xref: /freebsd/sys/netinet/ip_output.c (revision f02f7422801bb39f5eaab8fc383fa7b70c467ff9)
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/vnet.h>
69 
70 #include <netinet/in.h>
71 #include <netinet/in_kdtrace.h>
72 #include <netinet/in_systm.h>
73 #include <netinet/ip.h>
74 #include <netinet/in_pcb.h>
75 #include <netinet/in_rss.h>
76 #include <netinet/in_var.h>
77 #include <netinet/ip_var.h>
78 #include <netinet/ip_options.h>
79 #ifdef SCTP
80 #include <netinet/sctp.h>
81 #include <netinet/sctp_crc32.h>
82 #endif
83 
84 #ifdef IPSEC
85 #include <netinet/ip_ipsec.h>
86 #include <netipsec/ipsec.h>
87 #endif /* IPSEC*/
88 
89 #include <machine/in_cksum.h>
90 
91 #include <security/mac/mac_framework.h>
92 
93 VNET_DEFINE(u_short, ip_id);
94 
95 #ifdef MBUF_STRESS_TEST
96 static int mbuf_frag_size = 0;
97 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
98 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
99 #endif
100 
101 static void	ip_mloopback
102 	(struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
103 
104 
105 extern int in_mcast_loop;
106 extern	struct protosw inetsw[];
107 
108 /*
109  * IP output.  The packet in mbuf chain m contains a skeletal IP
110  * header (with len, off, ttl, proto, tos, src, dst).
111  * The mbuf chain containing the packet will be freed.
112  * The mbuf opt, if present, will not be freed.
113  * If route ro is present and has ro_rt initialized, route lookup would be
114  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
115  * then result of route lookup is stored in ro->ro_rt.
116  *
117  * In the IP forwarding case, the packet will arrive with options already
118  * inserted, so must have a NULL opt pointer.
119  */
120 int
121 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
122     struct ip_moptions *imo, struct inpcb *inp)
123 {
124 	struct ip *ip;
125 	struct ifnet *ifp = NULL;	/* keep compiler happy */
126 	struct mbuf *m0;
127 	int hlen = sizeof (struct ip);
128 	int mtu;
129 	int error = 0;
130 	struct sockaddr_in *dst;
131 	const struct sockaddr_in *gw;
132 	struct in_ifaddr *ia;
133 	int isbroadcast;
134 	uint16_t ip_len, ip_off;
135 	struct route iproute;
136 	struct rtentry *rte;	/* cache for ro->ro_rt */
137 	struct in_addr odst;
138 	struct m_tag *fwd_tag = NULL;
139 	int have_ia_ref;
140 #ifdef IPSEC
141 	int no_route_but_check_spd = 0;
142 #endif
143 	M_ASSERTPKTHDR(m);
144 
145 	if (inp != NULL) {
146 		INP_LOCK_ASSERT(inp);
147 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
148 		if (((flags & IP_NODEFAULTFLOWID) == 0) &&
149 		    inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
150 			m->m_pkthdr.flowid = inp->inp_flowid;
151 			M_HASHTYPE_SET(m, inp->inp_flowtype);
152 			m->m_flags |= M_FLOWID;
153 		}
154 	}
155 
156 	if (ro == NULL) {
157 		ro = &iproute;
158 		bzero(ro, sizeof (*ro));
159 	}
160 
161 #ifdef FLOWTABLE
162 	if (ro->ro_rt == NULL)
163 		(void )flowtable_lookup(AF_INET, m, ro);
164 #endif
165 
166 	if (opt) {
167 		int len = 0;
168 		m = ip_insertoptions(m, opt, &len);
169 		if (len != 0)
170 			hlen = len; /* ip->ip_hl is updated above */
171 	}
172 	ip = mtod(m, struct ip *);
173 	ip_len = ntohs(ip->ip_len);
174 	ip_off = ntohs(ip->ip_off);
175 
176 	/*
177 	 * Fill in IP header.  If we are not allowing fragmentation,
178 	 * then the ip_id field is meaningless, but we don't set it
179 	 * to zero.  Doing so causes various problems when devices along
180 	 * the path (routers, load balancers, firewalls, etc.) illegally
181 	 * disable DF on our packet.  Note that a 16-bit counter
182 	 * will wrap around in less than 10 seconds at 100 Mbit/s on a
183 	 * medium with MTU 1500.  See Steven M. Bellovin, "A Technique
184 	 * for Counting NATted Hosts", Proc. IMW'02, available at
185 	 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
186 	 */
187 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
188 		ip->ip_v = IPVERSION;
189 		ip->ip_hl = hlen >> 2;
190 		ip->ip_id = ip_newid();
191 		IPSTAT_INC(ips_localout);
192 	} else {
193 		/* Header already set, fetch hlen from there */
194 		hlen = ip->ip_hl << 2;
195 	}
196 
197 	/*
198 	 * dst/gw handling:
199 	 *
200 	 * dst can be rewritten but always points to &ro->ro_dst.
201 	 * gw is readonly but can point either to dst OR rt_gateway,
202 	 * therefore we need restore gw if we're redoing lookup.
203 	 */
204 	gw = dst = (struct sockaddr_in *)&ro->ro_dst;
205 again:
206 	ia = NULL;
207 	have_ia_ref = 0;
208 	/*
209 	 * If there is a cached route, check that it is to the same
210 	 * destination and is still up.  If not, free it and try again.
211 	 * The address family should also be checked in case of sharing
212 	 * the cache with IPv6.
