xref: /freebsd/sys/netinet/ip_output.c (revision 4ec234c813eed05c166859bba82c882e40826eb9)
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_ipfw.h"
36 #include "opt_ipsec.h"
37 #include "opt_mbuf_stress_test.h"
38 #include "opt_mpath.h"
39 #include "opt_route.h"
40 #include "opt_sctp.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/malloc.h>
46 #include <sys/mbuf.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/protosw.h>
50 #include <sys/sdt.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/sysctl.h>
54 #include <sys/ucred.h>
55 
56 #include <net/if.h>
57 #include <net/if_var.h>
58 #include <net/if_llatbl.h>
59 #include <net/netisr.h>
60 #include <net/pfil.h>
61 #include <net/route.h>
62 #include <net/flowtable.h>
63 #ifdef RADIX_MPATH
64 #include <net/radix_mpath.h>
65 #endif
66 #include <net/vnet.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_kdtrace.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/ip.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/in_var.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_options.h>
76 #ifdef SCTP
77 #include <netinet/sctp.h>
78 #include <netinet/sctp_crc32.h>
79 #endif
80 
81 #ifdef IPSEC
82 #include <netinet/ip_ipsec.h>
83 #include <netipsec/ipsec.h>
84 #endif /* IPSEC*/
85 
86 #include <machine/in_cksum.h>
87 
88 #include <security/mac/mac_framework.h>
89 
90 VNET_DEFINE(u_short, ip_id);
91 
92 #ifdef MBUF_STRESS_TEST
93 static int mbuf_frag_size = 0;
94 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
95 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
96 #endif
97 
98 static void	ip_mloopback
99 	(struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
100 
101 
102 extern int in_mcast_loop;
103 extern	struct protosw inetsw[];
104 
105 /*
106  * IP output.  The packet in mbuf chain m contains a skeletal IP
107  * header (with len, off, ttl, proto, tos, src, dst).
108  * The mbuf chain containing the packet will be freed.
109  * The mbuf opt, if present, will not be freed.
110  * If route ro is present and has ro_rt initialized, route lookup would be
111  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
112  * then result of route lookup is stored in ro->ro_rt.
113  *
114  * In the IP forwarding case, the packet will arrive with options already
115  * inserted, so must have a NULL opt pointer.
116  */
117 int
118 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
119     struct ip_moptions *imo, struct inpcb *inp)
120 {
121 	struct ip *ip;
122 	struct ifnet *ifp = NULL;	/* keep compiler happy */
123 	struct mbuf *m0;
124 	int hlen = sizeof (struct ip);
125 	int mtu;
126 #if 0
127 	int n;	/* scratchpad */
128 #endif
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 #ifdef IPSEC
140 	int no_route_but_check_spd = 0;
141 #endif
142 	M_ASSERTPKTHDR(m);
143 
144 	if (inp != NULL) {
145 		INP_LOCK_ASSERT(inp);
146 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
147 		if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
148 			m->m_pkthdr.flowid = inp->inp_flowid;
149 			m->m_flags |= M_FLOWID;
150 		}
151 	}
152 
153 	if (ro == NULL) {
154 		ro = &iproute;
155 		bzero(ro, sizeof (*ro));
156 	}
157 
158 #ifdef FLOWTABLE
159 	if (ro->ro_rt == NULL) {
160 		struct flentry *fle;
161 
162 		/*
163 		 * The flow table returns route entries valid for up to 30
164 		 * seconds; we rely on the remainder of ip_output() taking no
165 		 * longer than that long for the stability of ro_rt. The
166 		 * flow ID assignment must have happened before this point.
167 		 */
168 		fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET);
169 		if (fle != NULL)
170 			flow_to_route(fle, ro);
171 	}
172 #endif
173 
174 	if (opt) {
175 		int len = 0;
176 		m = ip_insertoptions(m, opt, &len);
177 		if (len != 0)
178 			hlen = len; /* ip->ip_hl is updated above */
179 	}
180 	ip = mtod(m, struct ip *);
181 	ip_len = ntohs(ip->ip_len);
182 	ip_off = ntohs(ip->ip_off);
183 
184 	/*
185 	 * Fill in IP header.  If we are not allowing fragmentation,
186 	 * then the ip_id field is meaningless, but we don't set it
187 	 * to zero.  Doing so causes various problems when devices along
188 	 * the path (routers, load balancers, firewalls, etc.) illegally
189 	 * disable DF on our packet.  Note that a 16-bit counter
190 	 * will wrap around in less than 10 seconds at 100 Mbit/s on a
191 	 * medium with MTU 1500.  See Steven M. Bellovin, "A Technique
192 	 * for Counting NATted Hosts", Proc. IMW'02, available at
193 	 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
194 	 */
195 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
196 		ip->ip_v = IPVERSION;
197 		ip->ip_hl = hlen >> 2;
198 		ip->ip_id = ip_newid();
199 		IPSTAT_INC(ips_localout);
200 	} else {
201 		/* Header already set, fetch hlen from there */
202 		hlen = ip->ip_hl << 2;
203 	}
204 
205 	gw = dst = (struct sockaddr_in *)&ro->ro_dst;
206 again:
207 	ia = NULL;
208 	/*
209 	 * If there is a cached route,
210 	 * check that it is to the same destination
211 	 * and is still up.  If not, free it and try again.
