xref: /freebsd/sys/netinet/ip_output.c (revision bdafb02fcb88389fd1ab684cfe734cb429d35618)
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 
84 #include <netinet/udp.h>
85 #include <netinet/udp_var.h>
86 
87 #ifdef SCTP
88 #include <netinet/sctp.h>
89 #include <netinet/sctp_crc32.h>
90 #endif
91 
92 #include <netipsec/ipsec_support.h>
93 
94 #include <machine/in_cksum.h>
95 
96 #include <security/mac/mac_framework.h>
97 
98 #ifdef MBUF_STRESS_TEST
99 static int mbuf_frag_size = 0;
100 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
101 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
102 #endif
103 
104 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
105 
106 
107 extern int in_mcast_loop;
108 extern	struct protosw inetsw[];
109 
110 static inline int
111 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
112     struct sockaddr_in *dst, int *fibnum, int *error)
113 {
114 	struct m_tag *fwd_tag = NULL;
115 	struct mbuf *m;
116 	struct in_addr odst;
117 	struct ip *ip;
118 
119 	m = *mp;
120 	ip = mtod(m, struct ip *);
121 
122 	/* Run through list of hooks for output packets. */
123 	odst.s_addr = ip->ip_dst.s_addr;
124 	*error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, 0, inp);
125 	m = *mp;
126 	if ((*error) != 0 || m == NULL)
127 		return 1; /* Finished */
128 
129 	ip = mtod(m, struct ip *);
130 
131 	/* See if destination IP address was changed by packet filter. */
132 	if (odst.s_addr != ip->ip_dst.s_addr) {
133 		m->m_flags |= M_SKIP_FIREWALL;
134 		/* If destination is now ourself drop to ip_input(). */
135 		if (in_localip(ip->ip_dst)) {
136 			m->m_flags |= M_FASTFWD_OURS;
137 			if (m->m_pkthdr.rcvif == NULL)
138 				m->m_pkthdr.rcvif = V_loif;
139 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
140 				m->m_pkthdr.csum_flags |=
141 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
142 				m->m_pkthdr.csum_data = 0xffff;
143 			}
144 			m->m_pkthdr.csum_flags |=
145 				CSUM_IP_CHECKED | CSUM_IP_VALID;
146 #ifdef SCTP
147 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
148 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
149 #endif
150 			*error = netisr_queue(NETISR_IP, m);
151 			return 1; /* Finished */
152 		}
153 
154 		bzero(dst, sizeof(*dst));
155 		dst->sin_family = AF_INET;
156 		dst->sin_len = sizeof(*dst);
157 		dst->sin_addr = ip->ip_dst;
158 
159 		return -1; /* Reloop */
160 	}
161 	/* See if fib was changed by packet filter. */
162 	if ((*fibnum) != M_GETFIB(m)) {
163 		m->m_flags |= M_SKIP_FIREWALL;
164 		*fibnum = M_GETFIB(m);
165 		return -1; /* Reloop for FIB change */
166 	}
167 
168 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
169 	if (m->m_flags & M_FASTFWD_OURS) {
170 		if (m->m_pkthdr.rcvif == NULL)
171 			m->m_pkthdr.rcvif = V_loif;
172 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
173 			m->m_pkthdr.csum_flags |=
174 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
175 			m->m_pkthdr.csum_data = 0xffff;
176 		}
177 #ifdef SCTP
178 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
179 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
180 #endif
181 		m->m_pkthdr.csum_flags |=
182 			CSUM_IP_CHECKED | CSUM_IP_VALID;
183 
184 		*error = netisr_queue(NETISR_IP, m);
185 		return 1; /* Finished */
186 	}
187 	/* Or forward to some other address? */
188 	if ((m->m_flags & M_IP_NEXTHOP) &&
189 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
190 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
191 		m->m_flags |= M_SKIP_FIREWALL;
192 		m->m_flags &= ~M_IP_NEXTHOP;
193 		m_tag_delete(m, fwd_tag);
194 
195 		return -1; /* Reloop for CHANGE of dst */
196 	}
197 
198 	return 0;
199 }
200 
201 /*
202  * IP output.  The packet in mbuf chain m contains a skeletal IP
203  * header (with len, off, ttl, proto, tos, src, dst).
204  * The mbuf chain containing the packet will be freed.
205  * The mbuf opt, if present, will not be freed.
206  * If route ro is present and has ro_rt initialized, route lookup would be
207  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
208  * then result of route lookup is stored in ro->ro_rt.
209  *
210  * In the IP forwarding case, the packet will arrive with options already
211  * inserted, so must have a NULL opt pointer.