213 	 */
214 	rte = ro->ro_rt;
215 	if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
216 		    rte->rt_ifp == NULL ||
217 		    !RT_LINK_IS_UP(rte->rt_ifp) ||
218 			  dst->sin_family != AF_INET ||
219 			  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
220 		RO_RTFREE(ro);
221 		ro->ro_lle = NULL;
222 		rte = NULL;
223 		gw = dst;
224 	}
225 	if (rte == NULL && fwd_tag == NULL) {
226 		bzero(dst, sizeof(*dst));
227 		dst->sin_family = AF_INET;
228 		dst->sin_len = sizeof(*dst);
229 		dst->sin_addr = ip->ip_dst;
230 	}
231 	/*
232 	 * If routing to interface only, short circuit routing lookup.
233 	 * The use of an all-ones broadcast address implies this; an
234 	 * interface is specified by the broadcast address of an interface,
235 	 * or the destination address of a ptp interface.
236 	 */
237 	if (flags & IP_SENDONES) {
238 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
239 						      M_GETFIB(m)))) == NULL &&
240 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
241 						    M_GETFIB(m)))) == NULL) {
242 			IPSTAT_INC(ips_noroute);
243 			error = ENETUNREACH;
244 			goto bad;
245 		}
246 		have_ia_ref = 1;
247 		ip->ip_dst.s_addr = INADDR_BROADCAST;
248 		dst->sin_addr = ip->ip_dst;
249 		ifp = ia->ia_ifp;
250 		ip->ip_ttl = 1;
251 		isbroadcast = 1;
252 	} else if (flags & IP_ROUTETOIF) {
253 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
254 						    M_GETFIB(m)))) == NULL &&
255 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
256 						M_GETFIB(m)))) == NULL) {
257 			IPSTAT_INC(ips_noroute);
258 			error = ENETUNREACH;
259 			goto bad;
260 		}
261 		have_ia_ref = 1;
262 		ifp = ia->ia_ifp;
263 		ip->ip_ttl = 1;
264 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
265 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
266 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
267 		/*
268 		 * Bypass the normal routing lookup for multicast
269 		 * packets if the interface is specified.
270 		 */
271 		ifp = imo->imo_multicast_ifp;
272 		IFP_TO_IA(ifp, ia);
273 		if (ia)
274 			have_ia_ref = 1;
275 		isbroadcast = 0;	/* fool gcc */
276 	} else {
277 		/*
278 		 * We want to do any cloning requested by the link layer,
279 		 * as this is probably required in all cases for correct
280 		 * operation (as it is for ARP).
281 		 */
282 		if (rte == NULL) {
283 #ifdef RADIX_MPATH
284 			rtalloc_mpath_fib(ro,
285 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
286 			    inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
287 #else
288 			in_rtalloc_ign(ro, 0,
289 			    inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
290 #endif
291 			rte = ro->ro_rt;
292 		}
293 		if (rte == NULL ||
294 		    rte->rt_ifp == NULL ||
295 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
296 #ifdef IPSEC
297 			/*
298 			 * There is no route for this packet, but it is
299 			 * possible that a matching SPD entry exists.
300 			 */
301 			no_route_but_check_spd = 1;
302 			mtu = 0; /* Silence GCC warning. */
303 			goto sendit;
304 #endif
305 			IPSTAT_INC(ips_noroute);
306 			error = EHOSTUNREACH;
307 			goto bad;
308 		}
309 		ia = ifatoia(rte->rt_ifa);
310 		ifp = rte->rt_ifp;
311 		counter_u64_add(rte->rt_pksent, 1);
312 		if (rte->rt_flags & RTF_GATEWAY)
313 			gw = (struct sockaddr_in *)rte->rt_gateway;
314 		if (rte->rt_flags & RTF_HOST)
315 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
316 		else
317 			isbroadcast = in_broadcast(gw->sin_addr, ifp);
318 	}
319 	/*
320 	 * Calculate MTU.  If we have a route that is up, use that,
321 	 * otherwise use the interface's MTU.
322 	 */
323 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
324 		/*
325 		 * This case can happen if the user changed the MTU
326 		 * of an interface after enabling IP on it.  Because
327 		 * most netifs don't keep track of routes pointing to
328 		 * them, there is no way for one to update all its
329 		 * routes when the MTU is changed.
330 		 */
331 		if (rte->rt_mtu > ifp->if_mtu)
332 			rte->rt_mtu = ifp->if_mtu;
333 		mtu = rte->rt_mtu;
334 	} else {
335 		mtu = ifp->if_mtu;
336 	}
337 	/* Catch a possible divide by zero later. */
338 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
339 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
340 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
341 		m->m_flags |= M_MCAST;
342 		/*
343 		 * IP destination address is multicast.  Make sure "gw"
344 		 * still points to the address in "ro".  (It may have been
345 		 * changed to point to a gateway address, above.)
346 		 */
347 		gw = dst;
348 		/*
349 		 * See if the caller provided any multicast options
350 		 */
351 		if (imo != NULL) {
352 			ip->ip_ttl = imo->imo_multicast_ttl;
353 			if (imo->imo_multicast_vif != -1)
354 				ip->ip_src.s_addr =
355 				    ip_mcast_src ?
356 				    ip_mcast_src(imo->imo_multicast_vif) :
357 				    INADDR_ANY;
358 		} else
359 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
360 		/*
361 		 * Confirm that the outgoing interface supports multicast.
362 		 */
363 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
364 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
365 				IPSTAT_INC(ips_noroute);
366 				error = ENETUNREACH;
367 				goto bad;
368 			}
369 		}
370 		/*
371 		 * If source address not specified yet, use address
372 		 * of outgoing interface.