212 	 * The address family should also be checked in case of sharing the
213 	 * cache with IPv6.
214 	 */
215 	rte = ro->ro_rt;
216 	if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
217 		    rte->rt_ifp == NULL ||
218 		    !RT_LINK_IS_UP(rte->rt_ifp) ||
219 			  dst->sin_family != AF_INET ||
220 			  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
221 		RO_RTFREE(ro);
222 		ro->ro_lle = NULL;
223 		rte = NULL;
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)))) == NULL &&
239 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
240 			IPSTAT_INC(ips_noroute);
241 			error = ENETUNREACH;
242 			goto bad;
243 		}
244 		ip->ip_dst.s_addr = INADDR_BROADCAST;
245 		dst->sin_addr = ip->ip_dst;
246 		ifp = ia->ia_ifp;
247 		ip->ip_ttl = 1;
248 		isbroadcast = 1;
249 	} else if (flags & IP_ROUTETOIF) {
250 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
251 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) {
252 			IPSTAT_INC(ips_noroute);
253 			error = ENETUNREACH;
254 			goto bad;
255 		}
256 		ifp = ia->ia_ifp;
257 		ip->ip_ttl = 1;
258 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
259 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
260 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
261 		/*
262 		 * Bypass the normal routing lookup for multicast
263 		 * packets if the interface is specified.
264 		 */
265 		ifp = imo->imo_multicast_ifp;
266 		IFP_TO_IA(ifp, ia);
267 		isbroadcast = 0;	/* fool gcc */
268 	} else {
269 		/*
270 		 * We want to do any cloning requested by the link layer,
271 		 * as this is probably required in all cases for correct
272 		 * operation (as it is for ARP).
273 		 */
274 		if (rte == NULL) {
275 #ifdef RADIX_MPATH
276 			rtalloc_mpath_fib(ro,
277 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
278 			    inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
279 #else
280 			in_rtalloc_ign(ro, 0,
281 			    inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
282 #endif
283 			rte = ro->ro_rt;
284 		}
285 		if (rte == NULL ||
286 		    rte->rt_ifp == NULL ||
287 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
288 #ifdef IPSEC
289 			/*
290 			 * There is no route for this packet, but it is
291 			 * possible that a matching SPD entry exists.
292 			 */
293 			no_route_but_check_spd = 1;
294 			mtu = 0; /* Silence GCC warning. */
295 			goto sendit;
296 #endif
297 			IPSTAT_INC(ips_noroute);
298 			error = EHOSTUNREACH;
299 			goto bad;
300 		}
301 		ia = ifatoia(rte->rt_ifa);
302 		ifa_ref(&ia->ia_ifa);
303 		ifp = rte->rt_ifp;
304 		rte->rt_rmx.rmx_pksent++;
305 		if (rte->rt_flags & RTF_GATEWAY)
306 			gw = (struct sockaddr_in *)rte->rt_gateway;
307 		if (rte->rt_flags & RTF_HOST)
308 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
309 		else
310 			isbroadcast = in_broadcast(gw->sin_addr, ifp);
311 	}
312 	/*
313 	 * Calculate MTU.  If we have a route that is up, use that,
314 	 * otherwise use the interface's MTU.
315 	 */
316 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
317 		/*
318 		 * This case can happen if the user changed the MTU
319 		 * of an interface after enabling IP on it.  Because
320 		 * most netifs don't keep track of routes pointing to
321 		 * them, there is no way for one to update all its
322 		 * routes when the MTU is changed.
323 		 */
324 		if (rte->rt_rmx.rmx_mtu > ifp->if_mtu)
325 			rte->rt_rmx.rmx_mtu = ifp->if_mtu;
326 		mtu = rte->rt_rmx.rmx_mtu;
327 	} else {
328 		mtu = ifp->if_mtu;
329 	}
330 	/* Catch a possible divide by zero later. */
331 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
332 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
333 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
334 		m->m_flags |= M_MCAST;
335 		/*
336 		 * IP destination address is multicast.  Make sure "gw"
337 		 * still points to the address in "ro".  (It may have been
338 		 * changed to point to a gateway address, above.)
