xref: /freebsd/sys/netinet/ip_output.c (revision 4ce386ff25d77954b8cfa11534f632172e848244)
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  * 4. 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_ipfw.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/malloc.h>
48 #include <sys/mbuf.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/protosw.h>
52 #include <sys/sdt.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sysctl.h>
56 #include <sys/ucred.h>
57 
58 #include <net/if.h>
59 #include <net/if_var.h>
60 #include <net/if_llatbl.h>
61 #include <net/netisr.h>
62 #include <net/pfil.h>
63 #include <net/route.h>
64 #include <net/flowtable.h>
65 #ifdef RADIX_MPATH
66 #include <net/radix_mpath.h>
67 #endif
68 #include <net/rss_config.h>
69 #include <net/vnet.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_kdtrace.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/ip.h>
75 #include <netinet/in_pcb.h>
76 #include <netinet/in_rss.h>
77 #include <netinet/in_var.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_options.h>
80 #ifdef SCTP
81 #include <netinet/sctp.h>
82 #include <netinet/sctp_crc32.h>
83 #endif
84 
85 #ifdef IPSEC
86 #include <netinet/ip_ipsec.h>
87 #include <netipsec/ipsec.h>
88 #endif /* IPSEC*/
89 
90 #include <machine/in_cksum.h>
91 
92 #include <security/mac/mac_framework.h>
93 
94 VNET_DEFINE(u_short, ip_id);
95 
96 #ifdef MBUF_STRESS_TEST
97 static int mbuf_frag_size = 0;
98 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
99 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
100 #endif
101 
102 static void	ip_mloopback
103 	(struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
104 
105 
106 extern int in_mcast_loop;
107 extern	struct protosw inetsw[];
108 
109 /*
110  * IP output.  The packet in mbuf chain m contains a skeletal IP
111  * header (with len, off, ttl, proto, tos, src, dst).
112  * The mbuf chain containing the packet will be freed.
113  * The mbuf opt, if present, will not be freed.
114  * If route ro is present and has ro_rt initialized, route lookup would be
115  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
116  * then result of route lookup is stored in ro->ro_rt.
117  *
118  * In the IP forwarding case, the packet will arrive with options already
119  * inserted, so must have a NULL opt pointer.
120  */
121 int
122 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
123     struct ip_moptions *imo, struct inpcb *inp)
124 {
125 	struct ip *ip;
126 	struct ifnet *ifp = NULL;	/* keep compiler happy */
127 	struct mbuf *m0;
128 	int hlen = sizeof (struct ip);
129 	int mtu;
130 	int error = 0;
131 	struct sockaddr_in *dst;
132 	const struct sockaddr_in *gw;
133 	struct in_ifaddr *ia;
134 	int isbroadcast;
135 	uint16_t ip_len, ip_off;
136 	struct route iproute;
137 	struct rtentry *rte;	/* cache for ro->ro_rt */
138 	struct in_addr odst;
139 	struct m_tag *fwd_tag = NULL;
140 	uint32_t fibnum;
141 	int have_ia_ref;
142 	int needfiblookup;
143 #ifdef IPSEC
144 	int no_route_but_check_spd = 0;
145 #endif
146 	M_ASSERTPKTHDR(m);
147 
148 	if (inp != NULL) {
149 		INP_LOCK_ASSERT(inp);
150 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
151 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
152 			m->m_pkthdr.flowid = inp->inp_flowid;
153 			M_HASHTYPE_SET(m, inp->inp_flowtype);
154 		}
155 	}
156 
157 	if (ro == NULL) {
158 		ro = &iproute;
159 		bzero(ro, sizeof (*ro));
160 	}
161 
162 #ifdef FLOWTABLE
163 	if (ro->ro_rt == NULL)
164 		(void )flowtable_lookup(AF_INET, m, ro);
165 #endif
166 
167 	if (opt) {
168 		int len = 0;
169 		m = ip_insertoptions(m, opt, &len);
170 		if (len != 0)
171 			hlen = len; /* ip->ip_hl is updated above */
172 	}
173 	ip = mtod(m, struct ip *);
174 	ip_len = ntohs(ip->ip_len);
175 	ip_off = ntohs(ip->ip_off);
176 
177 	/*
178 	 * Fill in IP header.  If we are not allowing fragmentation,
179 	 * then the ip_id field is meaningless, but we don't set it
180 	 * to zero.  Doing so causes various problems when devices along
181 	 * the path (routers, load balancers, firewalls, etc.) illegally
182 	 * disable DF on our packet.  Note that a 16-bit counter
183 	 * will wrap around in less than 10 seconds at 100 Mbit/s on a
184 	 * medium with MTU 1500.  See Steven M. Bellovin, "A Technique
185 	 * for Counting NATted Hosts", Proc. IMW'02, available at
186 	 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
187 	 */
188 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
189 		ip->ip_v = IPVERSION;
190 		ip->ip_hl = hlen >> 2;
191 		ip->ip_id = ip_newid();
192 		IPSTAT_INC(ips_localout);
193 	} else {
194 		/* Header already set, fetch hlen from there */
195 		hlen = ip->ip_hl << 2;
196 	}
197 
198 	/*
199 	 * dst/gw handling:
200 	 *
201 	 * dst can be rewritten but always points to &ro->ro_dst.
202 	 * gw is readonly but can point either to dst OR rt_gateway,
203 	 * therefore we need restore gw if we're redoing lookup.
204 	 */
205 	gw = dst = (struct sockaddr_in *)&ro->ro_dst;
206 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
207 again:
208 	ia = NULL;
209 	have_ia_ref = 0;
210 	/*
211 	 * If there is a cached route, check that it is to the same
212 	 * destination and is still up.  If not, free it and try again.