212  */
213 int
214 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
215     struct ip_moptions *imo, struct inpcb *inp)
216 {
217 	struct rm_priotracker in_ifa_tracker;
218 	struct ip *ip;
219 	struct ifnet *ifp = NULL;	/* keep compiler happy */
220 	struct mbuf *m0;
221 	int hlen = sizeof (struct ip);
222 	int mtu;
223 	int error = 0;
224 	struct sockaddr_in *dst;
225 	const struct sockaddr_in *gw;
226 	struct in_ifaddr *ia;
227 	int isbroadcast;
228 	uint16_t ip_len, ip_off;
229 	struct route iproute;
230 	struct rtentry *rte;	/* cache for ro->ro_rt */
231 	uint32_t fibnum;
232 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
233 	int no_route_but_check_spd = 0;
234 #endif
235 	M_ASSERTPKTHDR(m);
236 
237 	if (inp != NULL) {
238 		INP_LOCK_ASSERT(inp);
239 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
240 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
241 			m->m_pkthdr.flowid = inp->inp_flowid;
242 			M_HASHTYPE_SET(m, inp->inp_flowtype);
243 		}
244 	}
245 
246 	if (ro == NULL) {
247 		ro = &iproute;
248 		bzero(ro, sizeof (*ro));
249 	}
250 
251 	if (opt) {
252 		int len = 0;
253 		m = ip_insertoptions(m, opt, &len);
254 		if (len != 0)
255 			hlen = len; /* ip->ip_hl is updated above */
256 	}
257 	ip = mtod(m, struct ip *);
258 	ip_len = ntohs(ip->ip_len);
259 	ip_off = ntohs(ip->ip_off);
260 
261 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
262 		ip->ip_v = IPVERSION;
263 		ip->ip_hl = hlen >> 2;
264 		ip_fillid(ip);
265 	} else {
266 		/* Header already set, fetch hlen from there */
267 		hlen = ip->ip_hl << 2;
268 	}
269 	if ((flags & IP_FORWARDING) == 0)
270 		IPSTAT_INC(ips_localout);
271 
272 	/*
273 	 * dst/gw handling:
274 	 *
275 	 * dst can be rewritten but always points to &ro->ro_dst.
276 	 * gw is readonly but can point either to dst OR rt_gateway,
277 	 * therefore we need restore gw if we're redoing lookup.
278 	 */
279 	gw = dst = (struct sockaddr_in *)&ro->ro_dst;
280 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
281 	rte = ro->ro_rt;
282 	if (rte == NULL) {
283 		bzero(dst, sizeof(*dst));
284 		dst->sin_family = AF_INET;
285 		dst->sin_len = sizeof(*dst);
286 		dst->sin_addr = ip->ip_dst;
287 	}
288 	NET_EPOCH_ENTER();
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 		RO_INVALIDATE_CACHE(ro);
311 		rte = NULL;
312 	}
313 	ia = NULL;
314 	/*
315 	 * If routing to interface only, short circuit routing lookup.
316 	 * The use of an all-ones broadcast address implies this; an
317 	 * interface is specified by the broadcast address of an interface,
318 	 * or the destination address of a ptp interface.
319 	 */
320 	if (flags & IP_SENDONES) {
321 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
322 						      M_GETFIB(m)))) == NULL &&
323 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
324 						    M_GETFIB(m)))) == NULL) {
325 			IPSTAT_INC(ips_noroute);
326 			error = ENETUNREACH;
327 			goto bad;
328 		}
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 		ifp = ia->ia_ifp;
344 		ip->ip_ttl = 1;
345 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
346 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
347 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
348 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
349 		/*
350 		 * Bypass the normal routing lookup for multicast
351 		 * packets if the interface is specified.
352 		 */
353 		ifp = imo->imo_multicast_ifp;
354 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
355 		isbroadcast = 0;	/* fool gcc */
356 	} else {
357 		/*
358 		 * We want to do any cloning requested by the link layer,
359 		 * as this is probably required in all cases for correct
360 		 * operation (as it is for ARP).
361 		 */
362 		if (rte == NULL) {
363 #ifdef RADIX_MPATH
364 			rtalloc_mpath_fib(ro,
365 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
366 			    fibnum);
367 #else
368 			in_rtalloc_ign(ro, 0, fibnum);
369 #endif
370 			rte = ro->ro_rt;
371 		}
372 		if (rte == NULL ||
373 		    (rte->rt_flags & RTF_UP) == 0 ||
374 		    rte->rt_ifp == NULL ||
375 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
376 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
377 			/*
378 			 * There is no route for this packet, but it is
379 			 * possible that a matching SPD entry exists.
380 			 */
381 			no_route_but_check_spd = 1;
382 			mtu = 0; /* Silence GCC warning. */
383 			goto sendit;
384 #endif
385 			IPSTAT_INC(ips_noroute);
386 			error = EHOSTUNREACH;
387 			goto bad;
388 		}
389 		ia = ifatoia(rte->rt_ifa);
390 		ifp = rte->rt_ifp;
391 		counter_u64_add(rte->rt_pksent, 1);
392 		rt_update_ro_flags(ro);
393 		if (rte->rt_flags & RTF_GATEWAY)
394 			gw = (struct sockaddr_in *)rte->rt_gateway;
395 		if (rte->rt_flags & RTF_HOST)
396 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
397 		else if (ifp->if_flags & IFF_BROADCAST)
398 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
399 		else
400 			isbroadcast = 0;
401 	}
402 
403 	/*
404 	 * Calculate MTU.  If we have a route that is up, use that,
405 	 * otherwise use the interface's MTU.
406 	 */
407 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
408 		mtu = rte->rt_mtu;
409 	else
410 		mtu = ifp->if_mtu;
411 	/* Catch a possible divide by zero later. */
412 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
413 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
414 
415 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
416 		m->m_flags |= M_MCAST;
417 		/*
418 		 * IP destination address is multicast.  Make sure "gw"
419 		 * still points to the address in "ro".  (It may have been
420 		 * changed to point to a gateway address, above.)
421 		 */
422 		gw = dst;
423 		/*
424 		 * See if the caller provided any multicast options
425 		 */
426 		if (imo != NULL) {
427 			ip->ip_ttl = imo->imo_multicast_ttl;
428 			if (imo->imo_multicast_vif != -1)
429 				ip->ip_src.s_addr =
430 				    ip_mcast_src ?
