xref: /freebsd/sys/netinet/ip_output.c (revision 3e5645b78f476816ca3b5acc28b29bbafbb9c444)
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 #ifdef MBUF_STRESS_TEST
95 static int mbuf_frag_size = 0;
96 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
97 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
98 #endif
99 
100 static void	ip_mloopback
101 	(struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
102 
103 
104 extern int in_mcast_loop;
105 extern	struct protosw inetsw[];
106 
107 /*
108  * IP output.  The packet in mbuf chain m contains a skeletal IP
109  * header (with len, off, ttl, proto, tos, src, dst).
110  * The mbuf chain containing the packet will be freed.
111  * The mbuf opt, if present, will not be freed.
112  * If route ro is present and has ro_rt initialized, route lookup would be
113  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
114  * then result of route lookup is stored in ro->ro_rt.
115  *
116  * In the IP forwarding case, the packet will arrive with options already
117  * inserted, so must have a NULL opt pointer.
118  */
119 int
120 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
121     struct ip_moptions *imo, struct inpcb *inp)
122 {
123 	struct ip *ip;
124 	struct ifnet *ifp = NULL;	/* keep compiler happy */
125 	struct mbuf *m0;
126 	int hlen = sizeof (struct ip);
127 	int mtu;
128 	int error = 0;
129 	struct sockaddr_in *dst;
130 	const struct sockaddr_in *gw;
131 	struct in_ifaddr *ia;
132 	int isbroadcast;
133 	uint16_t ip_len, ip_off;
134 	struct route iproute;
135 	struct rtentry *rte;	/* cache for ro->ro_rt */
136 	struct in_addr odst;
137 	struct m_tag *fwd_tag = NULL;
138 	uint32_t fibnum;
139 	int have_ia_ref;
140 	int needfiblookup;
141 #ifdef IPSEC
142 	int no_route_but_check_spd = 0;
143 #endif
144 	M_ASSERTPKTHDR(m);
145 
146 	if (inp != NULL) {
147 		INP_LOCK_ASSERT(inp);
148 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
149 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
150 			m->m_pkthdr.flowid = inp->inp_flowid;
151 			M_HASHTYPE_SET(m, inp->inp_flowtype);
152 		}
153 	}
154 
155 	if (ro == NULL) {
156 		ro = &iproute;
157 		bzero(ro, sizeof (*ro));
158 	}
159 
160 #ifdef FLOWTABLE
161 	if (ro->ro_rt == NULL)
162 		(void )flowtable_lookup(AF_INET, m, ro);
163 #endif
164 
165 	if (opt) {
166 		int len = 0;
167 		m = ip_insertoptions(m, opt, &len);
168 		if (len != 0)
169 			hlen = len; /* ip->ip_hl is updated above */
170 	}
171 	ip = mtod(m, struct ip *);
172 	ip_len = ntohs(ip->ip_len);
173 	ip_off = ntohs(ip->ip_off);
174 
175 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
176 		ip->ip_v = IPVERSION;
177 		ip->ip_hl = hlen >> 2;
178 		ip_fillid(ip);
179 		IPSTAT_INC(ips_localout);
180 	} else {
181 		/* Header already set, fetch hlen from there */
182 		hlen = ip->ip_hl << 2;
183 	}
184 
185 	/*
186 	 * dst/gw handling:
187 	 *
188 	 * dst can be rewritten but always points to &ro->ro_dst.
189 	 * gw is readonly but can point either to dst OR rt_gateway,
190 	 * therefore we need restore gw if we're redoing lookup.
191 	 */
192 	gw = dst = (struct sockaddr_in *)&ro->ro_dst;
193 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
194 again:
195 	ia = NULL;
196 	have_ia_ref = 0;
197 	/*
198 	 * If there is a cached route, check that it is to the same
199 	 * destination and is still up.  If not, free it and try again.
200 	 * The address family should also be checked in case of sharing
201 	 * the cache with IPv6.
202 	 */
203 	rte = ro->ro_rt;
204 	if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
205 		    rte->rt_ifp == NULL ||
206 		    !RT_LINK_IS_UP(rte->rt_ifp) ||
207 			  dst->sin_family != AF_INET ||
208 			  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
209 		RO_RTFREE(ro);
210 		ro->ro_lle = NULL;
211 		rte = NULL;
212 		gw = dst;
213 	}
214 	if (rte == NULL && fwd_tag == NULL) {
215 		bzero(dst, sizeof(*dst));
216 		dst->sin_family = AF_INET;
217 		dst->sin_len = sizeof(*dst);
218 		dst->sin_addr = ip->ip_dst;
219 	}
220 	/*
221 	 * If routing to interface only, short circuit routing lookup.
222 	 * The use of an all-ones broadcast address implies this; an
223 	 * interface is specified by the broadcast address of an interface,
224 	 * or the destination address of a ptp interface.