213 	 * The address family should also be checked in case of sharing
214 	 * the cache with IPv6.
215 	 */
216 	rte = ro->ro_rt;
217 	if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
218 		    rte->rt_ifp == NULL ||
219 		    !RT_LINK_IS_UP(rte->rt_ifp) ||
220 			  dst->sin_family != AF_INET ||
221 			  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
222 		RO_RTFREE(ro);
223 		ro->ro_lle = NULL;
224 		rte = NULL;
225 		gw = dst;
226 	}
227 	if (rte == NULL && fwd_tag == NULL) {
228 		bzero(dst, sizeof(*dst));
229 		dst->sin_family = AF_INET;
230 		dst->sin_len = sizeof(*dst);
231 		dst->sin_addr = ip->ip_dst;
232 	}
233 	/*
234 	 * If routing to interface only, short circuit routing lookup.
235 	 * The use of an all-ones broadcast address implies this; an
236 	 * interface is specified by the broadcast address of an interface,
237 	 * or the destination address of a ptp interface.
238 	 */
239 	if (flags & IP_SENDONES) {
240 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
241 						      M_GETFIB(m)))) == NULL &&
242 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
243 						    M_GETFIB(m)))) == NULL) {
244 			IPSTAT_INC(ips_noroute);
245 			error = ENETUNREACH;
246 			goto bad;
247 		}
248 		have_ia_ref = 1;
249 		ip->ip_dst.s_addr = INADDR_BROADCAST;
250 		dst->sin_addr = ip->ip_dst;
251 		ifp = ia->ia_ifp;
252 		ip->ip_ttl = 1;
253 		isbroadcast = 1;
254 	} else if (flags & IP_ROUTETOIF) {
255 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
256 						    M_GETFIB(m)))) == NULL &&
257 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
258 						M_GETFIB(m)))) == NULL) {
259 			IPSTAT_INC(ips_noroute);
260 			error = ENETUNREACH;
261 			goto bad;
262 		}
263 		have_ia_ref = 1;
264 		ifp = ia->ia_ifp;
265 		ip->ip_ttl = 1;
266 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
267 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
268 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
269 		/*
270 		 * Bypass the normal routing lookup for multicast
271 		 * packets if the interface is specified.
272 		 */
273 		ifp = imo->imo_multicast_ifp;
274 		IFP_TO_IA(ifp, ia);
275 		if (ia)
276 			have_ia_ref = 1;
277 		isbroadcast = 0;	/* fool gcc */
278 	} else {
279 		/*
280 		 * We want to do any cloning requested by the link layer,
281 		 * as this is probably required in all cases for correct
282 		 * operation (as it is for ARP).
283 		 */
284 		if (rte == NULL) {
285 #ifdef RADIX_MPATH
286 			rtalloc_mpath_fib(ro,
287 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
288 			    fibnum);
289 #else
290 			in_rtalloc_ign(ro, 0, fibnum);
291 #endif
292 			rte = ro->ro_rt;
293 		}
294 		if (rte == NULL ||
295 		    rte->rt_ifp == NULL ||
296 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
297 #ifdef IPSEC
298 			/*
299 			 * There is no route for this packet, but it is
300 			 * possible that a matching SPD entry exists.
301 			 */
302 			no_route_but_check_spd = 1;
303 			mtu = 0; /* Silence GCC warning. */
304 			goto sendit;
305 #endif
306 			IPSTAT_INC(ips_noroute);
307 			error = EHOSTUNREACH;
308 			goto bad;
309 		}
310 		ia = ifatoia(rte->rt_ifa);
311 		ifp = rte->rt_ifp;
312 		counter_u64_add(rte->rt_pksent, 1);
313 		if (rte->rt_flags & RTF_GATEWAY)
314 			gw = (struct sockaddr_in *)rte->rt_gateway;
315 		if (rte->rt_flags & RTF_HOST)
316 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
317 		else
318 			isbroadcast = in_broadcast(gw->sin_addr, ifp);
319 	}
320 	/*
321 	 * Calculate MTU.  If we have a route that is up, use that,
322 	 * otherwise use the interface's MTU.
323 	 */
324 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
325 		mtu = rte->rt_mtu;
326 	else
327 		mtu = ifp->if_mtu;
328 	/* Catch a possible divide by zero later. */
329 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
330 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
331 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
332 		m->m_flags |= M_MCAST;
333 		/*
334 		 * IP destination address is multicast.  Make sure "gw"
335 		 * still points to the address in "ro".  (It may have been
336 		 * changed to point to a gateway address, above.)
337 		 */
338 		gw = dst;
339 		/*
340 		 * See if the caller provided any multicast options
341 		 */
342 		if (imo != NULL) {
343 			ip->ip_ttl = imo->imo_multicast_ttl;
344 			if (imo->imo_multicast_vif != -1)
345 				ip->ip_src.s_addr =
346 				    ip_mcast_src ?
347 				    ip_mcast_src(imo->imo_multicast_vif) :
348 				    INADDR_ANY;
349 		} else
350 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
351 		/*
352 		 * Confirm that the outgoing interface supports multicast.
353 		 */
354 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
355 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
356 				IPSTAT_INC(ips_noroute);
357 				error = ENETUNREACH;
358 				goto bad;
359 			}
360 		}
361 		/*
362 		 * If source address not specified yet, use address
363 		 * of outgoing interface.
