xref: /freebsd/sys/netinet/ip_output.c (revision c8dfaf382fa6df9dc6fd1e1c3356e0c8bf607e6a)
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 	 * Both in the SMP world, pre-emption world if_transmit() world,
450 	 * the following code doesn't really function as intended any further.
451 	 *
452 	 * + There can and will be multiple CPUs running this code path
453 	 *   in parallel, and we do no lock holding when checking the
454 	 *   queue depth;
455 	 * + And since other threads can be running concurrently, even if
456 	 *   we do pass this check, another thread may queue some frames
457 	 *   before this thread does and it will end up partially or fully
458 	 *   failing to send anyway;
459 	 * + if_transmit() based drivers don't necessarily set ifq_len
460 	 *   at all.
461 	 *
462 	 * This should be replaced with a method of pushing an entire list
463 	 * of fragment frames to the driver and have the driver decide
464 	 * whether it can queue or not queue the entire set.
465 	 */
466 #if 0
467 	/*
468 	 * Verify that we have any chance at all of being able to queue the
469 	 * packet or packet fragments, unless ALTQ is enabled on the given
470 	 * interface in which case packetdrop should be done by queueing.
471 	 */
472 	n = ip_len / mtu + 1; /* how many fragments ? */
473 	if (
474 #ifdef ALTQ
475 	    (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
476 #endif /* ALTQ */
477 	    (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
478 		error = ENOBUFS;
479 		IPSTAT_INC(ips_odropped);
480 		ifp->if_snd.ifq_drops += n;
481 		goto bad;
482 	}
483 #endif
484 
485 	/*
486 	 * Look for broadcast address and
487 	 * verify user is allowed to send
488 	 * such a packet.
489 	 */
490 	if (isbroadcast) {
491 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
492 			error = EADDRNOTAVAIL;
493 			goto bad;
494 		}
495 		if ((flags & IP_ALLOWBROADCAST) == 0) {
496 			error = EACCES;
497 			goto bad;
498 		}
499 		/* don't allow broadcast messages to be fragmented */
500 		if (ip_len > mtu) {
501 			error = EMSGSIZE;
502 			goto bad;
503 		}
504 		m->m_flags |= M_BCAST;
505 	} else {
506 		m->m_flags &= ~M_BCAST;
507 	}
508 
509 sendit:
510 #ifdef IPSEC
511 	switch(ip_ipsec_output(&m, inp, &flags, &error)) {
512 	case 1:
513 		goto bad;
514 	case -1:
515 		goto done;
516 	case 0:
517 	default:
518 		break;	/* Continue with packet processing. */
519 	}
520 	/*
521 	 * Check if there was a route for this packet; return error if not.
522 	 */
523 	if (no_route_but_check_spd) {
524 		IPSTAT_INC(ips_noroute);
525 		error = EHOSTUNREACH;
526 		goto bad;
527 	}
528 	/* Update variables that are affected by ipsec4_output(). */
529 	ip = mtod(m, struct ip *);
530 	hlen = ip->ip_hl << 2;
531 #endif /* IPSEC */
532 
533 	/* Jump over all PFIL processing if hooks are not active. */
534 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
535 		goto passout;
536 
537 	/* Run through list of hooks for output packets. */
538 	odst.s_addr = ip->ip_dst.s_addr;
539 	error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
540 	if (error != 0 || m == NULL)
541 		goto done;
542 
543 	ip = mtod(m, struct ip *);
544 
545 	/* See if destination IP address was changed by packet filter. */
546 	if (odst.s_addr != ip->ip_dst.s_addr) {
547 		m->m_flags |= M_SKIP_FIREWALL;
548 		/* If destination is now ourself drop to ip_input(). */
549 		if (in_localip(ip->ip_dst)) {
550 			m->m_flags |= M_FASTFWD_OURS;
551 			if (m->m_pkthdr.rcvif == NULL)
552 				m->m_pkthdr.rcvif = V_loif;
553 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
554 				m->m_pkthdr.csum_flags |=
555 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
556 				m->m_pkthdr.csum_data = 0xffff;
557 			}
558 			m->m_pkthdr.csum_flags |=
559 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
560 #ifdef SCTP
561 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
562 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
563 #endif
564 			error = netisr_queue(NETISR_IP, m);
565 			goto done;
566 		} else {
567 			if (have_ia_ref)
568 				ifa_free(&ia->ia_ifa);
569 			goto again;	/* Redo the routing table lookup. */
570 		}
571 	}