225 	 */
226 	if (flags & IP_SENDONES) {
227 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
228 						      M_GETFIB(m)))) == NULL &&
229 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
230 						    M_GETFIB(m)))) == NULL) {
231 			IPSTAT_INC(ips_noroute);
232 			error = ENETUNREACH;
233 			goto bad;
234 		}
235 		have_ia_ref = 1;
236 		ip->ip_dst.s_addr = INADDR_BROADCAST;
237 		dst->sin_addr = ip->ip_dst;
238 		ifp = ia->ia_ifp;
239 		ip->ip_ttl = 1;
240 		isbroadcast = 1;
241 	} else if (flags & IP_ROUTETOIF) {
242 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
243 						    M_GETFIB(m)))) == NULL &&
244 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
245 						M_GETFIB(m)))) == NULL) {
246 			IPSTAT_INC(ips_noroute);
247 			error = ENETUNREACH;
248 			goto bad;
249 		}
250 		have_ia_ref = 1;
251 		ifp = ia->ia_ifp;
252 		ip->ip_ttl = 1;
253 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
254 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
255 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
256 		/*
257 		 * Bypass the normal routing lookup for multicast
258 		 * packets if the interface is specified.
259 		 */
260 		ifp = imo->imo_multicast_ifp;
261 		IFP_TO_IA(ifp, ia);
262 		if (ia)
263 			have_ia_ref = 1;
264 		isbroadcast = 0;	/* fool gcc */
265 	} else {
266 		/*
267 		 * We want to do any cloning requested by the link layer,
268 		 * as this is probably required in all cases for correct
269 		 * operation (as it is for ARP).
270 		 */
271 		if (rte == NULL) {
272 #ifdef RADIX_MPATH
273 			rtalloc_mpath_fib(ro,
274 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
275 			    fibnum);
276 #else
277 			in_rtalloc_ign(ro, 0, fibnum);
278 #endif
279 			rte = ro->ro_rt;
280 		}
281 		if (rte == NULL ||
282 		    rte->rt_ifp == NULL ||
283 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
284 #ifdef IPSEC
285 			/*
286 			 * There is no route for this packet, but it is
287 			 * possible that a matching SPD entry exists.
288 			 */
289 			no_route_but_check_spd = 1;
290 			mtu = 0; /* Silence GCC warning. */
291 			goto sendit;
292 #endif
293 			IPSTAT_INC(ips_noroute);
294 			error = EHOSTUNREACH;
295 			goto bad;
296 		}
297 		ia = ifatoia(rte->rt_ifa);
298 		ifp = rte->rt_ifp;
299 		counter_u64_add(rte->rt_pksent, 1);
300 		if (rte->rt_flags & RTF_GATEWAY)
301 			gw = (struct sockaddr_in *)rte->rt_gateway;
302 		if (rte->rt_flags & RTF_HOST)
303 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
304 		else
305 			isbroadcast = in_broadcast(gw->sin_addr, ifp);
306 	}
307 	/*
308 	 * Calculate MTU.  If we have a route that is up, use that,
309 	 * otherwise use the interface's MTU.
310 	 */
311 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
312 		mtu = rte->rt_mtu;
313 	else
314 		mtu = ifp->if_mtu;
315 	/* Catch a possible divide by zero later. */
316 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
317 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
318 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
319 		m->m_flags |= M_MCAST;
320 		/*
321 		 * IP destination address is multicast.  Make sure "gw"
322 		 * still points to the address in "ro".  (It may have been
323 		 * changed to point to a gateway address, above.)
324 		 */
325 		gw = dst;
326 		/*
327 		 * See if the caller provided any multicast options
328 		 */
329 		if (imo != NULL) {
330 			ip->ip_ttl = imo->imo_multicast_ttl;
331 			if (imo->imo_multicast_vif != -1)
332 				ip->ip_src.s_addr =
333 				    ip_mcast_src ?
334 				    ip_mcast_src(imo->imo_multicast_vif) :
335 				    INADDR_ANY;
336 		} else
337 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
338 		/*
339 		 * Confirm that the outgoing interface supports multicast.
340 		 */
341 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
342 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
343 				IPSTAT_INC(ips_noroute);
344 				error = ENETUNREACH;
345 				goto bad;
346 			}
347 		}
348 		/*
349 		 * If source address not specified yet, use address
350 		 * of outgoing interface.
351 		 */
352 		if (ip->ip_src.s_addr == INADDR_ANY) {
353 			/* Interface may have no addresses. */
354 			if (ia != NULL)
355 				ip->ip_src = IA_SIN(ia)->sin_addr;
356 		}
357 
358 		if ((imo == NULL && in_mcast_loop) ||
359 		    (imo && imo->imo_multicast_loop)) {
360 			/*
361 			 * Loop back multicast datagram if not expressly
362 			 * forbidden to do so, even if we are not a member
363 			 * of the group; ip_input() will filter it later,
364 			 * thus deferring a hash lookup and mutex acquisition
365 			 * at the expense of a cheap copy using m_copym().
366 			 */
367 			ip_mloopback(ifp, m, dst, hlen);
368 		} else {
369 			/*
370 			 * If we are acting as a multicast router, perform
371 			 * multicast forwarding as if the packet had just
372 			 * arrived on the interface to which we are about
373 			 * to send.  The multicast forwarding function
374 			 * recursively calls this function, using the
375 			 * IP_FORWARDING flag to prevent infinite recursion.
376 			 *
377 			 * Multicasts that are looped back by ip_mloopback(),
378 			 * above, will be forwarded by the ip_input() routine,
379 			 * if necessary.
380 			 */
381 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
382 				/*
383 				 * If rsvp daemon is not running, do not
384 				 * set ip_moptions. This ensures that the packet
385 				 * is multicast and not just sent down one link
386 				 * as prescribed by rsvpd.
