 	if (rte == NULL) {
284 		bzero(dst, sizeof(*dst));
285 		dst->sin_family = AF_INET;
286 		dst->sin_len = sizeof(*dst);
287 		dst->sin_addr = ip->ip_dst;
288 	}
289 again:
290 	/*
291 	 * Validate route against routing table additions;
292 	 * a better/more specific route might have been added.
293 	 */
294 	if (inp)
295 		RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
296 	/*
297 	 * If there is a cached route,
298 	 * check that it is to the same destination
299 	 * and is still up.  If not, free it and try again.
300 	 * The address family should also be checked in case of sharing the
301 	 * cache with IPv6.
302 	 * Also check whether routing cache needs invalidation.
303 	 */
304 	rte = ro->ro_rt;
305 	if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
306 		    rte->rt_ifp == NULL ||
307 		    !RT_LINK_IS_UP(rte->rt_ifp) ||
308 			  dst->sin_family != AF_INET ||
309 			  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
310 		RTFREE(rte);
311 		rte = ro->ro_rt = (struct rtentry *)NULL;
312 		if (ro->ro_lle)
313 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
314 		ro->ro_lle = (struct llentry *)NULL;
315 	}
316 	ia = NULL;
317 	have_ia_ref = 0;
318 	/*
319 	 * If routing to interface only, short circuit routing lookup.
320 	 * The use of an all-ones broadcast address implies this; an
321 	 * interface is specified by the broadcast address of an interface,
322 	 * or the destination address of a ptp interface.
323 	 */
324 	if (flags & IP_SENDONES) {
325 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
326 						      M_GETFIB(m)))) == NULL &&
327 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
328 						    M_GETFIB(m)))) == NULL) {
329 			IPSTAT_INC(ips_noroute);
330 			error = ENETUNREACH;
331 			goto bad;
332 		}
333 		have_ia_ref = 1;
334 		ip->ip_dst.s_addr = INADDR_BROADCAST;
335 		dst->sin_addr = ip->ip_dst;
336 		ifp = ia->ia_ifp;
337 		ip->ip_ttl = 1;
338 		isbroadcast = 1;
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 		have_ia_ref = 1;
349 		ifp = ia->ia_ifp;
350 		ip->ip_ttl = 1;
351 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
352 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
353 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
354 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
355 		/*
356 		 * Bypass the normal routing lookup for multicast
357 		 * packets if the interface is specified.
358 		 */
359 		ifp = imo->imo_multicast_ifp;
360 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
361 		if (ia)
362 			have_ia_ref = 1;
363 		isbroadcast = 0;	/* fool gcc */
364 	} else {
365 		/*
366 		 * We want to do any cloning requested by the link layer,
367 		 * as this is probably required in all cases for correct
368 		 * operation (as it is for ARP).
369 		 */
370 		if (rte == NULL) {
371 #ifdef RADIX_MPATH
372 			rtalloc_mpath_fib(ro,
373 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
374 			    fibnum);
375 #else
376 			in_rtalloc_ign(ro, 0, fibnum);
377 #endif
378 			rte = ro->ro_rt;
379 		}
380 		if (rte == NULL ||
381 		    (rte->rt_flags & RTF_UP) == 0 ||
382 		    rte->rt_ifp == NULL ||
383 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
384 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
385 			/*
386 			 * There is no route for this packet, but it is
387 			 * possible that a matching SPD entry exists.
388 			 */
389 			no_route_but_check_spd = 1;
390 			mtu = 0; /* Silence GCC warning. */
391 			goto sendit;
392 #endif
393 			IPSTAT_INC(ips_noroute);
394 			error = EHOSTUNREACH;
395 			goto bad;
396 		}
397 		ia = ifatoia(rte->rt_ifa);
398 		ifp = rte->rt_ifp;
399 		counter_u64_add(rte->rt_pksent, 1);
400 		rt_update_ro_flags(ro);
401 		if (rte->rt_flags & RTF_GATEWAY)
402 			gw = (struct sockaddr_in *)rte->rt_gateway;
403 		if (rte->rt_flags & RTF_HOST)
404 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
405 		else if (ifp->if_flags & IFF_BROADCAST)
406 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
407 		else
408 			isbroadcast = 0;
409 	}
410 
411 	/*
412 	 * Calculate MTU.  If we have a route that is up, use that,
413 	 * otherwise use the interface's MTU.