373 		 */
374 		if (ip->ip_src.s_addr == INADDR_ANY) {
375 			/* Interface may have no addresses. */
376 			if (ia != NULL)
377 				ip->ip_src = IA_SIN(ia)->sin_addr;
378 		}
379 
380 		if ((imo == NULL && in_mcast_loop) ||
381 		    (imo && imo->imo_multicast_loop)) {
382 			/*
383 			 * Loop back multicast datagram if not expressly
384 			 * forbidden to do so, even if we are not a member
385 			 * of the group; ip_input() will filter it later,
386 			 * thus deferring a hash lookup and mutex acquisition
387 			 * at the expense of a cheap copy using m_copym().
388 			 */
389 			ip_mloopback(ifp, m, dst, hlen);
390 		} else {
391 			/*
392 			 * If we are acting as a multicast router, perform
393 			 * multicast forwarding as if the packet had just
394 			 * arrived on the interface to which we are about
395 			 * to send.  The multicast forwarding function
396 			 * recursively calls this function, using the
397 			 * IP_FORWARDING flag to prevent infinite recursion.
398 			 *
399 			 * Multicasts that are looped back by ip_mloopback(),
400 			 * above, will be forwarded by the ip_input() routine,
401 			 * if necessary.
402 			 */
403 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
404 				/*
405 				 * If rsvp daemon is not running, do not
406 				 * set ip_moptions. This ensures that the packet
407 				 * is multicast and not just sent down one link
408 				 * as prescribed by rsvpd.
409 				 */
410 				if (!V_rsvp_on)
411 					imo = NULL;
412 				if (ip_mforward &&
413 				    ip_mforward(ip, ifp, m, imo) != 0) {
414 					m_freem(m);
415 					goto done;
416 				}
417 			}
418 		}
419 
420 		/*
421 		 * Multicasts with a time-to-live of zero may be looped-
422 		 * back, above, but must not be transmitted on a network.
423 		 * Also, multicasts addressed to the loopback interface
424 		 * are not sent -- the above call to ip_mloopback() will
425 		 * loop back a copy. ip_input() will drop the copy if
426 		 * this host does not belong to the destination group on
427 		 * the loopback interface.
428 		 */
429 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
430 			m_freem(m);
431 			goto done;
432 		}
433 
434 		goto sendit;
435 	}
436 
437 	/*
438 	 * If the source address is not specified yet, use the address
439 	 * of the outoing interface.
440 	 */
441 	if (ip->ip_src.s_addr == INADDR_ANY) {
442 		/* Interface may have no addresses. */
443 		if (ia != NULL) {
444 			ip->ip_src = IA_SIN(ia)->sin_addr;
445 		}
446 	}
447 
448 	/*
449 	 * Look for broadcast address and
450 	 * verify user is allowed to send
451 	 * such a packet.
452 	 */
453 	if (isbroadcast) {
454 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
455 			error = EADDRNOTAVAIL;
456 			goto bad;
457 		}
458 		if ((flags & IP_ALLOWBROADCAST) == 0) {
459 			error = EACCES;
460 			goto bad;
461 		}
462 		/* don't allow broadcast messages to be fragmented */
463 		if (ip_len > mtu) {
464 			error = EMSGSIZE;
465 			goto bad;
466 		}
467 		m->m_flags |= M_BCAST;
468 	} else {
469 		m->m_flags &= ~M_BCAST;
470 	}
471 
472 sendit:
473 #ifdef IPSEC
474 	switch(ip_ipsec_output(&m, inp, &flags, &error)) {
475 	case 1:
476 		goto bad;
477 	case -1:
478 		goto done;
479 	case 0:
480 	default:
481 		break;	/* Continue with packet processing. */
482 	}
483 	/*
484 	 * Check if there was a route for this packet; return error if not.
485 	 */
486 	if (no_route_but_check_spd) {
487 		IPSTAT_INC(ips_noroute);
488 		error = EHOSTUNREACH;
489 		goto bad;
490 	}
491 	/* Update variables that are affected by ipsec4_output(). */
492 	ip = mtod(m, struct ip *);
493 	hlen = ip->ip_hl << 2;
494 #endif /* IPSEC */
495 
496 	/* Jump over all PFIL processing if hooks are not active. */
497 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
498 		goto passout;
499 
500 	/* Run through list of hooks for output packets. */
501 	odst.s_addr = ip->ip_dst.s_addr;
502 	error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
503 	if (error != 0 || m == NULL)
504 		goto done;
505 
506 	ip = mtod(m, struct ip *);
507 
508 	/* See if destination IP address was changed by packet filter. */
509 	if (odst.s_addr != ip->ip_dst.s_addr) {
510 		m->m_flags |= M_SKIP_FIREWALL;
511 		/* If destination is now ourself drop to ip_input(). */
512 		if (in_localip(ip->ip_dst)) {
513 			m->m_flags |= M_FASTFWD_OURS;
514 			if (m->m_pkthdr.rcvif == NULL)
515 				m->m_pkthdr.rcvif = V_loif;
516 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
517 				m->m_pkthdr.csum_flags |=
518 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
519 				m->m_pkthdr.csum_data = 0xffff;
520 			}
521 			m->m_pkthdr.csum_flags |=
522 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
523 #ifdef SCTP
524 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
525 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
526 #endif
527 			error = netisr_queue(NETISR_IP, m);
528 			goto done;
529 		} else {
530 			if (have_ia_ref)
531 				ifa_free(&ia->ia_ifa);
532 			goto again;	/* Redo the routing table lookup. */
533 		}
534 	}
535 
536 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
537 	if (m->m_flags & M_FASTFWD_OURS) {
538 		if (m->m_pkthdr.rcvif == NULL)
539 			m->m_pkthdr.rcvif = V_loif;
540 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
541 			m->m_pkthdr.csum_flags |=
542 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
543 			m->m_pkthdr.csum_data = 0xffff;
544 		}
545 #ifdef SCTP
546 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
547 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
548 #endif
549 		m->m_pkthdr.csum_flags |=
550 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
551 
552 		error = netisr_queue(NETISR_IP, m);
553 		goto done;
554 	}
555 	/* Or forward to some other address? */
556 	if ((m->m_flags & M_IP_NEXTHOP) &&
557 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
558 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
559 		m->m_flags |= M_SKIP_FIREWALL;
560 		m->m_flags &= ~M_IP_NEXTHOP;
561 		m_tag_delete(m, fwd_tag);
562 		if (have_ia_ref)
563 			ifa_free(&ia->ia_ifa);
564 		goto again;
565 	}
566 
567 passout:
568 	/* 127/8 must not appear on wire - RFC1122. */
569 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
570 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
571 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
572 			IPSTAT_INC(ips_badaddr);
573 			error = EADDRNOTAVAIL;
574 			goto bad;
575 		}
576 	}
577 
578 	m->m_pkthdr.csum_flags |= CSUM_IP;
579 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
580 		in_delayed_cksum(m);
581 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
582 	}
583 #ifdef SCTP
584 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
585 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
586 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
587 	}
588 #endif
589 
590 	/*
591 	 * If small enough for interface, or the interface will take
592 	 * care of the fragmentation for us, we can just send directly.