572 
573 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
574 	if (m->m_flags & M_FASTFWD_OURS) {
575 		if (m->m_pkthdr.rcvif == NULL)
576 			m->m_pkthdr.rcvif = V_loif;
577 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
578 			m->m_pkthdr.csum_flags |=
579 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
580 			m->m_pkthdr.csum_data = 0xffff;
581 		}
582 #ifdef SCTP
583 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
584 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
585 #endif
586 		m->m_pkthdr.csum_flags |=
587 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
588 
589 		error = netisr_queue(NETISR_IP, m);
590 		goto done;
591 	}
592 	/* Or forward to some other address? */
593 	if ((m->m_flags & M_IP_NEXTHOP) &&
594 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
595 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
596 		m->m_flags |= M_SKIP_FIREWALL;
597 		m->m_flags &= ~M_IP_NEXTHOP;
598 		m_tag_delete(m, fwd_tag);
599 		if (have_ia_ref)
600 			ifa_free(&ia->ia_ifa);
601 		goto again;
602 	}
603 
604 passout:
605 	/* 127/8 must not appear on wire - RFC1122. */
606 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
607 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
608 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
609 			IPSTAT_INC(ips_badaddr);
610 			error = EADDRNOTAVAIL;
611 			goto bad;
612 		}
613 	}
614 
615 	m->m_pkthdr.csum_flags |= CSUM_IP;
616 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
617 		in_delayed_cksum(m);
618 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
619 	}
620 #ifdef SCTP
621 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
622 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
623 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
624 	}
625 #endif
626 
627 	/*
628 	 * If small enough for interface, or the interface will take
629 	 * care of the fragmentation for us, we can just send directly.
630 	 */
631 	if (ip_len <= mtu ||
632 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
633 		ip->ip_sum = 0;
634 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
635 			ip->ip_sum = in_cksum(m, hlen);
636 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
637 		}
638 
639 		/*
640 		 * Record statistics for this interface address.
641 		 * With CSUM_TSO the byte/packet count will be slightly
642 		 * incorrect because we count the IP+TCP headers only
643 		 * once instead of for every generated packet.
644 		 */
645 		if (!(flags & IP_FORWARDING) && ia) {
646 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
647 				counter_u64_add(ia->ia_ifa.ifa_opackets,
648 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
649 			else
650 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
651 
652 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
653 		}
654 #ifdef MBUF_STRESS_TEST
655 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
656 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
657 #endif
658 		/*
659 		 * Reset layer specific mbuf flags
660 		 * to avoid confusing lower layers.
661 		 */
662 		m_clrprotoflags(m);
663 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
664 		error = (*ifp->if_output)(ifp, m,
665 		    (const struct sockaddr *)gw, ro);
666 		goto done;
667 	}
668 
669 	/* Balk when DF bit is set or the interface didn't support TSO. */
670 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
671 		error = EMSGSIZE;
672 		IPSTAT_INC(ips_cantfrag);
673 		goto bad;
674 	}
675 
676 	/*
677 	 * Too large for interface; fragment if possible. If successful,
678 	 * on return, m will point to a list of packets to be sent.
679 	 */
680 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
681 	if (error)
682 		goto bad;
683 	for (; m; m = m0) {
684 		m0 = m->m_nextpkt;
685 		m->m_nextpkt = 0;
686 		if (error == 0) {
687 			/* Record statistics for this interface address. */
688 			if (ia != NULL) {
689 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
690 				counter_u64_add(ia->ia_ifa.ifa_obytes,
691 				    m->m_pkthdr.len);
692 			}
693 			/*
694 			 * Reset layer specific mbuf flags
695 			 * to avoid confusing upper layers.