414 	 */
415 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
416 		mtu = rte->rt_mtu;
417 	else
418 		mtu = ifp->if_mtu;
419 	/* Catch a possible divide by zero later. */
420 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
421 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
422 
423 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
424 		m->m_flags |= M_MCAST;
425 		/*
426 		 * IP destination address is multicast.  Make sure "gw"
427 		 * still points to the address in "ro".  (It may have been
428 		 * changed to point to a gateway address, above.)
429 		 */
430 		gw = dst;
431 		/*
432 		 * See if the caller provided any multicast options
433 		 */
434 		if (imo != NULL) {
435 			ip->ip_ttl = imo->imo_multicast_ttl;
436 			if (imo->imo_multicast_vif != -1)
437 				ip->ip_src.s_addr =
438 				    ip_mcast_src ?
439 				    ip_mcast_src(imo->imo_multicast_vif) :
440 				    INADDR_ANY;
441 		} else
442 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
443 		/*
444 		 * Confirm that the outgoing interface supports multicast.
445 		 */
446 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
447 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
448 				IPSTAT_INC(ips_noroute);
449 				error = ENETUNREACH;
450 				goto bad;
451 			}
452 		}
453 		/*
454 		 * If source address not specified yet, use address
455 		 * of outgoing interface.
456 		 */
457 		if (ip->ip_src.s_addr == INADDR_ANY) {
458 			/* Interface may have no addresses. */
459 			if (ia != NULL)
460 				ip->ip_src = IA_SIN(ia)->sin_addr;
461 		}
462 
463 		if ((imo == NULL && in_mcast_loop) ||
464 		    (imo && imo->imo_multicast_loop)) {
465 			/*
466 			 * Loop back multicast datagram if not expressly
467 			 * forbidden to do so, even if we are not a member
468 			 * of the group; ip_input() will filter it later,
469 			 * thus deferring a hash lookup and mutex acquisition
470 			 * at the expense of a cheap copy using m_copym().
471 			 */
472 			ip_mloopback(ifp, m, hlen);
473 		} else {
474 			/*
475 			 * If we are acting as a multicast router, perform
476 			 * multicast forwarding as if the packet had just
477 			 * arrived on the interface to which we are about
478 			 * to send.  The multicast forwarding function
479 			 * recursively calls this function, using the
480 			 * IP_FORWARDING flag to prevent infinite recursion.
481 			 *
482 			 * Multicasts that are looped back by ip_mloopback(),
483 			 * above, will be forwarded by the ip_input() routine,
484 			 * if necessary.
485 			 */
486 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
487 				/*
488 				 * If rsvp daemon is not running, do not
489 				 * set ip_moptions. This ensures that the packet
490 				 * is multicast and not just sent down one link
491 				 * as prescribed by rsvpd.
492 				 */
493 				if (!V_rsvp_on)
494 					imo = NULL;
495 				if (ip_mforward &&
496 				    ip_mforward(ip, ifp, m, imo) != 0) {
497 					m_freem(m);
498 					goto done;
499 				}
500 			}
501 		}
502 
503 		/*
504 		 * Multicasts with a time-to-live of zero may be looped-
505 		 * back, above, but must not be transmitted on a network.
506 		 * Also, multicasts addressed to the loopback interface
507 		 * are not sent -- the above call to ip_mloopback() will
508 		 * loop back a copy. ip_input() will drop the copy if
509 		 * this host does not belong to the destination group on
510 		 * the loopback interface.
511 		 */
512 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
513 			m_freem(m);
514 			goto done;
515 		}
516 
517 		goto sendit;
518 	}
519 
520 	/*
521 	 * If the source address is not specified yet, use the address
522 	 * of the outoing interface.
523 	 */
524 	if (ip->ip_src.s_addr == INADDR_ANY) {
525 		/* Interface may have no addresses. */
526 		if (ia != NULL) {
527 			ip->ip_src = IA_SIN(ia)->sin_addr;
528 		}
529 	}
530 
531 	/*
532 	 * Look for broadcast address and
533 	 * verify user is allowed to send
534 	 * such a packet.