431 				    ip_mcast_src(imo->imo_multicast_vif) :
432 				    INADDR_ANY;
433 		} else
434 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
435 		/*
436 		 * Confirm that the outgoing interface supports multicast.
437 		 */
438 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
439 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
440 				IPSTAT_INC(ips_noroute);
441 				error = ENETUNREACH;
442 				goto bad;
443 			}
444 		}
445 		/*
446 		 * If source address not specified yet, use address
447 		 * of outgoing interface.
448 		 */
449 		if (ip->ip_src.s_addr == INADDR_ANY) {
450 			/* Interface may have no addresses. */
451 			if (ia != NULL)
452 				ip->ip_src = IA_SIN(ia)->sin_addr;
453 		}
454 
455 		if ((imo == NULL && in_mcast_loop) ||
456 		    (imo && imo->imo_multicast_loop)) {
457 			/*
458 			 * Loop back multicast datagram if not expressly
459 			 * forbidden to do so, even if we are not a member
460 			 * of the group; ip_input() will filter it later,
461 			 * thus deferring a hash lookup and mutex acquisition
462 			 * at the expense of a cheap copy using m_copym().
463 			 */
464 			ip_mloopback(ifp, m, hlen);
465 		} else {
466 			/*
467 			 * If we are acting as a multicast router, perform
468 			 * multicast forwarding as if the packet had just
469 			 * arrived on the interface to which we are about
470 			 * to send.  The multicast forwarding function
471 			 * recursively calls this function, using the
472 			 * IP_FORWARDING flag to prevent infinite recursion.
473 			 *
474 			 * Multicasts that are looped back by ip_mloopback(),
475 			 * above, will be forwarded by the ip_input() routine,
476 			 * if necessary.
477 			 */
478 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
479 				/*
480 				 * If rsvp daemon is not running, do not
481 				 * set ip_moptions. This ensures that the packet
482 				 * is multicast and not just sent down one link
483 				 * as prescribed by rsvpd.
484 				 */
485 				if (!V_rsvp_on)
486 					imo = NULL;
487 				if (ip_mforward &&
488 				    ip_mforward(ip, ifp, m, imo) != 0) {
489 					m_freem(m);
490 					goto done;
491 				}
492 			}
493 		}
494 
495 		/*
496 		 * Multicasts with a time-to-live of zero may be looped-
497 		 * back, above, but must not be transmitted on a network.
498 		 * Also, multicasts addressed to the loopback interface
499 		 * are not sent -- the above call to ip_mloopback() will
500 		 * loop back a copy. ip_input() will drop the copy if
501 		 * this host does not belong to the destination group on
502 		 * the loopback interface.
503 		 */
504 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
505 			m_freem(m);
506 			goto done;
507 		}
508 
509 		goto sendit;
510 	}
511 
512 	/*
513 	 * If the source address is not specified yet, use the address
514 	 * of the outoing interface.
515 	 */
516 	if (ip->ip_src.s_addr == INADDR_ANY) {
517 		/* Interface may have no addresses. */
518 		if (ia != NULL) {
519 			ip->ip_src = IA_SIN(ia)->sin_addr;
520 		}
521 	}
522 
523 	/*
524 	 * Look for broadcast address and
525 	 * verify user is allowed to send
526 	 * such a packet.
527 	 */
528 	if (isbroadcast) {
529 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
530 			error = EADDRNOTAVAIL;
531 			goto bad;
532 		}
533 		if ((flags & IP_ALLOWBROADCAST) == 0) {
534 			error = EACCES;
535 			goto bad;
536 		}
537 		/* don't allow broadcast messages to be fragmented */
538 		if (ip_len > mtu) {
539 			error = EMSGSIZE;
540 			goto bad;
541 		}
542 		m->m_flags |= M_BCAST;
543 	} else {
544 		m->m_flags &= ~M_BCAST;
545 	}
546 
547 sendit:
548 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
549 	if (IPSEC_ENABLED(ipv4)) {
550 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
551 			if (error == EINPROGRESS)
552 				error = 0;
553 			goto done;
554 		}
555 	}
556 	/*
557 	 * Check if there was a route for this packet; return error if not.
558 	 */
559 	if (no_route_but_check_spd) {
560 		IPSTAT_INC(ips_noroute);
561 		error = EHOSTUNREACH;
562 		goto bad;
563 	}
564 	/* Update variables that are affected by ipsec4_output(). */
565 	ip = mtod(m, struct ip *);
566 	hlen = ip->ip_hl << 2;
567 #endif /* IPSEC */
568 
569 	/* Jump over all PFIL processing if hooks are not active. */
570 	if (PFIL_HOOKED(&V_inet_pfil_hook)) {
571 		switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
572 		case 1: /* Finished */
573 			goto done;
574 
575 		case 0: /* Continue normally */
576 			ip = mtod(m, struct ip *);
577 			break;
578 
579 		case -1: /* Need to try again */
580 			/* Reset everything for a new round */
581 			RO_RTFREE(ro);
582 			ro->ro_prepend = NULL;
583 			rte = NULL;
584 			gw = dst;
585 			ip = mtod(m, struct ip *);
586 			goto again;
587 
588 		}
589 	}
590 
591 	/* 127/8 must not appear on wire - RFC1122. */
592 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
593 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
594 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
595 			IPSTAT_INC(ips_badaddr);
596 			error = EADDRNOTAVAIL;
597 			goto bad;
598 		}
599 	}
600 
601 	m->m_pkthdr.csum_flags |= CSUM_IP;
602 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
603 		in_delayed_cksum(m);
604 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
605 	}
606 #ifdef SCTP
607 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
608 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
609 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
610 	}
611 #endif
612 
613 	/*
614 	 * If small enough for interface, or the interface will take
615 	 * care of the fragmentation for us, we can just send directly.