593 	 */
594 	if (ip_len <= mtu ||
595 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
596 		ip->ip_sum = 0;
597 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
598 			ip->ip_sum = in_cksum(m, hlen);
599 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
600 		}
601 
602 		/*
603 		 * Record statistics for this interface address.
604 		 * With CSUM_TSO the byte/packet count will be slightly
605 		 * incorrect because we count the IP+TCP headers only
606 		 * once instead of for every generated packet.
607 		 */
608 		if (!(flags & IP_FORWARDING) && ia) {
609 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
610 				counter_u64_add(ia->ia_ifa.ifa_opackets,
611 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
612 			else
613 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
614 
615 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
616 		}
617 #ifdef MBUF_STRESS_TEST
618 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
619 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
620 #endif
621 		/*
622 		 * Reset layer specific mbuf flags
623 		 * to avoid confusing lower layers.
624 		 */
625 		m_clrprotoflags(m);
626 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
627 		error = (*ifp->if_output)(ifp, m,
628 		    (const struct sockaddr *)gw, ro);
629 		goto done;
630 	}
631 
632 	/* Balk when DF bit is set or the interface didn't support TSO. */
633 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
634 		error = EMSGSIZE;
635 		IPSTAT_INC(ips_cantfrag);
636 		goto bad;
637 	}
638 
639 	/*
640 	 * Too large for interface; fragment if possible. If successful,
641 	 * on return, m will point to a list of packets to be sent.
642 	 */
643 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
644 	if (error)
645 		goto bad;
646 	for (; m; m = m0) {
647 		m0 = m->m_nextpkt;
648 		m->m_nextpkt = 0;
649 		if (error == 0) {
650 			/* Record statistics for this interface address. */
651 			if (ia != NULL) {
652 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
653 				counter_u64_add(ia->ia_ifa.ifa_obytes,
654 				    m->m_pkthdr.len);
655 			}
656 			/*
657 			 * Reset layer specific mbuf flags
658 			 * to avoid confusing upper layers.
659 			 */
660 			m_clrprotoflags(m);
661 
662 			IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
663 			error = (*ifp->if_output)(ifp, m,
664 			    (const struct sockaddr *)gw, ro);
665 		} else
666 			m_freem(m);
667 	}
668 
669 	if (error == 0)
670 		IPSTAT_INC(ips_fragmented);
671 
672 done:
673 	if (ro == &iproute)
674 		RO_RTFREE(ro);
675 	if (have_ia_ref)
676 		ifa_free(&ia->ia_ifa);
677 	return (error);
678 bad:
679 	m_freem(m);
680 	goto done;
681 }
682 
683 /*
684  * Create a chain of fragments which fit the given mtu. m_frag points to the
685  * mbuf to be fragmented; on return it points to the chain with the fragments.
686  * Return 0 if no error. If error, m_frag may contain a partially built
687  * chain of fragments that should be freed by the caller.
688  *
689  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
690  */
691 int
692 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
693     u_long if_hwassist_flags)
694 {
695 	int error = 0;
696 	int hlen = ip->ip_hl << 2;
697 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
698 	int off;
699 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
700 	int firstlen;
701 	struct mbuf **mnext;
702 	int nfrags;
703 	uint16_t ip_len, ip_off;
704 
705 	ip_len = ntohs(ip->ip_len);
706 	ip_off = ntohs(ip->ip_off);
707 
708 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
709 		IPSTAT_INC(ips_cantfrag);
710 		return EMSGSIZE;
711 	}
712 
713 	/*
714 	 * Must be able to put at least 8 bytes per fragment.
715 	 */
716 	if (len < 8)
717 		return EMSGSIZE;
718 
719 	/*
720 	 * If the interface will not calculate checksums on
721 	 * fragmented packets, then do it here.
722 	 */
723 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
724 		in_delayed_cksum(m0);
725 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
726 	}
727 #ifdef SCTP
728 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
729 		sctp_delayed_cksum(m0, hlen);
730 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
731 	}
732 #endif
733 	if (len > PAGE_SIZE) {
734 		/*
735 		 * Fragment large datagrams such that each segment
736 		 * contains a multiple of PAGE_SIZE amount of data,
737 		 * plus headers. This enables a receiver to perform
738 		 * page-flipping zero-copy optimizations.