535 	 */
536 	if (isbroadcast) {
537 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
538 			error = EADDRNOTAVAIL;
539 			goto bad;
540 		}
541 		if ((flags & IP_ALLOWBROADCAST) == 0) {
542 			error = EACCES;
543 			goto bad;
544 		}
545 		/* don't allow broadcast messages to be fragmented */
546 		if (ip_len > mtu) {
547 			error = EMSGSIZE;
548 			goto bad;
549 		}
550 		m->m_flags |= M_BCAST;
551 	} else {
552 		m->m_flags &= ~M_BCAST;
553 	}
554 
555 sendit:
556 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
557 	if (IPSEC_ENABLED(ipv4)) {
558 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
559 			if (error == EINPROGRESS)
560 				error = 0;
561 			goto done;
562 		}
563 	}
564 	/*
565 	 * Check if there was a route for this packet; return error if not.
566 	 */
567 	if (no_route_but_check_spd) {
568 		IPSTAT_INC(ips_noroute);
569 		error = EHOSTUNREACH;
570 		goto bad;
571 	}
572 	/* Update variables that are affected by ipsec4_output(). */
573 	ip = mtod(m, struct ip *);
574 	hlen = ip->ip_hl << 2;
575 #endif /* IPSEC */
576 
577 	/* Jump over all PFIL processing if hooks are not active. */
578 	if (PFIL_HOOKED(&V_inet_pfil_hook)) {
579 		switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
580 		case 1: /* Finished */
581 			goto done;
582 
583 		case 0: /* Continue normally */
584 			ip = mtod(m, struct ip *);
585 			break;
586 
587 		case -1: /* Need to try again */
588 			/* Reset everything for a new round */
589 			RO_RTFREE(ro);
590 			if (have_ia_ref)
591 				ifa_free(&ia->ia_ifa);
592 			ro->ro_prepend = NULL;
593 			rte = NULL;
594 			gw = dst;
595 			ip = mtod(m, struct ip *);
596 			goto again;
597 
598 		}
599 	}
600 
601 	/* 127/8 must not appear on wire - RFC1122. */
602 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
603 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
604 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
605 			IPSTAT_INC(ips_badaddr);
606 			error = EADDRNOTAVAIL;
607 			goto bad;
608 		}
609 	}
610 
611 	m->m_pkthdr.csum_flags |= CSUM_IP;
612 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
613 		in_delayed_cksum(m);
614 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
615 	}
616 #ifdef SCTP
617 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
618 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
619 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
620 	}
621 #endif
622 
623 	/*
624 	 * If small enough for interface, or the interface will take
625 	 * care of the fragmentation for us, we can just send directly.
626 	 */
627 	if (ip_len <= mtu ||
628 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
629 		ip->ip_sum = 0;
630 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
631 			ip->ip_sum = in_cksum(m, hlen);
632 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
633 		}
634 
635 		/*
636 		 * Record statistics for this interface address.
637 		 * With CSUM_TSO the byte/packet count will be slightly
638 		 * incorrect because we count the IP+TCP headers only
639 		 * once instead of for every generated packet.
640 		 */
641 		if (!(flags & IP_FORWARDING) && ia) {
642 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
643 				counter_u64_add(ia->ia_ifa.ifa_opackets,
644 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
645 			else
646 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
647 
648 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
649 		}
650 #ifdef MBUF_STRESS_TEST
651 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
652 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
653 #endif
654 		/*
655 		 * Reset layer specific mbuf flags
656 		 * to avoid confusing lower layers.
657 		 */
658 		m_clrprotoflags(m);
659 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
660 #ifdef RATELIMIT
661 		if (inp != NULL) {
662 			if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
663 				in_pcboutput_txrtlmt(inp, ifp, m);
664 			/* stamp send tag on mbuf */
665 			m->m_pkthdr.snd_tag = inp->inp_snd_tag;
666 		} else {
667 			m->m_pkthdr.snd_tag = NULL;
668 		}
669 #endif
670 		error = (*ifp->if_output)(ifp, m,
671 		    (const struct sockaddr *)gw, ro);
672 #ifdef RATELIMIT
673 		/* check for route change */
674 		if (error == EAGAIN)
675 			in_pcboutput_eagain(inp);
676 #endif
677 		goto done;
678 	}
679 
680 	/* Balk when DF bit is set or the interface didn't support TSO. */
681 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
682 		error = EMSGSIZE;
683 		IPSTAT_INC(ips_cantfrag);
684 		goto bad;
685 	}
686 
687 	/*
688 	 * Too large for interface; fragment if possible. If successful,
689 	 * on return, m will point to a list of packets to be sent.