696 			 */
697 			m_clrprotoflags(m);
698 
699 			IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
700 			error = (*ifp->if_output)(ifp, m,
701 			    (const struct sockaddr *)gw, ro);
702 		} else
703 			m_freem(m);
704 	}
705 
706 	if (error == 0)
707 		IPSTAT_INC(ips_fragmented);
708 
709 done:
710 	if (ro == &iproute)
711 		RO_RTFREE(ro);
712 	if (have_ia_ref)
713 		ifa_free(&ia->ia_ifa);
714 	return (error);
715 bad:
716 	m_freem(m);
717 	goto done;
718 }
719 
720 /*
721  * Create a chain of fragments which fit the given mtu. m_frag points to the
722  * mbuf to be fragmented; on return it points to the chain with the fragments.
723  * Return 0 if no error. If error, m_frag may contain a partially built
724  * chain of fragments that should be freed by the caller.
725  *
726  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
727  */
728 int
729 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
730     u_long if_hwassist_flags)
731 {
732 	int error = 0;
733 	int hlen = ip->ip_hl << 2;
734 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
735 	int off;
736 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
737 	int firstlen;
738 	struct mbuf **mnext;
739 	int nfrags;
740 	uint16_t ip_len, ip_off;
741 
742 	ip_len = ntohs(ip->ip_len);
743 	ip_off = ntohs(ip->ip_off);
744 
745 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
746 		IPSTAT_INC(ips_cantfrag);
747 		return EMSGSIZE;
748 	}
749 
750 	/*
751 	 * Must be able to put at least 8 bytes per fragment.
752 	 */
753 	if (len < 8)
754 		return EMSGSIZE;
755 
756 	/*
757 	 * If the interface will not calculate checksums on
758 	 * fragmented packets, then do it here.
759 	 */
760 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
761 		in_delayed_cksum(m0);
762 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
763 	}
764 #ifdef SCTP
765 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
766 		sctp_delayed_cksum(m0, hlen);
767 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
768 	}
769 #endif
770 	if (len > PAGE_SIZE) {
771 		/*
772 		 * Fragment large datagrams such that each segment
773 		 * contains a multiple of PAGE_SIZE amount of data,
774 		 * plus headers. This enables a receiver to perform
775 		 * page-flipping zero-copy optimizations.
776 		 *
777 		 * XXX When does this help given that sender and receiver
778 		 * could have different page sizes, and also mtu could
779 		 * be less than the receiver's page size ?
780 		 */
781 		int newlen;
782 		struct mbuf *m;
783 
784 		for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
785 			off += m->m_len;
786 
787 		/*
788 		 * firstlen (off - hlen) must be aligned on an
789 		 * 8-byte boundary
790 		 */
791 		if (off < hlen)
792 			goto smart_frag_failure;
793 		off = ((off - hlen) & ~7) + hlen;
794 		newlen = (~PAGE_MASK) & mtu;
795 		if ((newlen + sizeof (struct ip)) > mtu) {
796 			/* we failed, go back the default */
797 smart_frag_failure:
798 			newlen = len;
799 			off = hlen + len;
800 		}
801 		len = newlen;
802 
803 	} else {
804 		off = hlen + len;
805 	}
806 
807 	firstlen = off - hlen;
808 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
809 
810 	/*
811 	 * Loop through length of segment after first fragment,
812 	 * make new header and copy data of each part and link onto chain.
813 	 * Here, m0 is the original packet, m is the fragment being created.
814 	 * The fragments are linked off the m_nextpkt of the original
815 	 * packet, which after processing serves as the first fragment.
816 	 */
817 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
818 		struct ip *mhip;	/* ip header on the fragment */
819 		struct mbuf *m;
820 		int mhlen = sizeof (struct ip);
821 
822 		m = m_gethdr(M_NOWAIT, MT_DATA);
823 		if (m == NULL) {
824 			error = ENOBUFS;
825 			IPSTAT_INC(ips_odropped);
826 			goto done;
827 		}
828 		m->m_flags |= (m0->m_flags & M_MCAST);
829 		/*
830 		 * In the first mbuf, leave room for the link header, then
831 		 * copy the original IP header including options. The payload
832 		 * goes into an additional mbuf chain returned by m_copym().