387 				 */
388 				if (!V_rsvp_on)
389 					imo = NULL;
390 				if (ip_mforward &&
391 				    ip_mforward(ip, ifp, m, imo) != 0) {
392 					m_freem(m);
393 					goto done;
394 				}
395 			}
396 		}
397 
398 		/*
399 		 * Multicasts with a time-to-live of zero may be looped-
400 		 * back, above, but must not be transmitted on a network.
401 		 * Also, multicasts addressed to the loopback interface
402 		 * are not sent -- the above call to ip_mloopback() will
403 		 * loop back a copy. ip_input() will drop the copy if
404 		 * this host does not belong to the destination group on
405 		 * the loopback interface.
406 		 */
407 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
408 			m_freem(m);
409 			goto done;
410 		}
411 
412 		goto sendit;
413 	}
414 
415 	/*
416 	 * If the source address is not specified yet, use the address
417 	 * of the outoing interface.
418 	 */
419 	if (ip->ip_src.s_addr == INADDR_ANY) {
420 		/* Interface may have no addresses. */
421 		if (ia != NULL) {
422 			ip->ip_src = IA_SIN(ia)->sin_addr;
423 		}
424 	}
425 
426 	/*
427 	 * Look for broadcast address and
428 	 * verify user is allowed to send
429 	 * such a packet.
430 	 */
431 	if (isbroadcast) {
432 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
433 			error = EADDRNOTAVAIL;
434 			goto bad;
435 		}
436 		if ((flags & IP_ALLOWBROADCAST) == 0) {
437 			error = EACCES;
438 			goto bad;
439 		}
440 		/* don't allow broadcast messages to be fragmented */
441 		if (ip_len > mtu) {
442 			error = EMSGSIZE;
443 			goto bad;
444 		}
445 		m->m_flags |= M_BCAST;
446 	} else {
447 		m->m_flags &= ~M_BCAST;
448 	}
449 
450 sendit:
451 #ifdef IPSEC
452 	switch(ip_ipsec_output(&m, inp, &error)) {
453 	case 1:
454 		goto bad;
455 	case -1:
456 		goto done;
457 	case 0:
458 	default:
459 		break;	/* Continue with packet processing. */
460 	}
461 	/*
462 	 * Check if there was a route for this packet; return error if not.
463 	 */
464 	if (no_route_but_check_spd) {
465 		IPSTAT_INC(ips_noroute);
466 		error = EHOSTUNREACH;
467 		goto bad;
468 	}
469 	/* Update variables that are affected by ipsec4_output(). */
470 	ip = mtod(m, struct ip *);
471 	hlen = ip->ip_hl << 2;
472 #endif /* IPSEC */
473 
474 	/* Jump over all PFIL processing if hooks are not active. */
475 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
476 		goto passout;
477 
478 	/* Run through list of hooks for output packets. */
479 	odst.s_addr = ip->ip_dst.s_addr;
480 	error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
481 	if (error != 0 || m == NULL)
482 		goto done;
483 
484 	ip = mtod(m, struct ip *);
485 	needfiblookup = 0;
486 
487 	/* See if destination IP address was changed by packet filter. */
488 	if (odst.s_addr != ip->ip_dst.s_addr) {
489 		m->m_flags |= M_SKIP_FIREWALL;
490 		/* If destination is now ourself drop to ip_input(). */
491 		if (in_localip(ip->ip_dst)) {
492 			m->m_flags |= M_FASTFWD_OURS;
493 			if (m->m_pkthdr.rcvif == NULL)
494 				m->m_pkthdr.rcvif = V_loif;
495 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
496 				m->m_pkthdr.csum_flags |=
497 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
498 				m->m_pkthdr.csum_data = 0xffff;
499 			}
500 			m->m_pkthdr.csum_flags |=
501 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
502 #ifdef SCTP
503 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
504 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
505 #endif
506 			error = netisr_queue(NETISR_IP, m);
507 			goto done;
508 		} else {
509 			if (have_ia_ref)
510 				ifa_free(&ia->ia_ifa);
511 			needfiblookup = 1; /* Redo the routing table lookup. */
512 		}
513 	}
514 	/* See if fib was changed by packet filter. */
515 	if (fibnum != M_GETFIB(m)) {
516 		m->m_flags |= M_SKIP_FIREWALL;
517 		fibnum = M_GETFIB(m);
518 		RO_RTFREE(ro);
519 		needfiblookup = 1;
520 	}
521 	if (needfiblookup)
522 		goto again;
523 
524 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
525 	if (m->m_flags & M_FASTFWD_OURS) {
526 		if (m->m_pkthdr.rcvif == NULL)
527 			m->m_pkthdr.rcvif = V_loif;
528 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
529 			m->m_pkthdr.csum_flags |=
530 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
531 			m->m_pkthdr.csum_data = 0xffff;
532 		}
533 #ifdef SCTP
534 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
535 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
536 #endif
537 		m->m_pkthdr.csum_flags |=
538 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
539 
540 		error = netisr_queue(NETISR_IP, m);
541 		goto done;
542 	}
543 	/* Or forward to some other address? */
544 	if ((m->m_flags & M_IP_NEXTHOP) &&
545 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
546 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
547 		m->m_flags |= M_SKIP_FIREWALL;
548 		m->m_flags &= ~M_IP_NEXTHOP;
549 		m_tag_delete(m, fwd_tag);
550 		if (have_ia_ref)
551 			ifa_free(&ia->ia_ifa);
552 		goto again;
553 	}
554 
555 passout:
556 	/* 127/8 must not appear on wire - RFC1122. */
557 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
558 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
559 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
560 			IPSTAT_INC(ips_badaddr);
561 			error = EADDRNOTAVAIL;
562 			goto bad;
563 		}
564 	}
565 
566 	m->m_pkthdr.csum_flags |= CSUM_IP;
567 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
568 		in_delayed_cksum(m);
569 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
570 	}
571 #ifdef SCTP
572 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
573 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
574 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
575 	}
576 #endif
577 
578 	/*
579 	 * If small enough for interface, or the interface will take
580 	 * care of the fragmentation for us, we can just send directly.