690 	 */
691 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
692 	if (error)
693 		goto bad;
694 	for (; m; m = m0) {
695 		m0 = m->m_nextpkt;
696 		m->m_nextpkt = 0;
697 		if (error == 0) {
698 			/* Record statistics for this interface address. */
699 			if (ia != NULL) {
700 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
701 				counter_u64_add(ia->ia_ifa.ifa_obytes,
702 				    m->m_pkthdr.len);
703 			}
704 			/*
705 			 * Reset layer specific mbuf flags
706 			 * to avoid confusing upper layers.
707 			 */
708 			m_clrprotoflags(m);
709 
710 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
711 			    mtod(m, struct ip *), NULL);
712 #ifdef RATELIMIT
713 			if (inp != NULL) {
714 				if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
715 					in_pcboutput_txrtlmt(inp, ifp, m);
716 				/* stamp send tag on mbuf */
717 				m->m_pkthdr.snd_tag = inp->inp_snd_tag;
718 			} else {
719 				m->m_pkthdr.snd_tag = NULL;
720 			}
721 #endif
722 			error = (*ifp->if_output)(ifp, m,
723 			    (const struct sockaddr *)gw, ro);
724 #ifdef RATELIMIT
725 			/* check for route change */
726 			if (error == EAGAIN)
727 				in_pcboutput_eagain(inp);
728 #endif
729 		} else
730 			m_freem(m);
731 	}
732 
733 	if (error == 0)
734 		IPSTAT_INC(ips_fragmented);
735 
736 done:
737 	if (ro == &iproute)
738 		RO_RTFREE(ro);
739 	else if (rte == NULL)
740 		/*
741 		 * If the caller supplied a route but somehow the reference
742 		 * to it has been released need to prevent the caller
743 		 * calling RTFREE on it again.
744 		 */
745 		ro->ro_rt = NULL;
746 	if (have_ia_ref)
747 		ifa_free(&ia->ia_ifa);
748 	return (error);
749 bad:
750 	m_freem(m);
751 	goto done;
752 }
753 
754 /*
755  * Create a chain of fragments which fit the given mtu. m_frag points to the
756  * mbuf to be fragmented; on return it points to the chain with the fragments.
757  * Return 0 if no error. If error, m_frag may contain a partially built
758  * chain of fragments that should be freed by the caller.
759  *
760  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
761  */
762 int
763 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
764     u_long if_hwassist_flags)
765 {
766 	int error = 0;
767 	int hlen = ip->ip_hl << 2;
768 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
769 	int off;
770 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
771 	int firstlen;
772 	struct mbuf **mnext;
773 	int nfrags;
774 	uint16_t ip_len, ip_off;
775 
776 	ip_len = ntohs(ip->ip_len);
777 	ip_off = ntohs(ip->ip_off);
778 
779 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
780 		IPSTAT_INC(ips_cantfrag);
781 		return EMSGSIZE;
782 	}
783 
784 	/*
785 	 * Must be able to put at least 8 bytes per fragment.
786 	 */
787 	if (len < 8)
788 		return EMSGSIZE;
789 
790 	/*
791 	 * If the interface will not calculate checksums on
792 	 * fragmented packets, then do it here.
793 	 */
794 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
795 		in_delayed_cksum(m0);
796 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
797 	}
798 #ifdef SCTP
799 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
800 		sctp_delayed_cksum(m0, hlen);
801 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
802 	}
803 #endif
804 	if (len > PAGE_SIZE) {
805 		/*
806 		 * Fragment large datagrams such that each segment
807 		 * contains a multiple of PAGE_SIZE amount of data,
808 		 * plus headers. This enables a receiver to perform
809 		 * page-flipping zero-copy optimizations.