739 		 *
740 		 * XXX When does this help given that sender and receiver
741 		 * could have different page sizes, and also mtu could
742 		 * be less than the receiver's page size ?
743 		 */
744 		int newlen;
745 		struct mbuf *m;
746 
747 		for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
748 			off += m->m_len;
749 
750 		/*
751 		 * firstlen (off - hlen) must be aligned on an
752 		 * 8-byte boundary
753 		 */
754 		if (off < hlen)
755 			goto smart_frag_failure;
756 		off = ((off - hlen) & ~7) + hlen;
757 		newlen = (~PAGE_MASK) & mtu;
758 		if ((newlen + sizeof (struct ip)) > mtu) {
759 			/* we failed, go back the default */
760 smart_frag_failure:
761 			newlen = len;
762 			off = hlen + len;
763 		}
764 		len = newlen;
765 
766 	} else {
767 		off = hlen + len;
768 	}
769 
770 	firstlen = off - hlen;
771 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
772 
773 	/*
774 	 * Loop through length of segment after first fragment,
775 	 * make new header and copy data of each part and link onto chain.
776 	 * Here, m0 is the original packet, m is the fragment being created.
777 	 * The fragments are linked off the m_nextpkt of the original
778 	 * packet, which after processing serves as the first fragment.
779 	 */
780 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
781 		struct ip *mhip;	/* ip header on the fragment */
782 		struct mbuf *m;
783 		int mhlen = sizeof (struct ip);
784 
785 		m = m_gethdr(M_NOWAIT, MT_DATA);
786 		if (m == NULL) {
787 			error = ENOBUFS;
788 			IPSTAT_INC(ips_odropped);
789 			goto done;
790 		}
791 		m->m_flags |= (m0->m_flags & M_MCAST);
792 		/*
793 		 * In the first mbuf, leave room for the link header, then
794 		 * copy the original IP header including options. The payload
795 		 * goes into an additional mbuf chain returned by m_copym().
796 		 */
797 		m->m_data += max_linkhdr;
798 		mhip = mtod(m, struct ip *);
799 		*mhip = *ip;
800 		if (hlen > sizeof (struct ip)) {
801 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
802 			mhip->ip_v = IPVERSION;
803 			mhip->ip_hl = mhlen >> 2;
804 		}
805 		m->m_len = mhlen;
806 		/* XXX do we need to add ip_off below ? */
807 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
808 		if (off + len >= ip_len)
809 			len = ip_len - off;
810 		else
811 			mhip->ip_off |= IP_MF;
812 		mhip->ip_len = htons((u_short)(len + mhlen));
813 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
814 		if (m->m_next == NULL) {	/* copy failed */
815 			m_free(m);
816 			error = ENOBUFS;	/* ??? */
817 			IPSTAT_INC(ips_odropped);
818 			goto done;
819 		}
820 		m->m_pkthdr.len = mhlen + len;
821 		m->m_pkthdr.rcvif = NULL;
822 #ifdef MAC
823 		mac_netinet_fragment(m0, m);
824 #endif
825 		m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
826 		mhip->ip_off = htons(mhip->ip_off);
827 		mhip->ip_sum = 0;
828 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
829 			mhip->ip_sum = in_cksum(m, mhlen);
830 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
831 		}
832 		*mnext = m;
833 		mnext = &m->m_nextpkt;
834 	}
835 	IPSTAT_ADD(ips_ofragments, nfrags);
836 
837 	/*
838 	 * Update first fragment by trimming what's been copied out
839 	 * and updating header.
840 	 */
841 	m_adj(m0, hlen + firstlen - ip_len);
842 	m0->m_pkthdr.len = hlen + firstlen;
843 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
844 	ip->ip_off = htons(ip_off | IP_MF);
845 	ip->ip_sum = 0;
846 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
847 		ip->ip_sum = in_cksum(m0, hlen);
848 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
849 	}
850 
851 done:
852 	*m_frag = m0;
853 	return error;
854 }
855 
856 void
857 in_delayed_cksum(struct mbuf *m)
858 {
859 	struct ip *ip;
860 	uint16_t csum, offset, ip_len;
861 
862 	ip = mtod(m, struct ip *);
863 	offset = ip->ip_hl << 2 ;
864 	ip_len = ntohs(ip->ip_len);
865 	csum = in_cksum_skip(m, ip_len, offset);
866 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
867 		csum = 0xffff;
868 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
869 
870 	/* find the mbuf in the chain where the checksum starts*/
871 	while ((m != NULL) && (offset >= m->m_len)) {
872 		offset -= m->m_len;
873 		m = m->m_next;
874 	}
875 	KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
876 	KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
877 	*(u_short *)(m->m_data + offset) = csum;
878 }
879 
880 /*
881  * IP socket option processing.