581 	 */
582 	if (ip_len <= mtu ||
583 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
584 		ip->ip_sum = 0;
585 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
586 			ip->ip_sum = in_cksum(m, hlen);
587 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
588 		}
589 
590 		/*
591 		 * Record statistics for this interface address.
592 		 * With CSUM_TSO the byte/packet count will be slightly
593 		 * incorrect because we count the IP+TCP headers only
594 		 * once instead of for every generated packet.
595 		 */
596 		if (!(flags & IP_FORWARDING) && ia) {
597 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
598 				counter_u64_add(ia->ia_ifa.ifa_opackets,
599 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
600 			else
601 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
602 
603 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
604 		}
605 #ifdef MBUF_STRESS_TEST
606 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
607 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
608 #endif
609 		/*
610 		 * Reset layer specific mbuf flags
611 		 * to avoid confusing lower layers.
612 		 */
613 		m_clrprotoflags(m);
614 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
615 		error = (*ifp->if_output)(ifp, m,
616 		    (const struct sockaddr *)gw, ro);
617 		goto done;
618 	}
619 
620 	/* Balk when DF bit is set or the interface didn't support TSO. */
621 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
622 		error = EMSGSIZE;
623 		IPSTAT_INC(ips_cantfrag);
624 		goto bad;
625 	}
626 
627 	/*
628 	 * Too large for interface; fragment if possible. If successful,
629 	 * on return, m will point to a list of packets to be sent.
630 	 */
631 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
632 	if (error)
633 		goto bad;
634 	for (; m; m = m0) {
635 		m0 = m->m_nextpkt;
636 		m->m_nextpkt = 0;
637 		if (error == 0) {
638 			/* Record statistics for this interface address. */
639 			if (ia != NULL) {
640 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
641 				counter_u64_add(ia->ia_ifa.ifa_obytes,
642 				    m->m_pkthdr.len);
643 			}
644 			/*
645 			 * Reset layer specific mbuf flags
646 			 * to avoid confusing upper layers.
647 			 */
648 			m_clrprotoflags(m);
649 
650 			IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
651 			error = (*ifp->if_output)(ifp, m,
652 			    (const struct sockaddr *)gw, ro);
653 		} else
654 			m_freem(m);
655 	}
656 
657 	if (error == 0)
658 		IPSTAT_INC(ips_fragmented);
659 
660 done:
661 	if (ro == &iproute)
662 		RO_RTFREE(ro);
663 	if (have_ia_ref)
664 		ifa_free(&ia->ia_ifa);
665 	return (error);
666 bad:
667 	m_freem(m);
668 	goto done;
669 }
670 
671 /*
672  * Create a chain of fragments which fit the given mtu. m_frag points to the
673  * mbuf to be fragmented; on return it points to the chain with the fragments.
674  * Return 0 if no error. If error, m_frag may contain a partially built
675  * chain of fragments that should be freed by the caller.
676  *
677  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
678  */
679 int
680 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
681     u_long if_hwassist_flags)
682 {
683 	int error = 0;
684 	int hlen = ip->ip_hl << 2;
685 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
686 	int off;
687 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
688 	int firstlen;
689 	struct mbuf **mnext;
690 	int nfrags;
691 	uint16_t ip_len, ip_off;
692 
693 	ip_len = ntohs(ip->ip_len);
694 	ip_off = ntohs(ip->ip_off);
695 
696 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
697 		IPSTAT_INC(ips_cantfrag);
698 		return EMSGSIZE;
699 	}
700 
701 	/*
702 	 * Must be able to put at least 8 bytes per fragment.
703 	 */
704 	if (len < 8)
705 		return EMSGSIZE;
706 
707 	/*
708 	 * If the interface will not calculate checksums on
709 	 * fragmented packets, then do it here.
710 	 */
711 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
712 		in_delayed_cksum(m0);
713 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
714 	}
715 #ifdef SCTP
716 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
717 		sctp_delayed_cksum(m0, hlen);
718 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
719 	}
720 #endif
721 	if (len > PAGE_SIZE) {
722 		/*
723 		 * Fragment large datagrams such that each segment
724 		 * contains a multiple of PAGE_SIZE amount of data,
725 		 * plus headers. This enables a receiver to perform
726 		 * page-flipping zero-copy optimizations.
727 		 *
728 		 * XXX When does this help given that sender and receiver
729 		 * could have different page sizes, and also mtu could
730 		 * be less than the receiver's page size ?