364 		 */
365 		if (ip->ip_src.s_addr == INADDR_ANY) {
366 			/* Interface may have no addresses. */
367 			if (ia != NULL)
368 				ip->ip_src = IA_SIN(ia)->sin_addr;
369 		}
370 
371 		if ((imo == NULL && in_mcast_loop) ||
372 		    (imo && imo->imo_multicast_loop)) {
373 			/*
374 			 * Loop back multicast datagram if not expressly
375 			 * forbidden to do so, even if we are not a member
376 			 * of the group; ip_input() will filter it later,
377 			 * thus deferring a hash lookup and mutex acquisition
378 			 * at the expense of a cheap copy using m_copym().
379 			 */
380 			ip_mloopback(ifp, m, dst, hlen);
381 		} else {
382 			/*
383 			 * If we are acting as a multicast router, perform
384 			 * multicast forwarding as if the packet had just
385 			 * arrived on the interface to which we are about
386 			 * to send.  The multicast forwarding function
387 			 * recursively calls this function, using the
388 			 * IP_FORWARDING flag to prevent infinite recursion.
389 			 *
390 			 * Multicasts that are looped back by ip_mloopback(),
391 			 * above, will be forwarded by the ip_input() routine,
392 			 * if necessary.
393 			 */
394 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
395 				/*
396 				 * If rsvp daemon is not running, do not
397 				 * set ip_moptions. This ensures that the packet
398 				 * is multicast and not just sent down one link
399 				 * as prescribed by rsvpd.
400 				 */
401 				if (!V_rsvp_on)
402 					imo = NULL;
403 				if (ip_mforward &&
404 				    ip_mforward(ip, ifp, m, imo) != 0) {
405 					m_freem(m);
406 					goto done;
407 				}
408 			}
409 		}
410 
411 		/*
412 		 * Multicasts with a time-to-live of zero may be looped-
413 		 * back, above, but must not be transmitted on a network.
414 		 * Also, multicasts addressed to the loopback interface
415 		 * are not sent -- the above call to ip_mloopback() will
416 		 * loop back a copy. ip_input() will drop the copy if
417 		 * this host does not belong to the destination group on
418 		 * the loopback interface.
419 		 */
420 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
421 			m_freem(m);
422 			goto done;
423 		}
424 
425 		goto sendit;
426 	}
427 
428 	/*
429 	 * If the source address is not specified yet, use the address
430 	 * of the outoing interface.
431 	 */
432 	if (ip->ip_src.s_addr == INADDR_ANY) {
433 		/* Interface may have no addresses. */
434 		if (ia != NULL) {
435 			ip->ip_src = IA_SIN(ia)->sin_addr;
436 		}
437 	}
438 
439 	/*
440 	 * Look for broadcast address and
441 	 * verify user is allowed to send
442 	 * such a packet.
443 	 */
444 	if (isbroadcast) {
445 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
446 			error = EADDRNOTAVAIL;
447 			goto bad;
448 		}
449 		if ((flags & IP_ALLOWBROADCAST) == 0) {
450 			error = EACCES;
451 			goto bad;
452 		}
453 		/* don't allow broadcast messages to be fragmented */
454 		if (ip_len > mtu) {
455 			error = EMSGSIZE;
456 			goto bad;
457 		}
458 		m->m_flags |= M_BCAST;
459 	} else {
460 		m->m_flags &= ~M_BCAST;
461 	}
462 
463 sendit:
464 #ifdef IPSEC
465 	switch(ip_ipsec_output(&m, inp, &error)) {
466 	case 1:
467 		goto bad;
468 	case -1:
469 		goto done;
470 	case 0:
471 	default:
472 		break;	/* Continue with packet processing. */
473 	}
474 	/*
475 	 * Check if there was a route for this packet; return error if not.
476 	 */
477 	if (no_route_but_check_spd) {
478 		IPSTAT_INC(ips_noroute);
479 		error = EHOSTUNREACH;
480 		goto bad;
481 	}
482 	/* Update variables that are affected by ipsec4_output(). */
483 	ip = mtod(m, struct ip *);
484 	hlen = ip->ip_hl << 2;
485 #endif /* IPSEC */
486 
487 	/* Jump over all PFIL processing if hooks are not active. */
488 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
489 		goto passout;
490 
491 	/* Run through list of hooks for output packets. */
492 	odst.s_addr = ip->ip_dst.s_addr;
493 	error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
494 	if (error != 0 || m == NULL)
495 		goto done;
496 
497 	ip = mtod(m, struct ip *);
498 	needfiblookup = 0;
499 
500 	/* See if destination IP address was changed by packet filter. */
501 	if (odst.s_addr != ip->ip_dst.s_addr) {
502 		m->m_flags |= M_SKIP_FIREWALL;
503 		/* If destination is now ourself drop to ip_input(). */
504 		if (in_localip(ip->ip_dst)) {
505 			m->m_flags |= M_FASTFWD_OURS;
506 			if (m->m_pkthdr.rcvif == NULL)
507 				m->m_pkthdr.rcvif = V_loif;
508 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
509 				m->m_pkthdr.csum_flags |=
510 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
511 				m->m_pkthdr.csum_data = 0xffff;
512 			}
513 			m->m_pkthdr.csum_flags |=
514 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
515 #ifdef SCTP
516 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
517 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
518 #endif
519 			error = netisr_queue(NETISR_IP, m);
520 			goto done;
521 		} else {
522 			if (have_ia_ref)
523 				ifa_free(&ia->ia_ifa);
524 			needfiblookup = 1; /* Redo the routing table lookup. */
525 		}
526 	}
527 	/* See if fib was changed by packet filter. */
528 	if (fibnum != M_GETFIB(m)) {
529 		m->m_flags |= M_SKIP_FIREWALL;
530 		fibnum = M_GETFIB(m);
531 		RO_RTFREE(ro);
532 		needfiblookup = 1;
533 	}
534 	if (needfiblookup)
535 		goto again;
536 
537 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
538 	if (m->m_flags & M_FASTFWD_OURS) {
539 		if (m->m_pkthdr.rcvif == NULL)
540 			m->m_pkthdr.rcvif = V_loif;
541 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
542 			m->m_pkthdr.csum_flags |=