833 		 */
834 		m->m_data += max_linkhdr;
835 		mhip = mtod(m, struct ip *);
836 		*mhip = *ip;
837 		if (hlen > sizeof (struct ip)) {
838 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
839 			mhip->ip_v = IPVERSION;
840 			mhip->ip_hl = mhlen >> 2;
841 		}
842 		m->m_len = mhlen;
843 		/* XXX do we need to add ip_off below ? */
844 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
845 		if (off + len >= ip_len)
846 			len = ip_len - off;
847 		else
848 			mhip->ip_off |= IP_MF;
849 		mhip->ip_len = htons((u_short)(len + mhlen));
850 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
851 		if (m->m_next == NULL) {	/* copy failed */
852 			m_free(m);
853 			error = ENOBUFS;	/* ??? */
854 			IPSTAT_INC(ips_odropped);
855 			goto done;
856 		}
857 		m->m_pkthdr.len = mhlen + len;
858 		m->m_pkthdr.rcvif = NULL;
859 #ifdef MAC
860 		mac_netinet_fragment(m0, m);
861 #endif
862 		m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
863 		mhip->ip_off = htons(mhip->ip_off);
864 		mhip->ip_sum = 0;
865 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
866 			mhip->ip_sum = in_cksum(m, mhlen);
867 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
868 		}
869 		*mnext = m;
870 		mnext = &m->m_nextpkt;
871 	}
872 	IPSTAT_ADD(ips_ofragments, nfrags);
873 
874 	/*
875 	 * Update first fragment by trimming what's been copied out
876 	 * and updating header.
877 	 */
878 	m_adj(m0, hlen + firstlen - ip_len);
879 	m0->m_pkthdr.len = hlen + firstlen;
880 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
881 	ip->ip_off = htons(ip_off | IP_MF);
882 	ip->ip_sum = 0;
883 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
884 		ip->ip_sum = in_cksum(m0, hlen);
885 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
886 	}
887 
888 done:
889 	*m_frag = m0;
890 	return error;
891 }
892 
893 void
894 in_delayed_cksum(struct mbuf *m)
895 {
896 	struct ip *ip;
897 	uint16_t csum, offset, ip_len;
898 
899 	ip = mtod(m, struct ip *);
900 	offset = ip->ip_hl << 2 ;
901 	ip_len = ntohs(ip->ip_len);
902 	csum = in_cksum_skip(m, ip_len, offset);
903 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
904 		csum = 0xffff;
905 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
906 
907 	/* find the mbuf in the chain where the checksum starts*/
908 	while ((m != NULL) && (offset >= m->m_len)) {
909 		offset -= m->m_len;
910 		m = m->m_next;
911 	}
912 	KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
913 	KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
914 	*(u_short *)(m->m_data + offset) = csum;
915 }
916 
917 /*
918  * IP socket option processing.