731 		 */
732 		int newlen;
733 
734 		off = MIN(mtu, m0->m_pkthdr.len);
735 
736 		/*
737 		 * firstlen (off - hlen) must be aligned on an
738 		 * 8-byte boundary
739 		 */
740 		if (off < hlen)
741 			goto smart_frag_failure;
742 		off = ((off - hlen) & ~7) + hlen;
743 		newlen = (~PAGE_MASK) & mtu;
744 		if ((newlen + sizeof (struct ip)) > mtu) {
745 			/* we failed, go back the default */
746 smart_frag_failure:
747 			newlen = len;
748 			off = hlen + len;
749 		}
750 		len = newlen;
751 
752 	} else {
753 		off = hlen + len;
754 	}
755 
756 	firstlen = off - hlen;
757 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
758 
759 	/*
760 	 * Loop through length of segment after first fragment,
761 	 * make new header and copy data of each part and link onto chain.
762 	 * Here, m0 is the original packet, m is the fragment being created.
763 	 * The fragments are linked off the m_nextpkt of the original
764 	 * packet, which after processing serves as the first fragment.
765 	 */
766 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
767 		struct ip *mhip;	/* ip header on the fragment */
768 		struct mbuf *m;
769 		int mhlen = sizeof (struct ip);
770 
771 		m = m_gethdr(M_NOWAIT, MT_DATA);
772 		if (m == NULL) {
773 			error = ENOBUFS;
774 			IPSTAT_INC(ips_odropped);
775 			goto done;
776 		}
777 		/*
778 		 * Make sure the complete packet header gets copied
779 		 * from the originating mbuf to the newly created
780 		 * mbuf. This also ensures that existing firewall
781 		 * classification(s), VLAN tags and so on get copied
782 		 * to the resulting fragmented packet(s):
783 		 */
784 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
785 			m_free(m);
786 			error = ENOBUFS;
787 			IPSTAT_INC(ips_odropped);
788 			goto done;
789 		}
790 		/*
791 		 * In the first mbuf, leave room for the link header, then
792 		 * copy the original IP header including options. The payload
793 		 * goes into an additional mbuf chain returned by m_copym().
794 		 */
795 		m->m_data += max_linkhdr;
796 		mhip = mtod(m, struct ip *);
797 		*mhip = *ip;
798 		if (hlen > sizeof (struct ip)) {
799 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
800 			mhip->ip_v = IPVERSION;
801 			mhip->ip_hl = mhlen >> 2;
802 		}
803 		m->m_len = mhlen;
804 		/* XXX do we need to add ip_off below ? */
805 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
806 		if (off + len >= ip_len)
807 			len = ip_len - off;
808 		else
809 			mhip->ip_off |= IP_MF;
810 		mhip->ip_len = htons((u_short)(len + mhlen));
811 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
812 		if (m->m_next == NULL) {	/* copy failed */
813 			m_free(m);
814 			error = ENOBUFS;	/* ??? */
815 			IPSTAT_INC(ips_odropped);
816 			goto done;
817 		}
818 		m->m_pkthdr.len = mhlen + len;
819 #ifdef MAC
820 		mac_netinet_fragment(m0, m);
821 #endif
822 		mhip->ip_off = htons(mhip->ip_off);
823 		mhip->ip_sum = 0;
824 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
825 			mhip->ip_sum = in_cksum(m, mhlen);
826 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
827 		}
828 		*mnext = m;
829 		mnext = &m->m_nextpkt;
830 	}
831 	IPSTAT_ADD(ips_ofragments, nfrags);
832 
833 	/*
834 	 * Update first fragment by trimming what's been copied out
835 	 * and updating header.
836 	 */
837 	m_adj(m0, hlen + firstlen - ip_len);
838 	m0->m_pkthdr.len = hlen + firstlen;
839 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
840 	ip->ip_off = htons(ip_off | IP_MF);
841 	ip->ip_sum = 0;
842 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
843 		ip->ip_sum = in_cksum(m0, hlen);
844 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
845 	}
846 
847 done:
848 	*m_frag = m0;
849 	return error;
850 }
851 
852 void
853 in_delayed_cksum(struct mbuf *m)
854 {
855 	struct ip *ip;
856 	uint16_t csum, offset, ip_len;
857 
858 	ip = mtod(m, struct ip *);
859 	offset = ip->ip_hl << 2 ;
860 	ip_len = ntohs(ip->ip_len);
861 	csum = in_cksum_skip(m, ip_len, offset);
862 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
863 		csum = 0xffff;
864 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
865 
866 	/* find the mbuf in the chain where the checksum starts*/
867 	while ((m != NULL) && (offset >= m->m_len)) {
868 		offset -= m->m_len;
869 		m = m->m_next;
870 	}
871 	KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
872 	KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
873 	*(u_short *)(m->m_data + offset) = csum;
874 }
875 
876 /*
877  * IP socket option processing.