810 		 *
811 		 * XXX When does this help given that sender and receiver
812 		 * could have different page sizes, and also mtu could
813 		 * be less than the receiver's page size ?
814 		 */
815 		int newlen;
816 
817 		off = MIN(mtu, m0->m_pkthdr.len);
818 
819 		/*
820 		 * firstlen (off - hlen) must be aligned on an
821 		 * 8-byte boundary
822 		 */
823 		if (off < hlen)
824 			goto smart_frag_failure;
825 		off = ((off - hlen) & ~7) + hlen;
826 		newlen = (~PAGE_MASK) & mtu;
827 		if ((newlen + sizeof (struct ip)) > mtu) {
828 			/* we failed, go back the default */
829 smart_frag_failure:
830 			newlen = len;
831 			off = hlen + len;
832 		}
833 		len = newlen;
834 
835 	} else {
836 		off = hlen + len;
837 	}
838 
839 	firstlen = off - hlen;
840 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
841 
842 	/*
843 	 * Loop through length of segment after first fragment,
844 	 * make new header and copy data of each part and link onto chain.
845 	 * Here, m0 is the original packet, m is the fragment being created.
846 	 * The fragments are linked off the m_nextpkt of the original
847 	 * packet, which after processing serves as the first fragment.
848 	 */
849 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
850 		struct ip *mhip;	/* ip header on the fragment */
851 		struct mbuf *m;
852 		int mhlen = sizeof (struct ip);
853 
854 		m = m_gethdr(M_NOWAIT, MT_DATA);
855 		if (m == NULL) {
856 			error = ENOBUFS;
857 			IPSTAT_INC(ips_odropped);
858 			goto done;
859 		}
860 		/*
861 		 * Make sure the complete packet header gets copied
862 		 * from the originating mbuf to the newly created
863 		 * mbuf. This also ensures that existing firewall
864 		 * classification(s), VLAN tags and so on get copied
865 		 * to the resulting fragmented packet(s):
866 		 */
867 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
868 			m_free(m);
869 			error = ENOBUFS;
870 			IPSTAT_INC(ips_odropped);
871 			goto done;
872 		}
873 		/*
874 		 * In the first mbuf, leave room for the link header, then
875 		 * copy the original IP header including options. The payload
876 		 * goes into an additional mbuf chain returned by m_copym().
877 		 */
878 		m->m_data += max_linkhdr;
879 		mhip = mtod(m, struct ip *);
880 		*mhip = *ip;
881 		if (hlen > sizeof (struct ip)) {
882 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
883 			mhip->ip_v = IPVERSION;
884 			mhip->ip_hl = mhlen >> 2;
885 		}
886 		m->m_len = mhlen;
887 		/* XXX do we need to add ip_off below ? */
888 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
889 		if (off + len >= ip_len)
890 			len = ip_len - off;
891 		else
892 			mhip->ip_off |= IP_MF;
893 		mhip->ip_len = htons((u_short)(len + mhlen));
894 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
895 		if (m->m_next == NULL) {	/* copy failed */
896 			m_free(m);
897 			error = ENOBUFS;	/* ??? */
898 			IPSTAT_INC(ips_odropped);
899 			goto done;
900 		}
901 		m->m_pkthdr.len = mhlen + len;
902 #ifdef MAC
903 		mac_netinet_fragment(m0, m);
904 #endif
905 		mhip->ip_off = htons(mhip->ip_off);
906 		mhip->ip_sum = 0;
907 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
908 			mhip->ip_sum = in_cksum(m, mhlen);
909 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
910 		}
911 		*mnext = m;
912 		mnext = &m->m_nextpkt;
913 	}
914 	IPSTAT_ADD(ips_ofragments, nfrags);
915 
916 	/*
917 	 * Update first fragment by trimming what's been copied out
918 	 * and updating header.