616 	 */
617 	if (ip_len <= mtu ||
618 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
619 		ip->ip_sum = 0;
620 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
621 			ip->ip_sum = in_cksum(m, hlen);
622 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
623 		}
624 
625 		/*
626 		 * Record statistics for this interface address.
627 		 * With CSUM_TSO the byte/packet count will be slightly
628 		 * incorrect because we count the IP+TCP headers only
629 		 * once instead of for every generated packet.
630 		 */
631 		if (!(flags & IP_FORWARDING) && ia) {
632 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
633 				counter_u64_add(ia->ia_ifa.ifa_opackets,
634 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
635 			else
636 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
637 
638 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
639 		}
640 #ifdef MBUF_STRESS_TEST
641 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
642 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
643 #endif
644 		/*
645 		 * Reset layer specific mbuf flags
646 		 * to avoid confusing lower layers.
647 		 */
648 		m_clrprotoflags(m);
649 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
650 #ifdef RATELIMIT
651 		if (inp != NULL) {
652 			if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
653 				in_pcboutput_txrtlmt(inp, ifp, m);
654 			/* stamp send tag on mbuf */
655 			m->m_pkthdr.snd_tag = inp->inp_snd_tag;
656 		} else {
657 			m->m_pkthdr.snd_tag = NULL;
658 		}
659 #endif
660 		error = (*ifp->if_output)(ifp, m,
661 		    (const struct sockaddr *)gw, ro);
662 #ifdef RATELIMIT
663 		/* check for route change */
664 		if (error == EAGAIN)
665 			in_pcboutput_eagain(inp);
666 #endif
667 		goto done;
668 	}
669 
670 	/* Balk when DF bit is set or the interface didn't support TSO. */
671 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
672 		error = EMSGSIZE;
673 		IPSTAT_INC(ips_cantfrag);
674 		goto bad;
675 	}
676 
677 	/*
678 	 * Too large for interface; fragment if possible. If successful,
679 	 * on return, m will point to a list of packets to be sent.
680 	 */
681 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
682 	if (error)
683 		goto bad;
684 	for (; m; m = m0) {
685 		m0 = m->m_nextpkt;
686 		m->m_nextpkt = 0;
687 		if (error == 0) {
688 			/* Record statistics for this interface address. */
689 			if (ia != NULL) {
690 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
691 				counter_u64_add(ia->ia_ifa.ifa_obytes,
692 				    m->m_pkthdr.len);
693 			}
694 			/*
695 			 * Reset layer specific mbuf flags
696 			 * to avoid confusing upper layers.
697 			 */
698 			m_clrprotoflags(m);
699 
700 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
701 			    mtod(m, struct ip *), NULL);
702 #ifdef RATELIMIT
703 			if (inp != NULL) {
704 				if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
705 					in_pcboutput_txrtlmt(inp, ifp, m);
706 				/* stamp send tag on mbuf */
707 				m->m_pkthdr.snd_tag = inp->inp_snd_tag;
708 			} else {
709 				m->m_pkthdr.snd_tag = NULL;
710 			}
711 #endif
712 			error = (*ifp->if_output)(ifp, m,
713 			    (const struct sockaddr *)gw, ro);
714 #ifdef RATELIMIT
715 			/* check for route change */
716 			if (error == EAGAIN)
717 				in_pcboutput_eagain(inp);
718 #endif
719 		} else
720 			m_freem(m);
721 	}
722 
723 	if (error == 0)
724 		IPSTAT_INC(ips_fragmented);
725 
726 done:
727 	if (ro == &iproute)
728 		RO_RTFREE(ro);
729 	else if (rte == NULL)
730 		/*
731 		 * If the caller supplied a route but somehow the reference
732 		 * to it has been released need to prevent the caller
733 		 * calling RTFREE on it again.
734 		 */
735 		ro->ro_rt = NULL;
736 	NET_EPOCH_EXIT();
737 	return (error);
738  bad:
739 	m_freem(m);
740 	goto done;
741 }
742 
743 /*
744  * Create a chain of fragments which fit the given mtu. m_frag points to the
745  * mbuf to be fragmented; on return it points to the chain with the fragments.
746  * Return 0 if no error. If error, m_frag may contain a partially built
747  * chain of fragments that should be freed by the caller.
748  *
749  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
750  */
751 int
752 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
753     u_long if_hwassist_flags)
754 {
755 	int error = 0;
756 	int hlen = ip->ip_hl << 2;
757 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
758 	int off;
759 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
760 	int firstlen;
761 	struct mbuf **mnext;
762 	int nfrags;
763 	uint16_t ip_len, ip_off;
764 
765 	ip_len = ntohs(ip->ip_len);
766 	ip_off = ntohs(ip->ip_off);
767 
768 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
769 		IPSTAT_INC(ips_cantfrag);
770 		return EMSGSIZE;
771 	}
772 
773 	/*
774 	 * Must be able to put at least 8 bytes per fragment.
775 	 */
776 	if (len < 8)
777 		return EMSGSIZE;
778 
779 	/*
780 	 * If the interface will not calculate checksums on
781 	 * fragmented packets, then do it here.