878  */
879 int
880 ip_ctloutput(struct socket *so, struct sockopt *sopt)
881 {
882 	struct	inpcb *inp = sotoinpcb(so);
883 	int	error, optval;
884 #ifdef	RSS
885 	uint32_t rss_bucket;
886 	int retval;
887 #endif
888 
889 	error = optval = 0;
890 	if (sopt->sopt_level != IPPROTO_IP) {
891 		error = EINVAL;
892 
893 		if (sopt->sopt_level == SOL_SOCKET &&
894 		    sopt->sopt_dir == SOPT_SET) {
895 			switch (sopt->sopt_name) {
896 			case SO_REUSEADDR:
897 				INP_WLOCK(inp);
898 				if ((so->so_options & SO_REUSEADDR) != 0)
899 					inp->inp_flags2 |= INP_REUSEADDR;
900 				else
901 					inp->inp_flags2 &= ~INP_REUSEADDR;
902 				INP_WUNLOCK(inp);
903 				error = 0;
904 				break;
905 			case SO_REUSEPORT:
906 				INP_WLOCK(inp);
907 				if ((so->so_options & SO_REUSEPORT) != 0)
908 					inp->inp_flags2 |= INP_REUSEPORT;
909 				else
910 					inp->inp_flags2 &= ~INP_REUSEPORT;
911 				INP_WUNLOCK(inp);
912 				error = 0;
913 				break;
914 			case SO_SETFIB:
915 				INP_WLOCK(inp);
916 				inp->inp_inc.inc_fibnum = so->so_fibnum;
917 				INP_WUNLOCK(inp);
918 				error = 0;
919 				break;
920 			default:
921 				break;
922 			}
923 		}
924 		return (error);
925 	}
926 
927 	switch (sopt->sopt_dir) {
928 	case SOPT_SET:
929 		switch (sopt->sopt_name) {
930 		case IP_OPTIONS:
931 #ifdef notyet
932 		case IP_RETOPTS:
933 #endif
934 		{
935 			struct mbuf *m;
936 			if (sopt->sopt_valsize > MLEN) {
937 				error = EMSGSIZE;
938 				break;
939 			}
940 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
941 			if (m == NULL) {
942 				error = ENOBUFS;
943 				break;
944 			}
945 			m->m_len = sopt->sopt_valsize;
946 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
947 					    m->m_len);
948 			if (error) {
949 				m_free(m);
950 				break;
951 			}
952 			INP_WLOCK(inp);
953 			error = ip_pcbopts(inp, sopt->sopt_name, m);
954 			INP_WUNLOCK(inp);
955 			return (error);
956 		}
957 
958 		case IP_BINDANY:
959 			if (sopt->sopt_td != NULL) {
960 				error = priv_check(sopt->sopt_td,
961 				    PRIV_NETINET_BINDANY);
962 				if (error)
963 					break;
964 			}
965 			/* FALLTHROUGH */
966 		case IP_BINDMULTI:
967 #ifdef	RSS
968 		case IP_RSS_LISTEN_BUCKET:
969 #endif
970 		case IP_TOS:
971 		case IP_TTL:
972 		case IP_MINTTL:
973 		case IP_RECVOPTS:
974 		case IP_RECVRETOPTS:
975 		case IP_RECVDSTADDR:
976 		case IP_RECVTTL:
977 		case IP_RECVIF:
978 		case IP_ONESBCAST:
979 		case IP_DONTFRAG:
980 		case IP_RECVTOS:
981 		case IP_RECVFLOWID:
982 #ifdef	RSS
983 		case IP_RECVRSSBUCKETID:
984 #endif
985 			error = sooptcopyin(sopt, &optval, sizeof optval,
986 					    sizeof optval);
987 			if (error)
988 				break;
989 
990 			switch (sopt->sopt_name) {
991 			case IP_TOS:
992 				inp->inp_ip_tos = optval;
993 				break;
994 
995 			case IP_TTL:
996 				inp->inp_ip_ttl = optval;
997 				break;
998 
999 			case IP_MINTTL:
1000 				if (optval >= 0 && optval <= MAXTTL)
1001 					inp->inp_ip_minttl = optval;
1002 				else
1003 					error = EINVAL;
1004 				break;
1005 
1006 #define	OPTSET(bit) do {						\
1007 	INP_WLOCK(inp);							\
1008 	if (optval)							\
1009 		inp->inp_flags |= bit;					\
1010 	else								\
1011 		inp->inp_flags &= ~bit;					\
1012 	INP_WUNLOCK(inp);						\
1013 } while (0)
1014 
1015 #define	OPTSET2(bit, val) do {						\
1016 	INP_WLOCK(inp);							\
1017 	if (val)							\
1018 		inp->inp_flags2 |= bit;					\
1019 	else								\
1020 		inp->inp_flags2 &= ~bit;				\
1021 	INP_WUNLOCK(inp);						\
1022 } while (0)
1023 
1024 			case IP_RECVOPTS:
1025 				OPTSET(INP_RECVOPTS);
1026 				break;
1027 
1028 			case IP_RECVRETOPTS:
1029 				OPTSET(INP_RECVRETOPTS);
1030 				break;
1031 
1032 			case IP_RECVDSTADDR:
1033 				OPTSET(INP_RECVDSTADDR);
1034 				break;
1035 
1036 			case IP_RECVTTL:
1037 				OPTSET(INP_RECVTTL);
1038 				break;
1039 
1040 			case IP_RECVIF:
1041 				OPTSET(INP_RECVIF);
1042 				break;
1043 
1044 			case IP_ONESBCAST:
1045 				OPTSET(INP_ONESBCAST);
1046 				break;
1047 			case IP_DONTFRAG:
1048 				OPTSET(INP_DONTFRAG);
1049 				break;
1050 			case IP_BINDANY:
1051 				OPTSET(INP_BINDANY);
1052 				break;
1053 			case IP_RECVTOS:
1054 				OPTSET(INP_RECVTOS);
1055 				break;
1056 			case IP_BINDMULTI:
1057 				OPTSET2(INP_BINDMULTI, optval);
1058 				break;
1059 			case IP_RECVFLOWID:
1060 				OPTSET2(INP_RECVFLOWID, optval);
1061 				break;
1062 #ifdef	RSS
1063 			case IP_RSS_LISTEN_BUCKET:
1064 				if ((optval >= 0) &&
1065 				    (optval < rss_getnumbuckets())) {
1066 					inp->inp_rss_listen_bucket = optval;
1067 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1068 				} else {
1069 					error = EINVAL;
1070 				}
1071 				break;
1072 			case IP_RECVRSSBUCKETID:
1073 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1074 				break;
1075 #endif
1076 			}
1077 			break;
1078 #undef OPTSET
1079 #undef OPTSET2
1080 
1081 		/*
1082 		 * Multicast socket options are processed by the in_mcast
1083 		 * module.