543 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
544 			m->m_pkthdr.csum_data = 0xffff;
545 		}
546 #ifdef SCTP
547 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
548 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
549 #endif
550 		m->m_pkthdr.csum_flags |=
551 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
552 
553 		error = netisr_queue(NETISR_IP, m);
554 		goto done;
555 	}
556 	/* Or forward to some other address? */
557 	if ((m->m_flags & M_IP_NEXTHOP) &&
558 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
559 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
560 		m->m_flags |= M_SKIP_FIREWALL;
561 		m->m_flags &= ~M_IP_NEXTHOP;
562 		m_tag_delete(m, fwd_tag);
563 		if (have_ia_ref)
564 			ifa_free(&ia->ia_ifa);
565 		goto again;
566 	}
567 
568 passout:
569 	/* 127/8 must not appear on wire - RFC1122. */
570 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
571 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
572 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
573 			IPSTAT_INC(ips_badaddr);
574 			error = EADDRNOTAVAIL;
575 			goto bad;
576 		}
577 	}
578 
579 	m->m_pkthdr.csum_flags |= CSUM_IP;
580 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
581 		in_delayed_cksum(m);
582 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
583 	}
584 #ifdef SCTP
585 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
586 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
587 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
588 	}
589 #endif
590 
591 	/*
592 	 * If small enough for interface, or the interface will take
593 	 * care of the fragmentation for us, we can just send directly.
594 	 */
595 	if (ip_len <= mtu ||
596 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
597 		ip->ip_sum = 0;
598 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
599 			ip->ip_sum = in_cksum(m, hlen);
600 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
601 		}
602 
603 		/*
604 		 * Record statistics for this interface address.
605 		 * With CSUM_TSO the byte/packet count will be slightly
606 		 * incorrect because we count the IP+TCP headers only
607 		 * once instead of for every generated packet.
608 		 */
609 		if (!(flags & IP_FORWARDING) && ia) {
610 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
611 				counter_u64_add(ia->ia_ifa.ifa_opackets,
612 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
613 			else
614 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
615 
616 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
617 		}
618 #ifdef MBUF_STRESS_TEST
619 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
620 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
621 #endif
622 		/*
623 		 * Reset layer specific mbuf flags
624 		 * to avoid confusing lower layers.
625 		 */
626 		m_clrprotoflags(m);
627 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
628 		error = (*ifp->if_output)(ifp, m,
629 		    (const struct sockaddr *)gw, ro);
630 		goto done;
631 	}
632 
633 	/* Balk when DF bit is set or the interface didn't support TSO. */
634 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
635 		error = EMSGSIZE;
636 		IPSTAT_INC(ips_cantfrag);
637 		goto bad;
638 	}
639 
640 	/*
641 	 * Too large for interface; fragment if possible. If successful,
642 	 * on return, m will point to a list of packets to be sent.
643 	 */
644 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
645 	if (error)
646 		goto bad;
647 	for (; m; m = m0) {
648 		m0 = m->m_nextpkt;
649 		m->m_nextpkt = 0;
650 		if (error == 0) {
651 			/* Record statistics for this interface address. */
652 			if (ia != NULL) {
653 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
654 				counter_u64_add(ia->ia_ifa.ifa_obytes,
655 				    m->m_pkthdr.len);
656 			}
657 			/*
658 			 * Reset layer specific mbuf flags
659 			 * to avoid confusing upper layers.
660 			 */
661 			m_clrprotoflags(m);
662 
663 			IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
664 			error = (*ifp->if_output)(ifp, m,
665 			    (const struct sockaddr *)gw, ro);
666 		} else
667 			m_freem(m);
668 	}
669 
670 	if (error == 0)
671 		IPSTAT_INC(ips_fragmented);
672 
673 done:
674 	if (ro == &iproute)
675 		RO_RTFREE(ro);
676 	if (have_ia_ref)
677 		ifa_free(&ia->ia_ifa);
678 	return (error);
679 bad:
680 	m_freem(m);
681 	goto done;
682 }
683 
684 /*
685  * Create a chain of fragments which fit the given mtu. m_frag points to the
686  * mbuf to be fragmented; on return it points to the chain with the fragments.
687  * Return 0 if no error. If error, m_frag may contain a partially built
688  * chain of fragments that should be freed by the caller.
689  *
690  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
691  */
692 int
693 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
694     u_long if_hwassist_flags)
695 {
696 	int error = 0;
697 	int hlen = ip->ip_hl << 2;
698 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
699 	int off;
700 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
701 	int firstlen;
702 	struct mbuf **mnext;
703 	int nfrags;
704 	uint16_t ip_len, ip_off;
705 
706 	ip_len = ntohs(ip->ip_len);
707 	ip_off = ntohs(ip->ip_off);
708 
709 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
710 		IPSTAT_INC(ips_cantfrag);
711 		return EMSGSIZE;
712 	}
713 
714 	/*
715 	 * Must be able to put at least 8 bytes per fragment.
716 	 */
717 	if (len < 8)
718 		return EMSGSIZE;
719 
720 	/*
721 	 * If the interface will not calculate checksums on
722 	 * fragmented packets, then do it here.