919  */
920 int
921 ip_ctloutput(struct socket *so, struct sockopt *sopt)
922 {
923 	struct	inpcb *inp = sotoinpcb(so);
924 	int	error, optval;
925 #ifdef	RSS
926 	uint32_t rss_bucket;
927 	int retval;
928 #endif
929 
930 	error = optval = 0;
931 	if (sopt->sopt_level != IPPROTO_IP) {
932 		error = EINVAL;
933 
934 		if (sopt->sopt_level == SOL_SOCKET &&
935 		    sopt->sopt_dir == SOPT_SET) {
936 			switch (sopt->sopt_name) {
937 			case SO_REUSEADDR:
938 				INP_WLOCK(inp);
939 				if ((so->so_options & SO_REUSEADDR) != 0)
940 					inp->inp_flags2 |= INP_REUSEADDR;
941 				else
942 					inp->inp_flags2 &= ~INP_REUSEADDR;
943 				INP_WUNLOCK(inp);
944 				error = 0;
945 				break;
946 			case SO_REUSEPORT:
947 				INP_WLOCK(inp);
948 				if ((so->so_options & SO_REUSEPORT) != 0)
949 					inp->inp_flags2 |= INP_REUSEPORT;
950 				else
951 					inp->inp_flags2 &= ~INP_REUSEPORT;
952 				INP_WUNLOCK(inp);
953 				error = 0;
954 				break;
955 			case SO_SETFIB:
956 				INP_WLOCK(inp);
957 				inp->inp_inc.inc_fibnum = so->so_fibnum;
958 				INP_WUNLOCK(inp);
959 				error = 0;
960 				break;
961 			default:
962 				break;
963 			}
964 		}
965 		return (error);
966 	}
967 
968 	switch (sopt->sopt_dir) {
969 	case SOPT_SET:
970 		switch (sopt->sopt_name) {
971 		case IP_OPTIONS:
972 #ifdef notyet
973 		case IP_RETOPTS:
974 #endif
975 		{
976 			struct mbuf *m;
977 			if (sopt->sopt_valsize > MLEN) {
978 				error = EMSGSIZE;
979 				break;
980 			}
981 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
982 			if (m == NULL) {
983 				error = ENOBUFS;
984 				break;
985 			}
986 			m->m_len = sopt->sopt_valsize;
987 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
988 					    m->m_len);
989 			if (error) {
990 				m_free(m);
991 				break;
992 			}
993 			INP_WLOCK(inp);
994 			error = ip_pcbopts(inp, sopt->sopt_name, m);
995 			INP_WUNLOCK(inp);
996 			return (error);
997 		}
998 
999 		case IP_BINDANY:
1000 			if (sopt->sopt_td != NULL) {
1001 				error = priv_check(sopt->sopt_td,
1002 				    PRIV_NETINET_BINDANY);
1003 				if (error)
1004 					break;
1005 			}
1006 			/* FALLTHROUGH */
1007 		case IP_BINDMULTI:
1008 #ifdef	RSS
1009 		case IP_RSS_LISTEN_BUCKET:
1010 #endif
1011 		case IP_TOS:
1012 		case IP_TTL:
1013 		case IP_MINTTL:
1014 		case IP_RECVOPTS:
1015 		case IP_RECVRETOPTS:
1016 		case IP_RECVDSTADDR:
1017 		case IP_RECVTTL:
1018 		case IP_RECVIF:
1019 		case IP_FAITH:
1020 		case IP_ONESBCAST:
1021 		case IP_DONTFRAG:
1022 		case IP_RECVTOS:
1023 		case IP_RECVFLOWID:
1024 #ifdef	RSS
1025 		case IP_RECVRSSBUCKETID:
1026 #endif
1027 			error = sooptcopyin(sopt, &optval, sizeof optval,
1028 					    sizeof optval);
1029 			if (error)
1030 				break;
1031 
1032 			switch (sopt->sopt_name) {
1033 			case IP_TOS:
1034 				inp->inp_ip_tos = optval;
1035 				break;
1036 
1037 			case IP_TTL:
1038 				inp->inp_ip_ttl = optval;
1039 				break;
1040 
1041 			case IP_MINTTL:
1042 				if (optval >= 0 && optval <= MAXTTL)
1043 					inp->inp_ip_minttl = optval;
1044 				else
1045 					error = EINVAL;
1046 				break;
1047 
1048 #define	OPTSET(bit) do {						\
1049 	INP_WLOCK(inp);							\
1050 	if (optval)							\
1051 		inp->inp_flags |= bit;					\
1052 	else								\
1053 		inp->inp_flags &= ~bit;					\
1054 	INP_WUNLOCK(inp);						\
1055 } while (0)
1056 
1057 #define	OPTSET2(bit, val) do {						\
1058 	INP_WLOCK(inp);							\
1059 	if (val)							\