919 	 */
920 	m_adj(m0, hlen + firstlen - ip_len);
921 	m0->m_pkthdr.len = hlen + firstlen;
922 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
923 	ip->ip_off = htons(ip_off | IP_MF);
924 	ip->ip_sum = 0;
925 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
926 		ip->ip_sum = in_cksum(m0, hlen);
927 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
928 	}
929 
930 done:
931 	*m_frag = m0;
932 	return error;
933 }
934 
935 void
936 in_delayed_cksum(struct mbuf *m)
937 {
938 	struct ip *ip;
939 	uint16_t csum, offset, ip_len;
940 
941 	ip = mtod(m, struct ip *);
942 	offset = ip->ip_hl << 2 ;
943 	ip_len = ntohs(ip->ip_len);
944 	csum = in_cksum_skip(m, ip_len, offset);
945 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
946 		csum = 0xffff;
947 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
948 
949 	/* find the mbuf in the chain where the checksum starts*/
950 	while ((m != NULL) && (offset >= m->m_len)) {
951 		offset -= m->m_len;
952 		m = m->m_next;
953 	}
954 	KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
955 	KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
956 	*(u_short *)(m->m_data + offset) = csum;
957 }
958 
959 /*
960  * IP socket option processing.
961  */
962 int
963 ip_ctloutput(struct socket *so, struct sockopt *sopt)
964 {
965 	struct	inpcb *inp = sotoinpcb(so);
966 	int	error, optval;
967 #ifdef	RSS
968 	uint32_t rss_bucket;
969 	int retval;
970 #endif
971 
972 	error = optval = 0;
973 	if (sopt->sopt_level != IPPROTO_IP) {
974 		error = EINVAL;
975 
976 		if (sopt->sopt_level == SOL_SOCKET &&
977 		    sopt->sopt_dir == SOPT_SET) {
978 			switch (sopt->sopt_name) {
979 			case SO_REUSEADDR:
980 				INP_WLOCK(inp);
981 				if ((so->so_options & SO_REUSEADDR) != 0)
982 					inp->inp_flags2 |= INP_REUSEADDR;
983 				else
984 					inp->inp_flags2 &= ~INP_REUSEADDR;
985 				INP_WUNLOCK(inp);
986 				error = 0;
987 				break;
988 			case SO_REUSEPORT:
989 				INP_WLOCK(inp);
990 				if ((so->so_options & SO_REUSEPORT) != 0)
991 					inp->inp_flags2 |= INP_REUSEPORT;
992 				else
993 					inp->inp_flags2 &= ~INP_REUSEPORT;
994 				INP_WUNLOCK(inp);
995 				error = 0;
996 				break;
997 			case SO_SETFIB:
998 				INP_WLOCK(inp);
999 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1000 				INP_WUNLOCK(inp);
1001 				error = 0;
1002 				break;
1003 			case SO_MAX_PACING_RATE:
1004 #ifdef RATELIMIT
1005 				INP_WLOCK(inp);
1006 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1007 				INP_WUNLOCK(inp);
1008 				error = 0;
1009 #else
1010 				error = EOPNOTSUPP;
1011 #endif
1012 				break;
1013 			default:
1014 				break;
1015 			}
1016 		}
1017 		return (error);
1018 	}
1019 
1020 	switch (sopt->sopt_dir) {
1021 	case SOPT_SET:
1022 		switch (sopt->sopt_name) {
1023 		case IP_OPTIONS:
1024 #ifdef notyet
1025 		case IP_RETOPTS:
1026 #endif
1027 		{
1028 			struct mbuf *m;
1029 			if (sopt->sopt_valsize > MLEN) {
1030 				error = EMSGSIZE;
1031 				break;
1032 			}
1033 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1034 			if (m == NULL) {
1035 				error = ENOBUFS;
1036 				break;
1037 			}
1038 			m->m_len = sopt->sopt_valsize;
1039 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1040 					    m->m_len);
1041 			if (error) {
1042 				m_free(m);
1043 				break;
1044 			}
1045 			INP_WLOCK(inp);
1046 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1047 			INP_WUNLOCK(inp);