723 	 */
724 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
725 		in_delayed_cksum(m0);
726 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
727 	}
728 #ifdef SCTP
729 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
730 		sctp_delayed_cksum(m0, hlen);
731 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
732 	}
733 #endif
734 	if (len > PAGE_SIZE) {
735 		/*
736 		 * Fragment large datagrams such that each segment
737 		 * contains a multiple of PAGE_SIZE amount of data,
738 		 * plus headers. This enables a receiver to perform
739 		 * page-flipping zero-copy optimizations.
740 		 *
741 		 * XXX When does this help given that sender and receiver
742 		 * could have different page sizes, and also mtu could
743 		 * be less than the receiver's page size ?
744 		 */
745 		int newlen;
746 
747 		off = MIN(mtu, m0->m_pkthdr.len);
748 
749 		/*
750 		 * firstlen (off - hlen) must be aligned on an
751 		 * 8-byte boundary
752 		 */
753 		if (off < hlen)
754 			goto smart_frag_failure;
755 		off = ((off - hlen) & ~7) + hlen;
756 		newlen = (~PAGE_MASK) & mtu;
757 		if ((newlen + sizeof (struct ip)) > mtu) {
758 			/* we failed, go back the default */
759 smart_frag_failure:
760 			newlen = len;
761 			off = hlen + len;
762 		}
763 		len = newlen;
764 
765 	} else {
766 		off = hlen + len;
767 	}
768 
769 	firstlen = off - hlen;
770 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
771 
772 	/*
773 	 * Loop through length of segment after first fragment,
774 	 * make new header and copy data of each part and link onto chain.
775 	 * Here, m0 is the original packet, m is the fragment being created.
776 	 * The fragments are linked off the m_nextpkt of the original
777 	 * packet, which after processing serves as the first fragment.
778 	 */
779 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
780 		struct ip *mhip;	/* ip header on the fragment */
781 		struct mbuf *m;
782 		int mhlen = sizeof (struct ip);
783 
784 		m = m_gethdr(M_NOWAIT, MT_DATA);
785 		if (m == NULL) {
786 			error = ENOBUFS;
787 			IPSTAT_INC(ips_odropped);
788 			goto done;
789 		}
790 		/* make sure the flowid is the same for the fragmented mbufs */
791 		M_HASHTYPE_SET(m, M_HASHTYPE_GET(m0));
792 		m->m_pkthdr.flowid = m0->m_pkthdr.flowid;
793 		/* copy multicast flag, if any */
794 		m->m_flags |= (m0->m_flags & M_MCAST);
795 		/*
796 		 * In the first mbuf, leave room for the link header, then
797 		 * copy the original IP header including options. The payload
798 		 * goes into an additional mbuf chain returned by m_copym().
799 		 */
800 		m->m_data += max_linkhdr;
801 		mhip = mtod(m, struct ip *);
802 		*mhip = *ip;
803 		if (hlen > sizeof (struct ip)) {
804 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
805 			mhip->ip_v = IPVERSION;
806 			mhip->ip_hl = mhlen >> 2;
807 		}
808 		m->m_len = mhlen;
809 		/* XXX do we need to add ip_off below ? */
810 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
811 		if (off + len >= ip_len)
812 			len = ip_len - off;
813 		else
814 			mhip->ip_off |= IP_MF;
815 		mhip->ip_len = htons((u_short)(len + mhlen));
816 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
817 		if (m->m_next == NULL) {	/* copy failed */
818 			m_free(m);
819 			error = ENOBUFS;	/* ??? */
820 			IPSTAT_INC(ips_odropped);
821 			goto done;
822 		}
823 		m->m_pkthdr.len = mhlen + len;
824 		m->m_pkthdr.rcvif = NULL;
825 #ifdef MAC
826 		mac_netinet_fragment(m0, m);
827 #endif
828 		m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
829 		mhip->ip_off = htons(mhip->ip_off);
830 		mhip->ip_sum = 0;
831 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
832 			mhip->ip_sum = in_cksum(m, mhlen);
833 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
834 		}
835 		*mnext = m;
836 		mnext = &m->m_nextpkt;
837 	}
838 	IPSTAT_ADD(ips_ofragments, nfrags);
839 
840 	/*
841 	 * Update first fragment by trimming what's been copied out
842 	 * and updating header.
843 	 */
844 	m_adj(m0, hlen + firstlen - ip_len);
845 	m0->m_pkthdr.len = hlen + firstlen;
846 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
847 	ip->ip_off = htons(ip_off | IP_MF);
848 	ip->ip_sum = 0;
849 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
850 		ip->ip_sum = in_cksum(m0, hlen);
851 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
852 	}
853 
854 done:
855 	*m_frag = m0;
856 	return error;
857 }
858 
859 void
860 in_delayed_cksum(struct mbuf *m)
861 {
862 	struct ip *ip;
863 	uint16_t csum, offset, ip_len;
864 
865 	ip = mtod(m, struct ip *);
866 	offset = ip->ip_hl << 2 ;
867 	ip_len = ntohs(ip->ip_len);
868 	csum = in_cksum_skip(m, ip_len, offset);
869 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
870 		csum = 0xffff;
871 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
872 
873 	/* find the mbuf in the chain where the checksum starts*/
874 	while ((m != NULL) && (offset >= m->m_len)) {
875 		offset -= m->m_len;
876 		m = m->m_next;
877 	}
878 	KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
879 	KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
880 	*(u_short *)(m->m_data + offset) = csum;
881 }
882 
883 /*
884  * IP socket option processing.