782 	 */
783 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
784 		in_delayed_cksum(m0);
785 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
786 	}
787 #ifdef SCTP
788 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
789 		sctp_delayed_cksum(m0, hlen);
790 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
791 	}
792 #endif
793 	if (len > PAGE_SIZE) {
794 		/*
795 		 * Fragment large datagrams such that each segment
796 		 * contains a multiple of PAGE_SIZE amount of data,
797 		 * plus headers. This enables a receiver to perform
798 		 * page-flipping zero-copy optimizations.
799 		 *
800 		 * XXX When does this help given that sender and receiver
801 		 * could have different page sizes, and also mtu could
802 		 * be less than the receiver's page size ?
803 		 */
804 		int newlen;
805 
806 		off = MIN(mtu, m0->m_pkthdr.len);
807 
808 		/*
809 		 * firstlen (off - hlen) must be aligned on an
810 		 * 8-byte boundary
811 		 */
812 		if (off < hlen)
813 			goto smart_frag_failure;
814 		off = ((off - hlen) & ~7) + hlen;
815 		newlen = (~PAGE_MASK) & mtu;
816 		if ((newlen + sizeof (struct ip)) > mtu) {
817 			/* we failed, go back the default */
818 smart_frag_failure:
819 			newlen = len;
820 			off = hlen + len;
821 		}
822 		len = newlen;
823 
824 	} else {
825 		off = hlen + len;
826 	}
827 
828 	firstlen = off - hlen;
829 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
830 
831 	/*
832 	 * Loop through length of segment after first fragment,
833 	 * make new header and copy data of each part and link onto chain.
834 	 * Here, m0 is the original packet, m is the fragment being created.
835 	 * The fragments are linked off the m_nextpkt of the original
836 	 * packet, which after processing serves as the first fragment.
837 	 */
838 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
839 		struct ip *mhip;	/* ip header on the fragment */
840 		struct mbuf *m;
841 		int mhlen = sizeof (struct ip);
842 
843 		m = m_gethdr(M_NOWAIT, MT_DATA);
844 		if (m == NULL) {
845 			error = ENOBUFS;
846 			IPSTAT_INC(ips_odropped);
847 			goto done;
848 		}
849 		/*
850 		 * Make sure the complete packet header gets copied
851 		 * from the originating mbuf to the newly created
852 		 * mbuf. This also ensures that existing firewall
853 		 * classification(s), VLAN tags and so on get copied
854 		 * to the resulting fragmented packet(s):
855 		 */
856 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
857 			m_free(m);
858 			error = ENOBUFS;
859 			IPSTAT_INC(ips_odropped);
860 			goto done;
861 		}
862 		/*
863 		 * In the first mbuf, leave room for the link header, then
864 		 * copy the original IP header including options. The payload
865 		 * goes into an additional mbuf chain returned by m_copym().
866 		 */
867 		m->m_data += max_linkhdr;
868 		mhip = mtod(m, struct ip *);
869 		*mhip = *ip;
870 		if (hlen > sizeof (struct ip)) {
871 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
872 			mhip->ip_v = IPVERSION;
873 			mhip->ip_hl = mhlen >> 2;
874 		}
875 		m->m_len = mhlen;
876 		/* XXX do we need to add ip_off below ? */
877 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
878 		if (off + len >= ip_len)
879 			len = ip_len - off;
880 		else
881 			mhip->ip_off |= IP_MF;
882 		mhip->ip_len = htons((u_short)(len + mhlen));
883 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
884 		if (m->m_next == NULL) {	/* copy failed */
885 			m_free(m);
886 			error = ENOBUFS;	/* ??? */
887 			IPSTAT_INC(ips_odropped);
888 			goto done;
889 		}
890 		m->m_pkthdr.len = mhlen + len;
891 #ifdef MAC
892 		mac_netinet_fragment(m0, m);
893 #endif
894 		mhip->ip_off = htons(mhip->ip_off);
895 		mhip->ip_sum = 0;
896 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
897 			mhip->ip_sum = in_cksum(m, mhlen);
898 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
899 		}
900 		*mnext = m;
901 		mnext = &m->m_nextpkt;
902 	}
903 	IPSTAT_ADD(ips_ofragments, nfrags);
904 
905 	/*
906 	 * Update first fragment by trimming what's been copied out
907 	 * and updating header.