1048 			return (error);
1049 		}
1050 
1051 		case IP_BINDANY:
1052 			if (sopt->sopt_td != NULL) {
1053 				error = priv_check(sopt->sopt_td,
1054 				    PRIV_NETINET_BINDANY);
1055 				if (error)
1056 					break;
1057 			}
1058 			/* FALLTHROUGH */
1059 		case IP_BINDMULTI:
1060 #ifdef	RSS
1061 		case IP_RSS_LISTEN_BUCKET:
1062 #endif
1063 		case IP_TOS:
1064 		case IP_TTL:
1065 		case IP_MINTTL:
1066 		case IP_RECVOPTS:
1067 		case IP_RECVRETOPTS:
1068 		case IP_RECVDSTADDR:
1069 		case IP_RECVTTL:
1070 		case IP_RECVIF:
1071 		case IP_ONESBCAST:
1072 		case IP_DONTFRAG:
1073 		case IP_RECVTOS:
1074 		case IP_RECVFLOWID:
1075 #ifdef	RSS
1076 		case IP_RECVRSSBUCKETID:
1077 #endif
1078 			error = sooptcopyin(sopt, &optval, sizeof optval,
1079 					    sizeof optval);
1080 			if (error)
1081 				break;
1082 
1083 			switch (sopt->sopt_name) {
1084 			case IP_TOS:
1085 				inp->inp_ip_tos = optval;
1086 				break;
1087 
1088 			case IP_TTL:
1089 				inp->inp_ip_ttl = optval;
1090 				break;
1091 
1092 			case IP_MINTTL:
1093 				if (optval >= 0 && optval <= MAXTTL)
1094 					inp->inp_ip_minttl = optval;
1095 				else
1096 					error = EINVAL;
1097 				break;
1098 
1099 #define	OPTSET(bit) do {						\
1100 	INP_WLOCK(inp);							\
1101 	if (optval)							\
1102 		inp->inp_flags |= bit;					\
1103 	else								\
1104 		inp->inp_flags &= ~bit;					\
1105 	INP_WUNLOCK(inp);						\
1106 } while (0)
1107 
1108 #define	OPTSET2(bit, val) do {						\
1109 	INP_WLOCK(inp);							\
1110 	if (val)							\
1111 		inp->inp_flags2 |= bit;					\
1112 	else								\
1113 		inp->inp_flags2 &= ~bit;				\
1114 	INP_WUNLOCK(inp);						\
1115 } while (0)
1116 
1117 			case IP_RECVOPTS:
1118 				OPTSET(INP_RECVOPTS);
1119 				break;
1120 
1121 			case IP_RECVRETOPTS:
1122 				OPTSET(INP_RECVRETOPTS);
1123 				break;
1124 
1125 			case IP_RECVDSTADDR:
1126 				OPTSET(INP_RECVDSTADDR);
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_RECVDSTADDR:
1262 		case IP_RECVTTL:
1263 		case IP_RECVIF:
1264 		case IP_PORTRANGE:
1265 		case IP_ONESBCAST:
1266 		case IP_DONTFRAG:
1267 		case IP_BINDANY:
1268 		case IP_RECVTOS:
1269 		case IP_BINDMULTI:
1270 		case IP_FLOWID:
1271 		case IP_FLOWTYPE:
1272 		case IP_RECVFLOWID:
1273 #ifdef	RSS
1274 		case IP_RSSBUCKETID:
1275 		case IP_RECVRSSBUCKETID:
1276 #endif
1277 			switch (sopt->sopt_name) {
1278 
1279 			case IP_TOS:
1280 				optval = inp->inp_ip_tos;
1281 				break;
1282 
1283 			case IP_TTL:
1284 				optval = inp->inp_ip_ttl;
1285 				break;
1286 
1287 			case IP_MINTTL:
1288 				optval = inp->inp_ip_minttl;
1289 				break;
1290 
1291 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1292 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1293 
1294 			case IP_RECVOPTS:
1295 				optval = OPTBIT(INP_RECVOPTS);
1296 				break;
1297 
1298 			case IP_RECVRETOPTS:
1299 				optval = OPTBIT(INP_RECVRETOPTS);
1300 				break;
1301 
1302 			case IP_RECVDSTADDR:
1303 				optval = OPTBIT(INP_RECVDSTADDR);
1304 				break;
1305 
1306 			case IP_RECVTTL:
1307 				optval = OPTBIT(INP_RECVTTL);
1308 				break;
1309 
1310 			case IP_RECVIF:
1311 				optval = OPTBIT(INP_RECVIF);
1312 				break;
1313 
1314 			case IP_PORTRANGE:
1315 				if (inp->inp_flags & INP_HIGHPORT)
1316 					optval = IP_PORTRANGE_HIGH;
1317 				else if (inp->inp_flags & INP_LOWPORT)
1318 					optval = IP_PORTRANGE_LOW;