885  */
886 int
887 ip_ctloutput(struct socket *so, struct sockopt *sopt)
888 {
889 	struct	inpcb *inp = sotoinpcb(so);
890 	int	error, optval;
891 #ifdef	RSS
892 	uint32_t rss_bucket;
893 	int retval;
894 #endif
895 
896 	error = optval = 0;
897 	if (sopt->sopt_level != IPPROTO_IP) {
898 		error = EINVAL;
899 
900 		if (sopt->sopt_level == SOL_SOCKET &&
901 		    sopt->sopt_dir == SOPT_SET) {
902 			switch (sopt->sopt_name) {
903 			case SO_REUSEADDR:
904 				INP_WLOCK(inp);
905 				if ((so->so_options & SO_REUSEADDR) != 0)
906 					inp->inp_flags2 |= INP_REUSEADDR;
907 				else
908 					inp->inp_flags2 &= ~INP_REUSEADDR;
909 				INP_WUNLOCK(inp);
910 				error = 0;
911 				break;
912 			case SO_REUSEPORT:
913 				INP_WLOCK(inp);
914 				if ((so->so_options & SO_REUSEPORT) != 0)
915 					inp->inp_flags2 |= INP_REUSEPORT;
916 				else
917 					inp->inp_flags2 &= ~INP_REUSEPORT;
918 				INP_WUNLOCK(inp);
919 				error = 0;
920 				break;
921 			case SO_SETFIB:
922 				INP_WLOCK(inp);
923 				inp->inp_inc.inc_fibnum = so->so_fibnum;
924 				INP_WUNLOCK(inp);
925 				error = 0;
926 				break;
927 			default:
928 				break;
929 			}
930 		}
931 		return (error);
932 	}
933 
934 	switch (sopt->sopt_dir) {
935 	case SOPT_SET:
936 		switch (sopt->sopt_name) {
937 		case IP_OPTIONS:
938 #ifdef notyet
939 		case IP_RETOPTS:
940 #endif
941 		{
942 			struct mbuf *m;
943 			if (sopt->sopt_valsize > MLEN) {
944 				error = EMSGSIZE;
945 				break;
946 			}
947 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
948 			if (m == NULL) {
949 				error = ENOBUFS;
950 				break;
951 			}
952 			m->m_len = sopt->sopt_valsize;
953 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
954 					    m->m_len);
955 			if (error) {
956 				m_free(m);
957 				break;
958 			}
959 			INP_WLOCK(inp);
960 			error = ip_pcbopts(inp, sopt->sopt_name, m);
961 			INP_WUNLOCK(inp);
962 			return (error);
963 		}
964 
965 		case IP_BINDANY:
966 			if (sopt->sopt_td != NULL) {
967 				error = priv_check(sopt->sopt_td,
968 				    PRIV_NETINET_BINDANY);
969 				if (error)
970 					break;
971 			}
972 			/* FALLTHROUGH */
973 		case IP_BINDMULTI:
974 #ifdef	RSS
975 		case IP_RSS_LISTEN_BUCKET:
976 #endif
977 		case IP_TOS:
978 		case IP_TTL:
979 		case IP_MINTTL:
980 		case IP_RECVOPTS:
981 		case IP_RECVRETOPTS:
982 		case IP_RECVDSTADDR:
983 		case IP_RECVTTL:
984 		case IP_RECVIF:
985 		case IP_ONESBCAST:
986 		case IP_DONTFRAG:
987 		case IP_RECVTOS:
988 		case IP_RECVFLOWID:
989 #ifdef	RSS
990 		case IP_RECVRSSBUCKETID:
991 #endif
992 			error = sooptcopyin(sopt, &optval, sizeof optval,
993 					    sizeof optval);
994 			if (error)
995 				break;
996 
997 			switch (sopt->sopt_name) {
998 			case IP_TOS:
999 				inp->inp_ip_tos = optval;
1000 				break;
1001 
1002 			case IP_TTL:
1003 				inp->inp_ip_ttl = optval;
1004 				break;
1005 
1006 			case IP_MINTTL:
1007 				if (optval >= 0 && optval <= MAXTTL)
1008 					inp->inp_ip_minttl = optval;
1009 				else
1010 					error = EINVAL;
1011 				break;
1012 
1013 #define	OPTSET(bit) do {						\
1014 	INP_WLOCK(inp);							\
1015 	if (optval)							\
1016 		inp->inp_flags |= bit;					\
1017 	else								\
1018 		inp->inp_flags &= ~bit;					\
1019 	INP_WUNLOCK(inp);						\
1020 } while (0)
1021 
1022 #define	OPTSET2(bit, val) do {						\
1023 	INP_WLOCK(inp);							\
1024 	if (val)							\
1025 		inp->inp_flags2 |= bit;					\
1026 	else								\
1027 		inp->inp_flags2 &= ~bit;				\
1028 	INP_WUNLOCK(inp);						\
1029 } while (0)
1030 
1031 			case IP_RECVOPTS:
1032 				OPTSET(INP_RECVOPTS);
1033 				break;
1034 
1035 			case IP_RECVRETOPTS:
1036 				OPTSET(INP_RECVRETOPTS);
1037 				break;
1038 
1039 			case IP_RECVDSTADDR:
1040 				OPTSET(INP_RECVDSTADDR);
1041 				break;
1042 
1043 			case IP_RECVTTL:
1044 				OPTSET(INP_RECVTTL);
1045 				break;
1046 
1047 			case IP_RECVIF:
1048 				OPTSET(INP_RECVIF);
1049 				break;
1050 
1051 			case IP_ONESBCAST:
1052 				OPTSET(INP_ONESBCAST);
1053 				break;
1054 			case IP_DONTFRAG:
1055 				OPTSET(INP_DONTFRAG);
1056 				break;
1057 			case IP_BINDANY:
1058 				OPTSET(INP_BINDANY);
1059 				break;
1060 			case IP_RECVTOS:
1061 				OPTSET(INP_RECVTOS);
1062 				break;
1063 			case IP_BINDMULTI:
1064 				OPTSET2(INP_BINDMULTI, optval);
1065 				break;
1066 			case IP_RECVFLOWID:
1067 				OPTSET2(INP_RECVFLOWID, optval);
1068 				break;
1069 #ifdef	RSS
1070 			case IP_RSS_LISTEN_BUCKET:
1071 				if ((optval >= 0) &&
1072 				    (optval < rss_getnumbuckets())) {
1073 					inp->inp_rss_listen_bucket = optval;
1074 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1075 				} else {
1076 					error = EINVAL;
1077 				}
1078 				break;
1079 			case IP_RECVRSSBUCKETID:
1080 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1081 				break;
1082 #endif
1083 			}
1084 			break;
1085 #undef OPTSET
1086 #undef OPTSET2
1087 
1088 		/*
1089 		 * Multicast socket options are processed by the in_mcast
1090 		 * module.