908 	 */
909 	m_adj(m0, hlen + firstlen - ip_len);
910 	m0->m_pkthdr.len = hlen + firstlen;
911 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
912 	ip->ip_off = htons(ip_off | IP_MF);
913 	ip->ip_sum = 0;
914 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
915 		ip->ip_sum = in_cksum(m0, hlen);
916 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
917 	}
918 
919 done:
920 	*m_frag = m0;
921 	return error;
922 }
923 
924 void
925 in_delayed_cksum(struct mbuf *m)
926 {
927 	struct ip *ip;
928 	struct udphdr *uh;
929 	uint16_t cklen, csum, offset;
930 
931 	ip = mtod(m, struct ip *);
932 	offset = ip->ip_hl << 2 ;
933 
934 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
935 		/* if udp header is not in the first mbuf copy udplen */
936 		if (offset + sizeof(struct udphdr) > m->m_len) {
937 			m_copydata(m, offset + offsetof(struct udphdr,
938 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
939 			cklen = ntohs(cklen);
940 		} else {
941 			uh = (struct udphdr *)mtodo(m, offset);
942 			cklen = ntohs(uh->uh_ulen);
943 		}
944 		csum = in_cksum_skip(m, cklen + offset, offset);
945 		if (csum == 0)
946 			csum = 0xffff;
947 	} else {
948 		cklen = ntohs(ip->ip_len);
949 		csum = in_cksum_skip(m, cklen, offset);
950 	}
951 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
952 
953 	if (offset + sizeof(csum) > m->m_len)
954 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
955 	else
956 		*(u_short *)mtodo(m, 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_REUSEPORT_LB:
998 				INP_WLOCK(inp);
999 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1000 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1001 				else
1002 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1003 				INP_WUNLOCK(inp);
1004 				error = 0;
1005 				break;
1006 			case SO_SETFIB:
1007 				INP_WLOCK(inp);
1008 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1009 				INP_WUNLOCK(inp);
1010 				error = 0;
1011 				break;
1012 			case SO_MAX_PACING_RATE:
1013 #ifdef RATELIMIT
1014 				INP_WLOCK(inp);
1015 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1016 				INP_WUNLOCK(inp);
1017 				error = 0;
1018 #else
1019 				error = EOPNOTSUPP;
1020 #endif
1021 				break;
1022 			default:
1023 				break;
1024 			}
1025 		}
1026 		return (error);
1027 	}
1028 
1029 	switch (sopt->sopt_dir) {
1030 	case SOPT_SET:
1031 		switch (sopt->sopt_name) {
1032 		case IP_OPTIONS:
1033 #ifdef notyet
1034 		case IP_RETOPTS:
1035 #endif
1036 		{
1037 			struct mbuf *m;
1038 			if (sopt->sopt_valsize > MLEN) {
1039 				error = EMSGSIZE;
1040 				break;
1041 			}
1042 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1043 			if (m == NULL) {
1044 				error = ENOBUFS;
1045 				break;
1046 			}
1047 			m->m_len = sopt->sopt_valsize;
1048 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1049 					    m->m_len);
1050 			if (error) {
1051 				m_free(m);
1052 				break;
1053 			}
1054 			INP_WLOCK(inp);
1055 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1056 			INP_WUNLOCK(inp);
1057 			return (error);
1058 		}
1059 
1060 		case IP_BINDANY:
1061 			if (sopt->sopt_td != NULL) {
1062 				error = priv_check(sopt->sopt_td,
1063 				    PRIV_NETINET_BINDANY);
1064 				if (error)
1065 					break;
1066 			}
1067 			/* FALLTHROUGH */
1068 		case IP_BINDMULTI:
1069 #ifdef	RSS
1070 		case IP_RSS_LISTEN_BUCKET:
1071 #endif
1072 		case IP_TOS:
1073 		case IP_TTL:
1074 		case IP_MINTTL:
1075 		case IP_RECVOPTS:
1076 		case IP_RECVRETOPTS:
1077 		case IP_ORIGDSTADDR:
1078 		case IP_RECVDSTADDR:
1079 		case IP_RECVTTL:
1080 		case IP_RECVIF:
1081 		case IP_ONESBCAST:
1082 		case IP_DONTFRAG:
1083 		case IP_RECVTOS:
1084 		case IP_RECVFLOWID:
1085 #ifdef	RSS
1086 		case IP_RECVRSSBUCKETID:
1087 #endif
1088 			error = sooptcopyin(sopt, &optval, sizeof optval,
1089 					    sizeof optval);
1090 			if (error)
1091 				break;
1092 
1093 			switch (sopt->sopt_name) {
1094 			case IP_TOS:
1095 				inp->inp_ip_tos = optval;
1096 				break;
1097 
1098 			case IP_TTL:
1099 				inp->inp_ip_ttl = optval;
1100 				break;
1101 
1102 			case IP_MINTTL:
1103 				if (optval >= 0 && optval <= MAXTTL)
1104 					inp->inp_ip_minttl = optval;
1105 				else
1106 					error = EINVAL;
1107 				break;
1108 
1109 #define	OPTSET(bit) do {						\
1110 	INP_WLOCK(inp);							\
1111 	if (optval)							\
1112 		inp->inp_flags |= bit;					\
1113 	else								\
1114 		inp->inp_flags &= ~bit;					\
1115 	INP_WUNLOCK(inp);						\
1116 } while (0)
1117 
1118 #define	OPTSET2(bit, val) do {						\
1119 	INP_WLOCK(inp);							\
1120 	if (val)							\
1121 		inp->inp_flags2 |= bit;					\
1122 	else								\
1123 		inp->inp_flags2 &= ~bit;				\
1124 	INP_WUNLOCK(inp);						\
1125 } while (0)
1126 
1127 			case IP_RECVOPTS:
1128 				OPTSET(INP_RECVOPTS);
1129 				break;
1130 
1131 			case IP_RECVRETOPTS:
1132 				OPTSET(INP_RECVRETOPTS);
1133 				break;
1134 
1135 			case IP_RECVDSTADDR:
1136 				OPTSET(INP_RECVDSTADDR);
1137 				break;
1138 
1139 			case IP_ORIGDSTADDR:
1140 				OPTSET2(INP_ORIGDSTADDR, optval);
1141 				break;
1142 
1143 			case IP_RECVTTL:
1144 				OPTSET(INP_RECVTTL);
1145 				break;
1146 
1147 			case IP_RECVIF:
1148 				OPTSET(INP_RECVIF);
1149 				break;
1150 
1151 			case IP_ONESBCAST:
1152 				OPTSET(INP_ONESBCAST);
1153 				break;
1154 			case IP_DONTFRAG:
1155 				OPTSET(INP_DONTFRAG);
1156 				break;
1157 			case IP_BINDANY:
1158 				OPTSET(INP_BINDANY);
1159 				break;
1160 			case IP_RECVTOS:
1161 				OPTSET(INP_RECVTOS);
1162 				break;
1163 			case IP_BINDMULTI:
1164 				OPTSET2(INP_BINDMULTI, optval);
1165 				break;
1166 			case IP_RECVFLOWID:
1167 				OPTSET2(INP_RECVFLOWID, optval);
1168 				break;
1169 #ifdef	RSS
1170 			case IP_RSS_LISTEN_BUCKET:
1171 				if ((optval >= 0) &&
1172 				    (optval < rss_getnumbuckets())) {
1173 					inp->inp_rss_listen_bucket = optval;
1174 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1175 				} else {
1176 					error = EINVAL;
1177 				}
1178 				break;
1179 			case IP_RECVRSSBUCKETID:
1180 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1181 				break;
1182 #endif
1183 			}
1184 			break;
1185 #undef OPTSET
1186 #undef OPTSET2
1187 
1188 		/*
1189 		 * Multicast socket options are processed by the in_mcast
1190 		 * module.