882  */
883 int
884 ip_ctloutput(struct socket *so, struct sockopt *sopt)
885 {
886 	struct	inpcb *inp = sotoinpcb(so);
887 	int	error, optval;
888 #ifdef	RSS
889 	uint32_t rss_bucket;
890 	int retval;
891 #endif
892 
893 	error = optval = 0;
894 	if (sopt->sopt_level != IPPROTO_IP) {
895 		error = EINVAL;
896 
897 		if (sopt->sopt_level == SOL_SOCKET &&
898 		    sopt->sopt_dir == SOPT_SET) {
899 			switch (sopt->sopt_name) {
900 			case SO_REUSEADDR:
901 				INP_WLOCK(inp);
902 				if ((so->so_options & SO_REUSEADDR) != 0)
903 					inp->inp_flags2 |= INP_REUSEADDR;
904 				else
905 					inp->inp_flags2 &= ~INP_REUSEADDR;
906 				INP_WUNLOCK(inp);
907 				error = 0;
908 				break;
909 			case SO_REUSEPORT:
910 				INP_WLOCK(inp);
911 				if ((so->so_options & SO_REUSEPORT) != 0)
912 					inp->inp_flags2 |= INP_REUSEPORT;
913 				else
914 					inp->inp_flags2 &= ~INP_REUSEPORT;
915 				INP_WUNLOCK(inp);
916 				error = 0;
917 				break;
918 			case SO_SETFIB:
919 				INP_WLOCK(inp);
920 				inp->inp_inc.inc_fibnum = so->so_fibnum;
921 				INP_WUNLOCK(inp);
922 				error = 0;
923 				break;
924 			default:
925 				break;
926 			}
927 		}
928 		return (error);
929 	}
930 
931 	switch (sopt->sopt_dir) {
932 	case SOPT_SET:
933 		switch (sopt->sopt_name) {
934 		case IP_OPTIONS:
935 #ifdef notyet
936 		case IP_RETOPTS:
937 #endif
938 		{
939 			struct mbuf *m;
940 			if (sopt->sopt_valsize > MLEN) {
941 				error = EMSGSIZE;
942 				break;
943 			}
944 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
945 			if (m == NULL) {
946 				error = ENOBUFS;
947 				break;
948 			}
949 			m->m_len = sopt->sopt_valsize;
950 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
951 					    m->m_len);
952 			if (error) {
953 				m_free(m);
954 				break;
955 			}
956 			INP_WLOCK(inp);
957 			error = ip_pcbopts(inp, sopt->sopt_name, m);
958 			INP_WUNLOCK(inp);
959 			return (error);
960 		}
961 
962 		case IP_BINDANY:
963 			if (sopt->sopt_td != NULL) {
964 				error = priv_check(sopt->sopt_td,
965 				    PRIV_NETINET_BINDANY);
966 				if (error)
967 					break;
968 			}
969 			/* FALLTHROUGH */
970 		case IP_BINDMULTI:
971 #ifdef	RSS
972 		case IP_RSS_LISTEN_BUCKET:
973 #endif
974 		case IP_TOS:
975 		case IP_TTL:
976 		case IP_MINTTL:
977 		case IP_RECVOPTS:
978 		case IP_RECVRETOPTS:
979 		case IP_RECVDSTADDR:
980 		case IP_RECVTTL:
981 		case IP_RECVIF:
982 		case IP_FAITH:
983 		case IP_ONESBCAST:
984 		case IP_DONTFRAG:
985 		case IP_RECVTOS:
986 		case IP_RECVFLOWID:
987 #ifdef	RSS
988 		case IP_RECVRSSBUCKETID:
989 #endif
990 			error = sooptcopyin(sopt, &optval, sizeof optval,
991 					    sizeof optval);
992 			if (error)
993 				break;
994 
995 			switch (sopt->sopt_name) {
996 			case IP_TOS:
997 				inp->inp_ip_tos = optval;
998 				break;
999 
1000 			case IP_TTL:
1001 				inp->inp_ip_ttl = optval;
1002 				break;
1003 
1004 			case IP_MINTTL:
1005 				if (optval >= 0 && optval <= MAXTTL)
1006 					inp->inp_ip_minttl = optval;
1007 				else
1008 					error = EINVAL;
1009 				break;
1010 
1011 #define	OPTSET(bit) do {						\
1012 	INP_WLOCK(inp);							\
1013 	if (optval)							\
1014 		inp->inp_flags |= bit;					\
1015 	else								\
1016 		inp->inp_flags &= ~bit;					\
1017 	INP_WUNLOCK(inp);						\
1018 } while (0)
1019 
1020 #define	OPTSET2(bit, val) do {						\
1021 	INP_WLOCK(inp);							\
1022 	if (val)							\
1023 		inp->inp_flags2 |= bit;					\
1024 	else								\
1025 		inp->inp_flags2 &= ~bit;				\
1026 	INP_WUNLOCK(inp);						\
1027 } while (0)
1028 
1029 			case IP_RECVOPTS:
1030 				OPTSET(INP_RECVOPTS);
1031 				break;
1032 
1033 			case IP_RECVRETOPTS:
1034 				OPTSET(INP_RECVRETOPTS);
1035 				break;
1036 
1037 			case IP_RECVDSTADDR:
1038 				OPTSET(INP_RECVDSTADDR);
1039 				break;
1040 
1041 			case IP_RECVTTL:
1042 				OPTSET(INP_RECVTTL);
1043 				break;
1044 
1045 			case IP_RECVIF:
1046 				OPTSET(INP_RECVIF);
1047 				break;
1048 
1049 			case IP_FAITH:
1050 				OPTSET(INP_FAITH);
1051 				break;
1052 
1053 			case IP_ONESBCAST:
1054 				OPTSET(INP_ONESBCAST);
1055 				break;
1056 			case IP_DONTFRAG:
1057 				OPTSET(INP_DONTFRAG);
1058 				break;
1059 			case IP_BINDANY:
1060 				OPTSET(INP_BINDANY);
1061 				break;
1062 			case IP_RECVTOS:
1063 				OPTSET(INP_RECVTOS);
1064 				break;
1065 			case IP_BINDMULTI:
1066 				OPTSET2(INP_BINDMULTI, optval);
1067 				break;
1068 			case IP_RECVFLOWID:
1069 				OPTSET2(INP_RECVFLOWID, optval);
1070 				break;
1071 #ifdef	RSS
1072 			case IP_RSS_LISTEN_BUCKET:
1073 				if ((optval >= 0) &&
1074 				    (optval < rss_getnumbuckets())) {
1075 					inp->inp_rss_listen_bucket = optval;
1076 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1077 				} else {
1078 					error = EINVAL;
1079 				}
1080 				break;
1081 			case IP_RECVRSSBUCKETID:
1082 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1083 				break;
1084 #endif
1085 			}
1086 			break;
1087 #undef OPTSET
1088 #undef OPTSET2
1089 
1090 		/*
1091 		 * Multicast socket options are processed by the in_mcast
1092 		 * module.