1319 				else
1320 					optval = 0;
1321 				break;
1322 
1323 			case IP_ONESBCAST:
1324 				optval = OPTBIT(INP_ONESBCAST);
1325 				break;
1326 			case IP_DONTFRAG:
1327 				optval = OPTBIT(INP_DONTFRAG);
1328 				break;
1329 			case IP_BINDANY:
1330 				optval = OPTBIT(INP_BINDANY);
1331 				break;
1332 			case IP_RECVTOS:
1333 				optval = OPTBIT(INP_RECVTOS);
1334 				break;
1335 			case IP_FLOWID:
1336 				optval = inp->inp_flowid;
1337 				break;
1338 			case IP_FLOWTYPE:
1339 				optval = inp->inp_flowtype;
1340 				break;
1341 			case IP_RECVFLOWID:
1342 				optval = OPTBIT2(INP_RECVFLOWID);
1343 				break;
1344 #ifdef	RSS
1345 			case IP_RSSBUCKETID:
1346 				retval = rss_hash2bucket(inp->inp_flowid,
1347 				    inp->inp_flowtype,
1348 				    &rss_bucket);
1349 				if (retval == 0)
1350 					optval = rss_bucket;
1351 				else
1352 					error = EINVAL;
1353 				break;
1354 			case IP_RECVRSSBUCKETID:
1355 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1356 				break;
1357 #endif
1358 			case IP_BINDMULTI:
1359 				optval = OPTBIT2(INP_BINDMULTI);
1360 				break;
1361 			}
1362 			error = sooptcopyout(sopt, &optval, sizeof optval);
1363 			break;
1364 
1365 		/*
1366 		 * Multicast socket options are processed by the in_mcast
1367 		 * module.
1368 		 */
1369 		case IP_MULTICAST_IF:
1370 		case IP_MULTICAST_VIF:
1371 		case IP_MULTICAST_TTL:
1372 		case IP_MULTICAST_LOOP:
1373 		case IP_MSFILTER:
1374 			error = inp_getmoptions(inp, sopt);
1375 			break;
1376 
1377 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1378 		case IP_IPSEC_POLICY:
1379 			if (IPSEC_ENABLED(ipv4)) {
1380 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1381 				break;
1382 			}
1383 			/* FALLTHROUGH */
1384 #endif /* IPSEC */
1385 
1386 		default:
1387 			error = ENOPROTOOPT;
1388 			break;
1389 		}
1390 		break;
1391 	}
1392 	return (error);
1393 }
1394 
1395 /*
1396  * Routine called from ip_output() to loop back a copy of an IP multicast
1397  * packet to the input queue of a specified interface.  Note that this
1398  * calls the output routine of the loopback "driver", but with an interface
1399  * pointer that might NOT be a loopback interface -- evil, but easier than
1400  * replicating that code here.
1401  */
1402 static void
1403 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1404 {
1405 	struct ip *ip;
1406 	struct mbuf *copym;
1407 
1408 	/*
1409 	 * Make a deep copy of the packet because we're going to
1410 	 * modify the pack in order to generate checksums.
1411 	 */
1412 	copym = m_dup(m, M_NOWAIT);
1413 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1414 		copym = m_pullup(copym, hlen);
1415 	if (copym != NULL) {
1416 		/* If needed, compute the checksum and mark it as valid. */
1417 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1418 			in_delayed_cksum(copym);
1419 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1420 			copym->m_pkthdr.csum_flags |=
1421 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1422 			copym->m_pkthdr.csum_data = 0xffff;
1423 		}
1424 		/*
1425 		 * We don't bother to fragment if the IP length is greater
1426 		 * than the interface's MTU.  Can this possibly matter?
1427 		 */
1428 		ip = mtod(copym, struct ip *);
1429 		ip->ip_sum = 0;
1430 		ip->ip_sum = in_cksum(copym, hlen);
1431 		if_simloop(ifp, copym, AF_INET, 0);
1432 	}
1433 }
1434