1191 		 */
1192 		case IP_MULTICAST_IF:
1193 		case IP_MULTICAST_VIF:
1194 		case IP_MULTICAST_TTL:
1195 		case IP_MULTICAST_LOOP:
1196 		case IP_ADD_MEMBERSHIP:
1197 		case IP_DROP_MEMBERSHIP:
1198 		case IP_ADD_SOURCE_MEMBERSHIP:
1199 		case IP_DROP_SOURCE_MEMBERSHIP:
1200 		case IP_BLOCK_SOURCE:
1201 		case IP_UNBLOCK_SOURCE:
1202 		case IP_MSFILTER:
1203 		case MCAST_JOIN_GROUP:
1204 		case MCAST_LEAVE_GROUP:
1205 		case MCAST_JOIN_SOURCE_GROUP:
1206 		case MCAST_LEAVE_SOURCE_GROUP:
1207 		case MCAST_BLOCK_SOURCE:
1208 		case MCAST_UNBLOCK_SOURCE:
1209 			error = inp_setmoptions(inp, sopt);
1210 			break;
1211 
1212 		case IP_PORTRANGE:
1213 			error = sooptcopyin(sopt, &optval, sizeof optval,
1214 					    sizeof optval);
1215 			if (error)
1216 				break;
1217 
1218 			INP_WLOCK(inp);
1219 			switch (optval) {
1220 			case IP_PORTRANGE_DEFAULT:
1221 				inp->inp_flags &= ~(INP_LOWPORT);
1222 				inp->inp_flags &= ~(INP_HIGHPORT);
1223 				break;
1224 
1225 			case IP_PORTRANGE_HIGH:
1226 				inp->inp_flags &= ~(INP_LOWPORT);
1227 				inp->inp_flags |= INP_HIGHPORT;
1228 				break;
1229 
1230 			case IP_PORTRANGE_LOW:
1231 				inp->inp_flags &= ~(INP_HIGHPORT);
1232 				inp->inp_flags |= INP_LOWPORT;
1233 				break;
1234 
1235 			default:
1236 				error = EINVAL;
1237 				break;
1238 			}
1239 			INP_WUNLOCK(inp);
1240 			break;
1241 
1242 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1243 		case IP_IPSEC_POLICY:
1244 			if (IPSEC_ENABLED(ipv4)) {
1245 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1246 				break;
1247 			}
1248 			/* FALLTHROUGH */
1249 #endif /* IPSEC */
1250 
1251 		default:
1252 			error = ENOPROTOOPT;
1253 			break;
1254 		}
1255 		break;
1256 
1257 	case SOPT_GET:
1258 		switch (sopt->sopt_name) {
1259 		case IP_OPTIONS:
1260 		case IP_RETOPTS:
1261 			INP_RLOCK(inp);
1262 			if (inp->inp_options) {
1263 				struct mbuf *options;
1264 
1265 				options = m_dup(inp->inp_options, M_NOWAIT);
1266 				INP_RUNLOCK(inp);
1267 				if (options != NULL) {
1268 					error = sooptcopyout(sopt,
1269 							     mtod(options, char *),
1270 							     options->m_len);
1271 					m_freem(options);
1272 				} else
1273 					error = ENOMEM;
1274 			} else {
1275 				INP_RUNLOCK(inp);
1276 				sopt->sopt_valsize = 0;
1277 			}
1278 			break;
1279 
1280 		case IP_TOS:
1281 		case IP_TTL:
1282 		case IP_MINTTL:
1283 		case IP_RECVOPTS:
1284 		case IP_RECVRETOPTS:
1285 		case IP_ORIGDSTADDR:
1286 		case IP_RECVDSTADDR:
1287 		case IP_RECVTTL:
1288 		case IP_RECVIF:
1289 		case IP_PORTRANGE:
1290 		case IP_ONESBCAST:
1291 		case IP_DONTFRAG:
1292 		case IP_BINDANY:
1293 		case IP_RECVTOS:
1294 		case IP_BINDMULTI:
1295 		case IP_FLOWID:
1296 		case IP_FLOWTYPE:
1297 		case IP_RECVFLOWID:
1298 #ifdef	RSS
1299 		case IP_RSSBUCKETID:
1300 		case IP_RECVRSSBUCKETID:
1301 #endif
1302 			switch (sopt->sopt_name) {
1303 
1304 			case IP_TOS:
1305 				optval = inp->inp_ip_tos;
1306 				break;
1307 
1308 			case IP_TTL:
1309 				optval = inp->inp_ip_ttl;
1310 				break;
1311 
1312 			case IP_MINTTL:
1313 				optval = inp->inp_ip_minttl;
1314 				break;
1315 
1316 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1317 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1318 
1319 			case IP_RECVOPTS:
1320 				optval = OPTBIT(INP_RECVOPTS);
1321 				break;
1322 
1323 			case IP_RECVRETOPTS:
1324 				optval = OPTBIT(INP_RECVRETOPTS);
1325 				break;
1326 
1327 			case IP_RECVDSTADDR:
1328 				optval = OPTBIT(INP_RECVDSTADDR);
1329 				break;
1330 
1331 			case IP_ORIGDSTADDR:
1332 				optval = OPTBIT2(INP_ORIGDSTADDR);
1333 				break;
1334 
1335 			case IP_RECVTTL:
1336 				optval = OPTBIT(INP_RECVTTL);