1091 		 */
1092 		case IP_MULTICAST_IF:
1093 		case IP_MULTICAST_VIF:
1094 		case IP_MULTICAST_TTL:
1095 		case IP_MULTICAST_LOOP:
1096 		case IP_ADD_MEMBERSHIP:
1097 		case IP_DROP_MEMBERSHIP:
1098 		case IP_ADD_SOURCE_MEMBERSHIP:
1099 		case IP_DROP_SOURCE_MEMBERSHIP:
1100 		case IP_BLOCK_SOURCE:
1101 		case IP_UNBLOCK_SOURCE:
1102 		case IP_MSFILTER:
1103 		case MCAST_JOIN_GROUP:
1104 		case MCAST_LEAVE_GROUP:
1105 		case MCAST_JOIN_SOURCE_GROUP:
1106 		case MCAST_LEAVE_SOURCE_GROUP:
1107 		case MCAST_BLOCK_SOURCE:
1108 		case MCAST_UNBLOCK_SOURCE:
1109 			error = inp_setmoptions(inp, sopt);
1110 			break;
1111 
1112 		case IP_PORTRANGE:
1113 			error = sooptcopyin(sopt, &optval, sizeof optval,
1114 					    sizeof optval);
1115 			if (error)
1116 				break;
1117 
1118 			INP_WLOCK(inp);
1119 			switch (optval) {
1120 			case IP_PORTRANGE_DEFAULT:
1121 				inp->inp_flags &= ~(INP_LOWPORT);
1122 				inp->inp_flags &= ~(INP_HIGHPORT);
1123 				break;
1124 
1125 			case IP_PORTRANGE_HIGH:
1126 				inp->inp_flags &= ~(INP_LOWPORT);
1127 				inp->inp_flags |= INP_HIGHPORT;
1128 				break;
1129 
1130 			case IP_PORTRANGE_LOW:
1131 				inp->inp_flags &= ~(INP_HIGHPORT);
1132 				inp->inp_flags |= INP_LOWPORT;
1133 				break;
1134 
1135 			default:
1136 				error = EINVAL;
1137 				break;
1138 			}
1139 			INP_WUNLOCK(inp);
1140 			break;
1141 
1142 #ifdef IPSEC
1143 		case IP_IPSEC_POLICY:
1144 		{
1145 			caddr_t req;
1146 			struct mbuf *m;
1147 
1148 			if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1149 				break;
1150 			if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1151 				break;
1152 			req = mtod(m, caddr_t);
1153 			error = ipsec_set_policy(inp, sopt->sopt_name, req,
1154 			    m->m_len, (sopt->sopt_td != NULL) ?
1155 			    sopt->sopt_td->td_ucred : NULL);
1156 			m_freem(m);
1157 			break;
1158 		}
1159 #endif /* IPSEC */
1160 
1161 		default:
1162 			error = ENOPROTOOPT;
1163 			break;
1164 		}
1165 		break;
1166 
1167 	case SOPT_GET:
1168 		switch (sopt->sopt_name) {
1169 		case IP_OPTIONS:
1170 		case IP_RETOPTS:
1171 			if (inp->inp_options)
1172 				error = sooptcopyout(sopt,
1173 						     mtod(inp->inp_options,
1174 							  char *),
1175 						     inp->inp_options->m_len);
1176 			else
1177 				sopt->sopt_valsize = 0;
1178 			break;
1179 
1180 		case IP_TOS:
1181 		case IP_TTL:
1182 		case IP_MINTTL:
1183 		case IP_RECVOPTS:
1184 		case IP_RECVRETOPTS:
1185 		case IP_RECVDSTADDR:
1186 		case IP_RECVTTL:
1187 		case IP_RECVIF:
1188 		case IP_PORTRANGE:
1189 		case IP_ONESBCAST:
1190 		case IP_DONTFRAG:
1191 		case IP_BINDANY:
1192 		case IP_RECVTOS:
1193 		case IP_BINDMULTI:
1194 		case IP_FLOWID:
1195 		case IP_FLOWTYPE:
1196 		case IP_RECVFLOWID:
1197 #ifdef	RSS
1198 		case IP_RSSBUCKETID:
1199 		case IP_RECVRSSBUCKETID:
1200 #endif
1201 			switch (sopt->sopt_name) {
1202 
1203 			case IP_TOS:
1204 				optval = inp->inp_ip_tos;
1205 				break;
1206 
1207 			case IP_TTL:
1208 				optval = inp->inp_ip_ttl;
1209 				break;
1210 
1211 			case IP_MINTTL:
1212 				optval = inp->inp_ip_minttl;
1213 				break;
1214 
1215 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1216 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1217 
1218 			case IP_RECVOPTS:
1219 				optval = OPTBIT(INP_RECVOPTS);
1220 				break;
1221 
1222 			case IP_RECVRETOPTS:
1223 				optval = OPTBIT(INP_RECVRETOPTS);
1224 				break;
1225 
1226 			case IP_RECVDSTADDR:
1227 				optval = OPTBIT(INP_RECVDSTADDR);
1228 				break;
1229 
1230 			case IP_RECVTTL:
1231 				optval = OPTBIT(INP_RECVTTL);
1232 				break;
1233 
1234 			case IP_RECVIF:
1235 				optval = OPTBIT(INP_RECVIF);
1236 				break;
1237 
1238 			case IP_PORTRANGE:
1239 				if (inp->inp_flags & INP_HIGHPORT)
1240 					optval = IP_PORTRANGE_HIGH;
1241 				else if (inp->inp_flags & INP_LOWPORT)