1060 		inp->inp_flags2 |= bit;					\
1061 	else								\
1062 		inp->inp_flags2 &= ~bit;				\
1063 	INP_WUNLOCK(inp);						\
1064 } while (0)
1065 
1066 			case IP_RECVOPTS:
1067 				OPTSET(INP_RECVOPTS);
1068 				break;
1069 
1070 			case IP_RECVRETOPTS:
1071 				OPTSET(INP_RECVRETOPTS);
1072 				break;
1073 
1074 			case IP_RECVDSTADDR:
1075 				OPTSET(INP_RECVDSTADDR);
1076 				break;
1077 
1078 			case IP_RECVTTL:
1079 				OPTSET(INP_RECVTTL);
1080 				break;
1081 
1082 			case IP_RECVIF:
1083 				OPTSET(INP_RECVIF);
1084 				break;
1085 
1086 			case IP_FAITH:
1087 				OPTSET(INP_FAITH);
1088 				break;
1089 
1090 			case IP_ONESBCAST:
1091 				OPTSET(INP_ONESBCAST);
1092 				break;
1093 			case IP_DONTFRAG:
1094 				OPTSET(INP_DONTFRAG);
1095 				break;
1096 			case IP_BINDANY:
1097 				OPTSET(INP_BINDANY);
1098 				break;
1099 			case IP_RECVTOS:
1100 				OPTSET(INP_RECVTOS);
1101 				break;
1102 			case IP_BINDMULTI:
1103 				OPTSET2(INP_BINDMULTI, optval);
1104 				break;
1105 			case IP_RECVFLOWID:
1106 				OPTSET2(INP_RECVFLOWID, optval);
1107 				break;
1108 #ifdef	RSS
1109 			case IP_RSS_LISTEN_BUCKET:
1110 				if ((optval >= 0) &&
1111 				    (optval < rss_getnumbuckets())) {
1112 					inp->inp_rss_listen_bucket = optval;
1113 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1114 				} else {
1115 					error = EINVAL;
1116 				}
1117 				break;
1118 			case IP_RECVRSSBUCKETID:
1119 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1120 				break;
1121 #endif
1122 			}
1123 			break;
1124 #undef OPTSET
1125 #undef OPTSET2
1126 
1127 		/*
1128 		 * Multicast socket options are processed by the in_mcast
1129 		 * module.
1130 		 */
1131 		case IP_MULTICAST_IF:
1132 		case IP_MULTICAST_VIF:
1133 		case IP_MULTICAST_TTL:
1134 		case IP_MULTICAST_LOOP:
1135 		case IP_ADD_MEMBERSHIP:
1136 		case IP_DROP_MEMBERSHIP:
1137 		case IP_ADD_SOURCE_MEMBERSHIP:
1138 		case IP_DROP_SOURCE_MEMBERSHIP:
1139 		case IP_BLOCK_SOURCE:
1140 		case IP_UNBLOCK_SOURCE:
1141 		case IP_MSFILTER:
1142 		case MCAST_JOIN_GROUP:
1143 		case MCAST_LEAVE_GROUP:
1144 		case MCAST_JOIN_SOURCE_GROUP:
1145 		case MCAST_LEAVE_SOURCE_GROUP:
1146 		case MCAST_BLOCK_SOURCE:
1147 		case MCAST_UNBLOCK_SOURCE:
1148 			error = inp_setmoptions(inp, sopt);
1149 			break;
1150 
1151 		case IP_PORTRANGE:
1152 			error = sooptcopyin(sopt, &optval, sizeof optval,
1153 					    sizeof optval);
1154 			if (error)
1155 				break;
1156 
1157 			INP_WLOCK(inp);
1158 			switch (optval) {
1159 			case IP_PORTRANGE_DEFAULT:
1160 				inp->inp_flags &= ~(INP_LOWPORT);
1161 				inp->inp_flags &= ~(INP_HIGHPORT);
1162 				break;
1163 
1164 			case IP_PORTRANGE_HIGH:
1165 				inp->inp_flags &= ~(INP_LOWPORT);
1166 				inp->inp_flags |= INP_HIGHPORT;
1167 				break;
1168 
1169 			case IP_PORTRANGE_LOW:
1170 				inp->inp_flags &= ~(INP_HIGHPORT);
1171 				inp->inp_flags |= INP_LOWPORT;
1172 				break;
1173 
1174 			default:
1175 				error = EINVAL;
1176 				break;
1177 			}
1178 			INP_WUNLOCK(inp);
1179 			break;
1180 
1181 #ifdef IPSEC
1182 		case IP_IPSEC_POLICY:
1183 		{
1184 			caddr_t req;
1185 			struct mbuf *m;
1186 
1187 			if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1188 				break;
1189 			if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1190 				break;
1191 			req = mtod(m, caddr_t);
1192 			error = ipsec_set_policy(inp, sopt->sopt_name, req,
1193 			    m->m_len, (sopt->sopt_td != NULL) ?