1084 		 */
1085 		case IP_MULTICAST_IF:
1086 		case IP_MULTICAST_VIF:
1087 		case IP_MULTICAST_TTL:
1088 		case IP_MULTICAST_LOOP:
1089 		case IP_ADD_MEMBERSHIP:
1090 		case IP_DROP_MEMBERSHIP:
1091 		case IP_ADD_SOURCE_MEMBERSHIP:
1092 		case IP_DROP_SOURCE_MEMBERSHIP:
1093 		case IP_BLOCK_SOURCE:
1094 		case IP_UNBLOCK_SOURCE:
1095 		case IP_MSFILTER:
1096 		case MCAST_JOIN_GROUP:
1097 		case MCAST_LEAVE_GROUP:
1098 		case MCAST_JOIN_SOURCE_GROUP:
1099 		case MCAST_LEAVE_SOURCE_GROUP:
1100 		case MCAST_BLOCK_SOURCE:
1101 		case MCAST_UNBLOCK_SOURCE:
1102 			error = inp_setmoptions(inp, sopt);
1103 			break;
1104 
1105 		case IP_PORTRANGE:
1106 			error = sooptcopyin(sopt, &optval, sizeof optval,
1107 					    sizeof optval);
1108 			if (error)
1109 				break;
1110 
1111 			INP_WLOCK(inp);
1112 			switch (optval) {
1113 			case IP_PORTRANGE_DEFAULT:
1114 				inp->inp_flags &= ~(INP_LOWPORT);
1115 				inp->inp_flags &= ~(INP_HIGHPORT);
1116 				break;
1117 
1118 			case IP_PORTRANGE_HIGH:
1119 				inp->inp_flags &= ~(INP_LOWPORT);
1120 				inp->inp_flags |= INP_HIGHPORT;
1121 				break;
1122 
1123 			case IP_PORTRANGE_LOW:
1124 				inp->inp_flags &= ~(INP_HIGHPORT);
1125 				inp->inp_flags |= INP_LOWPORT;
1126 				break;
1127 
1128 			default:
1129 				error = EINVAL;
1130 				break;
1131 			}
1132 			INP_WUNLOCK(inp);
1133 			break;
1134 
1135 #ifdef IPSEC
1136 		case IP_IPSEC_POLICY:
1137 		{
1138 			caddr_t req;
1139 			struct mbuf *m;
1140 
1141 			if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1142 				break;
1143 			if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1144 				break;
1145 			req = mtod(m, caddr_t);
1146 			error = ipsec_set_policy(inp, sopt->sopt_name, req,
1147 			    m->m_len, (sopt->sopt_td != NULL) ?
1148 			    sopt->sopt_td->td_ucred : NULL);
1149 			m_freem(m);
1150 			break;
1151 		}
1152 #endif /* IPSEC */
1153 
1154 		default:
1155 			error = ENOPROTOOPT;
1156 			break;
1157 		}
1158 		break;
1159 
1160 	case SOPT_GET:
1161 		switch (sopt->sopt_name) {
1162 		case IP_OPTIONS:
1163 		case IP_RETOPTS:
1164 			if (inp->inp_options)
1165 				error = sooptcopyout(sopt,
1166 						     mtod(inp->inp_options,
1167 							  char *),
1168 						     inp->inp_options->m_len);
1169 			else
1170 				sopt->sopt_valsize = 0;
1171 			break;
1172 
1173 		case IP_TOS:
1174 		case IP_TTL:
1175 		case IP_MINTTL:
1176 		case IP_RECVOPTS:
1177 		case IP_RECVRETOPTS:
1178 		case IP_RECVDSTADDR:
1179 		case IP_RECVTTL:
1180 		case IP_RECVIF:
1181 		case IP_PORTRANGE:
1182 		case IP_ONESBCAST:
1183 		case IP_DONTFRAG:
1184 		case IP_BINDANY:
1185 		case IP_RECVTOS:
1186 		case IP_BINDMULTI:
1187 		case IP_FLOWID:
1188 		case IP_FLOWTYPE:
1189 		case IP_RECVFLOWID:
1190 #ifdef	RSS
1191 		case IP_RSSBUCKETID:
1192 		case IP_RECVRSSBUCKETID:
1193 #endif
1194 			switch (sopt->sopt_name) {
1195 
1196 			case IP_TOS:
1197 				optval = inp->inp_ip_tos;
1198 				break;
1199 
1200 			case IP_TTL:
1201 				optval = inp->inp_ip_ttl;
1202 				break;
1203 
1204 			case IP_MINTTL:
1205 				optval = inp->inp_ip_minttl;
1206 				break;
1207 
1208 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1209 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1210 
1211 			case IP_RECVOPTS:
1212 				optval = OPTBIT(INP_RECVOPTS);
1213 				break;
1214 
1215 			case IP_RECVRETOPTS:
1216 				optval = OPTBIT(INP_RECVRETOPTS);
1217 				break;
1218 
1219 			case IP_RECVDSTADDR:
1220 				optval = OPTBIT(INP_RECVDSTADDR);
1221 				break;
1222 
1223 			case IP_RECVTTL:
1224 				optval = OPTBIT(INP_RECVTTL);
1225 				break;
1226 
1227 			case IP_RECVIF:
1228 				optval = OPTBIT(INP_RECVIF);
1229 				break;
1230 
1231 			case IP_PORTRANGE:
1232 				if (inp->inp_flags & INP_HIGHPORT)
1233 					optval = IP_PORTRANGE_HIGH;
1234 				else if (inp->inp_flags & INP_LOWPORT)