1093 		 */
1094 		case IP_MULTICAST_IF:
1095 		case IP_MULTICAST_VIF:
1096 		case IP_MULTICAST_TTL:
1097 		case IP_MULTICAST_LOOP:
1098 		case IP_ADD_MEMBERSHIP:
1099 		case IP_DROP_MEMBERSHIP:
1100 		case IP_ADD_SOURCE_MEMBERSHIP:
1101 		case IP_DROP_SOURCE_MEMBERSHIP:
1102 		case IP_BLOCK_SOURCE:
1103 		case IP_UNBLOCK_SOURCE:
1104 		case IP_MSFILTER:
1105 		case MCAST_JOIN_GROUP:
1106 		case MCAST_LEAVE_GROUP:
1107 		case MCAST_JOIN_SOURCE_GROUP:
1108 		case MCAST_LEAVE_SOURCE_GROUP:
1109 		case MCAST_BLOCK_SOURCE:
1110 		case MCAST_UNBLOCK_SOURCE:
1111 			error = inp_setmoptions(inp, sopt);
1112 			break;
1113 
1114 		case IP_PORTRANGE:
1115 			error = sooptcopyin(sopt, &optval, sizeof optval,
1116 					    sizeof optval);
1117 			if (error)
1118 				break;
1119 
1120 			INP_WLOCK(inp);
1121 			switch (optval) {
1122 			case IP_PORTRANGE_DEFAULT:
1123 				inp->inp_flags &= ~(INP_LOWPORT);
1124 				inp->inp_flags &= ~(INP_HIGHPORT);
1125 				break;
1126 
1127 			case IP_PORTRANGE_HIGH:
1128 				inp->inp_flags &= ~(INP_LOWPORT);
1129 				inp->inp_flags |= INP_HIGHPORT;
1130 				break;
1131 
1132 			case IP_PORTRANGE_LOW:
1133 				inp->inp_flags &= ~(INP_HIGHPORT);
1134 				inp->inp_flags |= INP_LOWPORT;
1135 				break;
1136 
1137 			default:
1138 				error = EINVAL;
1139 				break;
1140 			}
1141 			INP_WUNLOCK(inp);
1142 			break;
1143 
1144 #ifdef IPSEC
1145 		case IP_IPSEC_POLICY:
1146 		{
1147 			caddr_t req;
1148 			struct mbuf *m;
1149 
1150 			if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1151 				break;
1152 			if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1153 				break;
1154 			req = mtod(m, caddr_t);
1155 			error = ipsec_set_policy(inp, sopt->sopt_name, req,
1156 			    m->m_len, (sopt->sopt_td != NULL) ?
1157 			    sopt->sopt_td->td_ucred : NULL);
1158 			m_freem(m);
1159 			break;
1160 		}
1161 #endif /* IPSEC */
1162 
1163 		default:
1164 			error = ENOPROTOOPT;
1165 			break;
1166 		}
1167 		break;
1168 
1169 	case SOPT_GET:
1170 		switch (sopt->sopt_name) {
1171 		case IP_OPTIONS:
1172 		case IP_RETOPTS:
1173 			if (inp->inp_options)
1174 				error = sooptcopyout(sopt,
1175 						     mtod(inp->inp_options,
1176 							  char *),
1177 						     inp->inp_options->m_len);
1178 			else
1179 				sopt->sopt_valsize = 0;
1180 			break;
1181 
1182 		case IP_TOS:
1183 		case IP_TTL:
1184 		case IP_MINTTL:
1185 		case IP_RECVOPTS:
1186 		case IP_RECVRETOPTS:
1187 		case IP_RECVDSTADDR:
1188 		case IP_RECVTTL:
1189 		case IP_RECVIF:
1190 		case IP_PORTRANGE:
1191 		case IP_FAITH:
1192 		case IP_ONESBCAST:
1193 		case IP_DONTFRAG:
1194 		case IP_BINDANY:
1195 		case IP_RECVTOS:
1196 		case IP_BINDMULTI:
1197 		case IP_FLOWID:
1198 		case IP_FLOWTYPE:
1199 		case IP_RECVFLOWID:
1200 #ifdef	RSS
1201 		case IP_RSSBUCKETID:
1202 		case IP_RECVRSSBUCKETID:
1203 #endif
1204 			switch (sopt->sopt_name) {
1205 
1206 			case IP_TOS:
1207 				optval = inp->inp_ip_tos;
1208 				break;
1209 
1210 			case IP_TTL:
1211 				optval = inp->inp_ip_ttl;
1212 				break;
1213 
1214 			case IP_MINTTL:
1215 				optval = inp->inp_ip_minttl;
1216 				break;
1217 
1218 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1219 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1220 
1221 			case IP_RECVOPTS:
1222 				optval = OPTBIT(INP_RECVOPTS);
1223 				break;
1224 
1225 			case IP_RECVRETOPTS:
1226 				optval = OPTBIT(INP_RECVRETOPTS);
1227 				break;
1228 
1229 			case IP_RECVDSTADDR:
1230 				optval = OPTBIT(INP_RECVDSTADDR);
1231 				break;
1232 
1233 			case IP_RECVTTL:
1234 				optval = OPTBIT(INP_RECVTTL);
1235 				break;
1236 
1237 			case IP_RECVIF:
1238 				optval = OPTBIT(INP_RECVIF);
1239 				break;
1240 
1241 			case IP_PORTRANGE:
1242 				if (inp->inp_flags & INP_HIGHPORT)
1243 					optval = IP_PORTRANGE_HIGH;
1244 				else if (inp->inp_flags & INP_LOWPORT)
1245 					optval = IP_PORTRANGE_LOW;
1246 				else