1194 			    sopt->sopt_td->td_ucred : NULL);
1195 			m_freem(m);
1196 			break;
1197 		}
1198 #endif /* IPSEC */
1199 
1200 		default:
1201 			error = ENOPROTOOPT;
1202 			break;
1203 		}
1204 		break;
1205 
1206 	case SOPT_GET:
1207 		switch (sopt->sopt_name) {
1208 		case IP_OPTIONS:
1209 		case IP_RETOPTS:
1210 			if (inp->inp_options)
1211 				error = sooptcopyout(sopt,
1212 						     mtod(inp->inp_options,
1213 							  char *),
1214 						     inp->inp_options->m_len);
1215 			else
1216 				sopt->sopt_valsize = 0;
1217 			break;
1218 
1219 		case IP_TOS:
1220 		case IP_TTL:
1221 		case IP_MINTTL:
1222 		case IP_RECVOPTS:
1223 		case IP_RECVRETOPTS:
1224 		case IP_RECVDSTADDR:
1225 		case IP_RECVTTL:
1226 		case IP_RECVIF:
1227 		case IP_PORTRANGE:
1228 		case IP_FAITH:
1229 		case IP_ONESBCAST:
1230 		case IP_DONTFRAG:
1231 		case IP_BINDANY:
1232 		case IP_RECVTOS:
1233 		case IP_BINDMULTI:
1234 		case IP_FLOWID:
1235 		case IP_FLOWTYPE:
1236 		case IP_RECVFLOWID:
1237 #ifdef	RSS
1238 		case IP_RSSBUCKETID:
1239 		case IP_RECVRSSBUCKETID:
1240 #endif
1241 			switch (sopt->sopt_name) {
1242 
1243 			case IP_TOS:
1244 				optval = inp->inp_ip_tos;
1245 				break;
1246 
1247 			case IP_TTL:
1248 				optval = inp->inp_ip_ttl;
1249 				break;
1250 
1251 			case IP_MINTTL:
1252 				optval = inp->inp_ip_minttl;
1253 				break;
1254 
1255 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1256 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1257 
1258 			case IP_RECVOPTS:
1259 				optval = OPTBIT(INP_RECVOPTS);
1260 				break;
1261 
1262 			case IP_RECVRETOPTS:
1263 				optval = OPTBIT(INP_RECVRETOPTS);
1264 				break;
1265 
1266 			case IP_RECVDSTADDR:
1267 				optval = OPTBIT(INP_RECVDSTADDR);
1268 				break;
1269 
1270 			case IP_RECVTTL:
1271 				optval = OPTBIT(INP_RECVTTL);
1272 				break;
1273 
1274 			case IP_RECVIF:
1275 				optval = OPTBIT(INP_RECVIF);
1276 				break;
1277 
1278 			case IP_PORTRANGE:
1279 				if (inp->inp_flags & INP_HIGHPORT)
1280 					optval = IP_PORTRANGE_HIGH;
1281 				else if (inp->inp_flags & INP_LOWPORT)
1282 					optval = IP_PORTRANGE_LOW;
1283 				else
1284 					optval = 0;
1285 				break;
1286 
1287 			case IP_FAITH:
1288 				optval = OPTBIT(INP_FAITH);
1289 				break;
1290 
1291 			case IP_ONESBCAST:
1292 				optval = OPTBIT(INP_ONESBCAST);
1293 				break;
1294 			case IP_DONTFRAG:
1295 				optval = OPTBIT(INP_DONTFRAG);
1296 				break;
1297 			case IP_BINDANY:
1298 				optval = OPTBIT(INP_BINDANY);
1299 				break;
1300 			case IP_RECVTOS:
1301 				optval = OPTBIT(INP_RECVTOS);
1302 				break;
1303 			case IP_FLOWID:
1304 				optval = inp->inp_flowid;
1305 				break;
1306 			case IP_FLOWTYPE:
1307 				optval = inp->inp_flowtype;
1308 				break;
1309 			case IP_RECVFLOWID:
1310 				optval = OPTBIT2(INP_RECVFLOWID);
1311 				break;
1312 #ifdef	RSS
1313 			case IP_RSSBUCKETID:
1314 				retval = rss_hash2bucket(inp->inp_flowid,
1315 				    inp->inp_flowtype,
1316 				    &rss_bucket);
1317 				if (retval == 0)
1318 					optval = rss_bucket;
1319 				else
1320 					error = EINVAL;
1321 				break;
1322 			case IP_RECVRSSBUCKETID:
1323 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1324 				break;
1325 #endif
1326 			case IP_BINDMULTI:
1327 				optval = OPTBIT2(INP_BINDMULTI);
1328 				break;
1329 			}
1330 			error = sooptcopyout(sopt, &optval, sizeof optval);
1331 			break;
1332 
1333 		/*
1334 		 * Multicast socket options are processed by the in_mcast
1335 		 * module.