1235 					optval = IP_PORTRANGE_LOW;
1236 				else
1237 					optval = 0;
1238 				break;
1239 
1240 			case IP_ONESBCAST:
1241 				optval = OPTBIT(INP_ONESBCAST);
1242 				break;
1243 			case IP_DONTFRAG:
1244 				optval = OPTBIT(INP_DONTFRAG);
1245 				break;
1246 			case IP_BINDANY:
1247 				optval = OPTBIT(INP_BINDANY);
1248 				break;
1249 			case IP_RECVTOS:
1250 				optval = OPTBIT(INP_RECVTOS);
1251 				break;
1252 			case IP_FLOWID:
1253 				optval = inp->inp_flowid;
1254 				break;
1255 			case IP_FLOWTYPE:
1256 				optval = inp->inp_flowtype;
1257 				break;
1258 			case IP_RECVFLOWID:
1259 				optval = OPTBIT2(INP_RECVFLOWID);
1260 				break;
1261 #ifdef	RSS
1262 			case IP_RSSBUCKETID:
1263 				retval = rss_hash2bucket(inp->inp_flowid,
1264 				    inp->inp_flowtype,
1265 				    &rss_bucket);
1266 				if (retval == 0)
1267 					optval = rss_bucket;
1268 				else
1269 					error = EINVAL;
1270 				break;
1271 			case IP_RECVRSSBUCKETID:
1272 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1273 				break;
1274 #endif
1275 			case IP_BINDMULTI:
1276 				optval = OPTBIT2(INP_BINDMULTI);
1277 				break;
1278 			}
1279 			error = sooptcopyout(sopt, &optval, sizeof optval);
1280 			break;
1281 
1282 		/*
1283 		 * Multicast socket options are processed by the in_mcast
1284 		 * module.
1285 		 */
1286 		case IP_MULTICAST_IF:
1287 		case IP_MULTICAST_VIF:
1288 		case IP_MULTICAST_TTL:
1289 		case IP_MULTICAST_LOOP:
1290 		case IP_MSFILTER:
1291 			error = inp_getmoptions(inp, sopt);
1292 			break;
1293 
1294 #ifdef IPSEC
1295 		case IP_IPSEC_POLICY:
1296 		{
1297 			struct mbuf *m = NULL;
1298 			caddr_t req = NULL;
1299 			size_t len = 0;
1300 
1301 			if (m != 0) {
1302 				req = mtod(m, caddr_t);
1303 				len = m->m_len;
1304 			}
1305 			error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1306 			if (error == 0)
1307 				error = soopt_mcopyout(sopt, m); /* XXX */
1308 			if (error == 0)
1309 				m_freem(m);
1310 			break;
1311 		}
1312 #endif /* IPSEC */
1313 
1314 		default:
1315 			error = ENOPROTOOPT;
1316 			break;
1317 		}
1318 		break;
1319 	}
1320 	return (error);
1321 }
1322 
1323 /*
1324  * Routine called from ip_output() to loop back a copy of an IP multicast
1325  * packet to the input queue of a specified interface.  Note that this
1326  * calls the output routine of the loopback "driver", but with an interface
1327  * pointer that might NOT be a loopback interface -- evil, but easier than
1328  * replicating that code here.
1329  */
1330 static void
1331 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1332     int hlen)
1333 {
1334 	register struct ip *ip;
1335 	struct mbuf *copym;
1336 
1337 	/*
1338 	 * Make a deep copy of the packet because we're going to
1339 	 * modify the pack in order to generate checksums.
1340 	 */
1341 	copym = m_dup(m, M_NOWAIT);
1342 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1343 		copym = m_pullup(copym, hlen);
1344 	if (copym != NULL) {
1345 		/* If needed, compute the checksum and mark it as valid. */
1346 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1347 			in_delayed_cksum(copym);
1348 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1349 			copym->m_pkthdr.csum_flags |=
1350 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1351 			copym->m_pkthdr.csum_data = 0xffff;
1352 		}
1353 		/*
1354 		 * We don't bother to fragment if the IP length is greater
1355 		 * than the interface's MTU.  Can this possibly matter?
1356 		 */
1357 		ip = mtod(copym, struct ip *);
1358 		ip->ip_sum = 0;
1359 		ip->ip_sum = in_cksum(copym, hlen);
1360 #if 1 /* XXX */
1361 		if (dst->sin_family != AF_INET) {
1362 			printf("ip_mloopback: bad address family %d\n",
1363 						dst->sin_family);
1364 			dst->sin_family = AF_INET;
1365 		}
1366 #endif
1367 		if_simloop(ifp, copym, dst->sin_family, 0);
1368 	}
1369 }
1370