1247 					optval = 0;
1248 				break;
1249 
1250 			case IP_FAITH:
1251 				optval = OPTBIT(INP_FAITH);
1252 				break;
1253 
1254 			case IP_ONESBCAST:
1255 				optval = OPTBIT(INP_ONESBCAST);
1256 				break;
1257 			case IP_DONTFRAG:
1258 				optval = OPTBIT(INP_DONTFRAG);
1259 				break;
1260 			case IP_BINDANY:
1261 				optval = OPTBIT(INP_BINDANY);
1262 				break;
1263 			case IP_RECVTOS:
1264 				optval = OPTBIT(INP_RECVTOS);
1265 				break;
1266 			case IP_FLOWID:
1267 				optval = inp->inp_flowid;
1268 				break;
1269 			case IP_FLOWTYPE:
1270 				optval = inp->inp_flowtype;
1271 				break;
1272 			case IP_RECVFLOWID:
1273 				optval = OPTBIT2(INP_RECVFLOWID);
1274 				break;
1275 #ifdef	RSS
1276 			case IP_RSSBUCKETID:
1277 				retval = rss_hash2bucket(inp->inp_flowid,
1278 				    inp->inp_flowtype,
1279 				    &rss_bucket);
1280 				if (retval == 0)
1281 					optval = rss_bucket;
1282 				else
1283 					error = EINVAL;
1284 				break;
1285 			case IP_RECVRSSBUCKETID:
1286 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1287 				break;
1288 #endif
1289 			case IP_BINDMULTI:
1290 				optval = OPTBIT2(INP_BINDMULTI);
1291 				break;
1292 			}
1293 			error = sooptcopyout(sopt, &optval, sizeof optval);
1294 			break;
1295 
1296 		/*
1297 		 * Multicast socket options are processed by the in_mcast
1298 		 * module.
1299 		 */
1300 		case IP_MULTICAST_IF:
1301 		case IP_MULTICAST_VIF:
1302 		case IP_MULTICAST_TTL:
1303 		case IP_MULTICAST_LOOP:
1304 		case IP_MSFILTER:
1305 			error = inp_getmoptions(inp, sopt);
1306 			break;
1307 
1308 #ifdef IPSEC
1309 		case IP_IPSEC_POLICY:
1310 		{
1311 			struct mbuf *m = NULL;
1312 			caddr_t req = NULL;
1313 			size_t len = 0;
1314 
1315 			if (m != 0) {
1316 				req = mtod(m, caddr_t);
1317 				len = m->m_len;
1318 			}
1319 			error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1320 			if (error == 0)
1321 				error = soopt_mcopyout(sopt, m); /* XXX */
1322 			if (error == 0)
1323 				m_freem(m);
1324 			break;
1325 		}
1326 #endif /* IPSEC */
1327 
1328 		default:
1329 			error = ENOPROTOOPT;
1330 			break;
1331 		}
1332 		break;
1333 	}
1334 	return (error);
1335 }
1336 
1337 /*
1338  * Routine called from ip_output() to loop back a copy of an IP multicast
1339  * packet to the input queue of a specified interface.  Note that this
1340  * calls the output routine of the loopback "driver", but with an interface
1341  * pointer that might NOT be a loopback interface -- evil, but easier than
1342  * replicating that code here.
1343  */
1344 static void
1345 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1346     int hlen)
1347 {
1348 	register struct ip *ip;
1349 	struct mbuf *copym;
1350 
1351 	/*
1352 	 * Make a deep copy of the packet because we're going to
1353 	 * modify the pack in order to generate checksums.
1354 	 */
1355 	copym = m_dup(m, M_NOWAIT);
1356 	if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1357 		copym = m_pullup(copym, hlen);
1358 	if (copym != NULL) {
1359 		/* If needed, compute the checksum and mark it as valid. */
1360 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1361 			in_delayed_cksum(copym);
1362 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1363 			copym->m_pkthdr.csum_flags |=
1364 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1365 			copym->m_pkthdr.csum_data = 0xffff;
1366 		}
1367 		/*
1368 		 * We don't bother to fragment if the IP length is greater
1369 		 * than the interface's MTU.  Can this possibly matter?
1370 		 */
1371 		ip = mtod(copym, struct ip *);
1372 		ip->ip_sum = 0;
1373 		ip->ip_sum = in_cksum(copym, hlen);
1374 #if 1 /* XXX */
1375 		if (dst->sin_family != AF_INET) {
1376 			printf("ip_mloopback: bad address family %d\n",
1377 						dst->sin_family);
1378 			dst->sin_family = AF_INET;
1379 		}
1380 #endif
1381 		if_simloop(ifp, copym, dst->sin_family, 0);
1382 	}
1383 }
1384