1337 				break;
1338 
1339 			case IP_RECVIF:
1340 				optval = OPTBIT(INP_RECVIF);
1341 				break;
1342 
1343 			case IP_PORTRANGE:
1344 				if (inp->inp_flags & INP_HIGHPORT)
1345 					optval = IP_PORTRANGE_HIGH;
1346 				else if (inp->inp_flags & INP_LOWPORT)
1347 					optval = IP_PORTRANGE_LOW;
1348 				else
1349 					optval = 0;
1350 				break;
1351 
1352 			case IP_ONESBCAST:
1353 				optval = OPTBIT(INP_ONESBCAST);
1354 				break;
1355 			case IP_DONTFRAG:
1356 				optval = OPTBIT(INP_DONTFRAG);
1357 				break;
1358 			case IP_BINDANY:
1359 				optval = OPTBIT(INP_BINDANY);
1360 				break;
1361 			case IP_RECVTOS:
1362 				optval = OPTBIT(INP_RECVTOS);
1363 				break;
1364 			case IP_FLOWID:
1365 				optval = inp->inp_flowid;
1366 				break;
1367 			case IP_FLOWTYPE:
1368 				optval = inp->inp_flowtype;
1369 				break;
1370 			case IP_RECVFLOWID:
1371 				optval = OPTBIT2(INP_RECVFLOWID);
1372 				break;
1373 #ifdef	RSS
1374 			case IP_RSSBUCKETID:
1375 				retval = rss_hash2bucket(inp->inp_flowid,
1376 				    inp->inp_flowtype,
1377 				    &rss_bucket);
1378 				if (retval == 0)
1379 					optval = rss_bucket;
1380 				else
1381 					error = EINVAL;
1382 				break;
1383 			case IP_RECVRSSBUCKETID:
1384 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1385 				break;
1386 #endif
1387 			case IP_BINDMULTI:
1388 				optval = OPTBIT2(INP_BINDMULTI);
1389 				break;
1390 			}
1391 			error = sooptcopyout(sopt, &optval, sizeof optval);
1392 			break;
1393 
1394 		/*
1395 		 * Multicast socket options are processed by the in_mcast
1396 		 * module.
1397 		 */
1398 		case IP_MULTICAST_IF:
1399 		case IP_MULTICAST_VIF:
1400 		case IP_MULTICAST_TTL:
1401 		case IP_MULTICAST_LOOP:
1402 		case IP_MSFILTER:
1403 			error = inp_getmoptions(inp, sopt);
1404 			break;
1405 
1406 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1407 		case IP_IPSEC_POLICY:
1408 			if (IPSEC_ENABLED(ipv4)) {
1409 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1410 				break;
1411 			}
1412 			/* FALLTHROUGH */
1413 #endif /* IPSEC */
1414 
1415 		default:
1416 			error = ENOPROTOOPT;
1417 			break;
1418 		}
1419 		break;
1420 	}
1421 	return (error);
1422 }
1423 
1424 /*
1425  * Routine called from ip_output() to loop back a copy of an IP multicast
1426  * packet to the input queue of a specified interface.  Note that this
1427  * calls the output routine of the loopback "driver", but with an interface
1428  * pointer that might NOT be a loopback interface -- evil, but easier than
1429  * replicating that code here.
1430  */
1431 static void
1432 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1433 {
1434 	struct ip *ip;
1435 	struct mbuf *copym;
1436 
1437 	/*
1438 	 * Make a deep copy of the packet because we're going to
1439 	 * modify the pack in order to generate checksums.
1440 	 */
1441 	copym = m_dup(m, M_NOWAIT);
1442 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1443 		copym = m_pullup(copym, hlen);
1444 	if (copym != NULL) {
1445 		/* If needed, compute the checksum and mark it as valid. */
1446 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1447 			in_delayed_cksum(copym);
1448 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1449 			copym->m_pkthdr.csum_flags |=
1450 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1451 			copym->m_pkthdr.csum_data = 0xffff;
1452 		}
1453 		/*
1454 		 * We don't bother to fragment if the IP length is greater
1455 		 * than the interface's MTU.  Can this possibly matter?
1456 		 */
1457 		ip = mtod(copym, struct ip *);
1458 		ip->ip_sum = 0;
1459 		ip->ip_sum = in_cksum(copym, hlen);
1460 		if_simloop(ifp, copym, AF_INET, 0);
1461 	}
1462 }
1463