1336 		 */
1337 		case IP_MULTICAST_IF:
1338 		case IP_MULTICAST_VIF:
1339 		case IP_MULTICAST_TTL:
1340 		case IP_MULTICAST_LOOP:
1341 		case IP_MSFILTER:
1342 			error = inp_getmoptions(inp, sopt);
1343 			break;
1344 
1345 #ifdef IPSEC
1346 		case IP_IPSEC_POLICY:
1347 		{
1348 			struct mbuf *m = NULL;
1349 			caddr_t req = NULL;
1350 			size_t len = 0;
1351 
1352 			if (m != 0) {
1353 				req = mtod(m, caddr_t);
1354 				len = m->m_len;
1355 			}
1356 			error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1357 			if (error == 0)
1358 				error = soopt_mcopyout(sopt, m); /* XXX */
1359 			if (error == 0)
1360 				m_freem(m);
1361 			break;
1362 		}
1363 #endif /* IPSEC */
1364 
1365 		default:
1366 			error = ENOPROTOOPT;
1367 			break;
1368 		}
1369 		break;
1370 	}
1371 	return (error);
1372 }
1373 
1374 /*
1375  * Routine called from ip_output() to loop back a copy of an IP multicast
1376  * packet to the input queue of a specified interface.  Note that this
1377  * calls the output routine of the loopback "driver", but with an interface
1378  * pointer that might NOT be a loopback interface -- evil, but easier than
1379  * replicating that code here.
1380  */
1381 static void
1382 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1383     int hlen)
1384 {
1385 	register struct ip *ip;
1386 	struct mbuf *copym;
1387 
1388 	/*
1389 	 * Make a deep copy of the packet because we're going to
1390 	 * modify the pack in order to generate checksums.
1391 	 */
1392 	copym = m_dup(m, M_NOWAIT);
1393 	if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1394 		copym = m_pullup(copym, hlen);
1395 	if (copym != NULL) {
1396 		/* If needed, compute the checksum and mark it as valid. */
1397 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1398 			in_delayed_cksum(copym);
1399 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1400 			copym->m_pkthdr.csum_flags |=
1401 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1402 			copym->m_pkthdr.csum_data = 0xffff;
1403 		}
1404 		/*
1405 		 * We don't bother to fragment if the IP length is greater
1406 		 * than the interface's MTU.  Can this possibly matter?
1407 		 */
1408 		ip = mtod(copym, struct ip *);
1409 		ip->ip_sum = 0;
1410 		ip->ip_sum = in_cksum(copym, hlen);
1411 #if 1 /* XXX */
1412 		if (dst->sin_family != AF_INET) {
1413 			printf("ip_mloopback: bad address family %d\n",
1414 						dst->sin_family);
1415 			dst->sin_family = AF_INET;
1416 		}
1417 #endif
1418 		if_simloop(ifp, copym, dst->sin_family, 0);
1419 	}
1420 }
1421