1242 					optval = IP_PORTRANGE_LOW;
1243 				else
1244 					optval = 0;
1245 				break;
1246 
1247 			case IP_ONESBCAST:
1248 				optval = OPTBIT(INP_ONESBCAST);
1249 				break;
1250 			case IP_DONTFRAG:
1251 				optval = OPTBIT(INP_DONTFRAG);
1252 				break;
1253 			case IP_BINDANY:
1254 				optval = OPTBIT(INP_BINDANY);
1255 				break;
1256 			case IP_RECVTOS:
1257 				optval = OPTBIT(INP_RECVTOS);
1258 				break;
1259 			case IP_FLOWID:
1260 				optval = inp->inp_flowid;
1261 				break;
1262 			case IP_FLOWTYPE:
1263 				optval = inp->inp_flowtype;
1264 				break;
1265 			case IP_RECVFLOWID:
1266 				optval = OPTBIT2(INP_RECVFLOWID);
1267 				break;
1268 #ifdef	RSS
1269 			case IP_RSSBUCKETID:
1270 				retval = rss_hash2bucket(inp->inp_flowid,
1271 				    inp->inp_flowtype,
1272 				    &rss_bucket);
1273 				if (retval == 0)
1274 					optval = rss_bucket;
1275 				else
1276 					error = EINVAL;
1277 				break;
1278 			case IP_RECVRSSBUCKETID:
1279 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1280 				break;
1281 #endif
1282 			case IP_BINDMULTI:
1283 				optval = OPTBIT2(INP_BINDMULTI);
1284 				break;
1285 			}
1286 			error = sooptcopyout(sopt, &optval, sizeof optval);
1287 			break;
1288 
1289 		/*
1290 		 * Multicast socket options are processed by the in_mcast
1291 		 * module.
1292 		 */
1293 		case IP_MULTICAST_IF:
1294 		case IP_MULTICAST_VIF:
1295 		case IP_MULTICAST_TTL:
1296 		case IP_MULTICAST_LOOP:
1297 		case IP_MSFILTER:
1298 			error = inp_getmoptions(inp, sopt);
1299 			break;
1300 
1301 #ifdef IPSEC
1302 		case IP_IPSEC_POLICY:
1303 		{
1304 			struct mbuf *m = NULL;
1305 			caddr_t req = NULL;
1306 			size_t len = 0;
1307 
1308 			if (m != 0) {
1309 				req = mtod(m, caddr_t);
1310 				len = m->m_len;
1311 			}
1312 			error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1313 			if (error == 0)
1314 				error = soopt_mcopyout(sopt, m); /* XXX */
1315 			if (error == 0)
1316 				m_freem(m);
1317 			break;
1318 		}
1319 #endif /* IPSEC */
1320 
1321 		default:
1322 			error = ENOPROTOOPT;
1323 			break;
1324 		}
1325 		break;
1326 	}
1327 	return (error);
1328 }
1329 
1330 /*
1331  * Routine called from ip_output() to loop back a copy of an IP multicast
1332  * packet to the input queue of a specified interface.  Note that this
1333  * calls the output routine of the loopback "driver", but with an interface
1334  * pointer that might NOT be a loopback interface -- evil, but easier than
1335  * replicating that code here.
1336  */
1337 static void
1338 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1339     int hlen)
1340 {
1341 	register struct ip *ip;
1342 	struct mbuf *copym;
1343 
1344 	/*
1345 	 * Make a deep copy of the packet because we're going to
1346 	 * modify the pack in order to generate checksums.
1347 	 */
1348 	copym = m_dup(m, M_NOWAIT);
1349 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1350 		copym = m_pullup(copym, hlen);
1351 	if (copym != NULL) {
1352 		/* If needed, compute the checksum and mark it as valid. */
1353 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1354 			in_delayed_cksum(copym);
1355 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1356 			copym->m_pkthdr.csum_flags |=
1357 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1358 			copym->m_pkthdr.csum_data = 0xffff;
1359 		}
1360 		/*
1361 		 * We don't bother to fragment if the IP length is greater
1362 		 * than the interface's MTU.  Can this possibly matter?
1363 		 */
1364 		ip = mtod(copym, struct ip *);
1365 		ip->ip_sum = 0;
1366 		ip->ip_sum = in_cksum(copym, hlen);
1367 #if 1 /* XXX */
1368 		if (dst->sin_family != AF_INET) {
1369 			printf("ip_mloopback: bad address family %d\n",
1370 						dst->sin_family);
1371 			dst->sin_family = AF_INET;
1372 		}
1373 #endif
1374 		if_simloop(ifp, copym, dst->sin_family, 0);
1375 	}
1376 }
1377