339 		 */
340 		gw = dst;
341 		/*
342 		 * See if the caller provided any multicast options
343 		 */
344 		if (imo != NULL) {
345 			ip->ip_ttl = imo->imo_multicast_ttl;
346 			if (imo->imo_multicast_vif != -1)
347 				ip->ip_src.s_addr =
348 				    ip_mcast_src ?
349 				    ip_mcast_src(imo->imo_multicast_vif) :
350 				    INADDR_ANY;
351 		} else
352 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
353 		/*
354 		 * Confirm that the outgoing interface supports multicast.
355 		 */
356 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
357 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
358 				IPSTAT_INC(ips_noroute);
359 				error = ENETUNREACH;
360 				goto bad;
361 			}
362 		}
363 		/*
364 		 * If source address not specified yet, use address
365 		 * of outgoing interface.
366 		 */
367 		if (ip->ip_src.s_addr == INADDR_ANY) {
368 			/* Interface may have no addresses. */
369 			if (ia != NULL)
370 				ip->ip_src = IA_SIN(ia)->sin_addr;
371 		}
372 
373 		if ((imo == NULL && in_mcast_loop) ||
374 		    (imo && imo->imo_multicast_loop)) {
375 			/*
376 			 * Loop back multicast datagram if not expressly
377 			 * forbidden to do so, even if we are not a member
378 			 * of the group; ip_input() will filter it later,
379 			 * thus deferring a hash lookup and mutex acquisition
380 			 * at the expense of a cheap copy using m_copym().
381 			 */
382 			ip_mloopback(ifp, m, dst, hlen);
383 		} else {
384 			/*
385 			 * If we are acting as a multicast router, perform
386 			 * multicast forwarding as if the packet had just
387 			 * arrived on the interface to which we are about
388 			 * to send.  The multicast forwarding function
389 			 * recursively calls this function, using the
390 			 * IP_FORWARDING flag to prevent infinite recursion.
391 			 *
392 			 * Multicasts that are looped back by ip_mloopback(),
393 			 * above, will be forwarded by the ip_input() routine,
394 			 * if necessary.
395 			 */
396 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
397 				/*
398 				 * If rsvp daemon is not running, do not
399 				 * set ip_moptions. This ensures that the packet
400 				 * is multicast and not just sent down one link
401 				 * as prescribed by rsvpd.
402 				 */
403 				if (!V_rsvp_on)
404 					imo = NULL;
405 				if (ip_mforward &&
406 				    ip_mforward(ip, ifp, m, imo) != 0) {
407 					m_freem(m);
408 					goto done;
409 				}
410 			}
411 		}
412 
413 		/*
414 		 * Multicasts with a time-to-live of zero may be looped-
415 		 * back, above, but must not be transmitted on a network.
416 		 * Also, multicasts addressed to the loopback interface
417 		 * are not sent -- the above call to ip_mloopback() will
418 		 * loop back a copy. ip_input() will drop the copy if
419 		 * this host does not belong to the destination group on
420 		 * the loopback interface.
421 		 */
422 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
423 			m_freem(m);
424 			goto done;
425 		}
426 
427 		goto sendit;
428 	}
429 
430 	/*
431 	 * If the source address is not specified yet, use the address
432 	 * of the outoing interface.
433 	 */
434 	if (ip->ip_src.s_addr == INADDR_ANY) {
435 		/* Interface may have no addresses. */
436 		if (ia != NULL) {
437 			ip->ip_src = IA_SIN(ia)->sin_addr;
438 		}
439 	}
440 
441 	/*
442 	 * Both in the SMP world, pre-emption world if_transmit() world,
443 	 * the following code doesn't really function as intended any further.
444 	 *
445 	 * + There can and will be multiple CPUs running this code path
446 	 *   in parallel, and we do no lock holding when checking the
447 	 *   queue depth;
448 	 * + And since other threads can be running concurrently, even if
449 	 *   we do pass this check, another thread may queue some frames
450 	 *   before this thread does and it will end up partially or fully
451 	 *   failing to send anyway;
452 	 * + if_transmit() based drivers don't necessarily set ifq_len
453 	 *   at all.
454 	 *
455 	 * This should be replaced with a method of pushing an entire list
456 	 * of fragment frames to the driver and have the driver decide
457 	 * whether it can queue or not queue the entire set.
458 	 */
459 #if 0
460 	/*
461 	 * Verify that we have any chance at all of being able to queue the
462 	 * packet or packet fragments, unless ALTQ is enabled on the given
463 	 * interface in which case packetdrop should be done by queueing.
464 	 */
465 	n = ip_len / mtu + 1; /* how many fragments ? */
466 	if (
467 #ifdef ALTQ
468 	    (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
469 #endif /* ALTQ */
470 	    (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
471 		error = ENOBUFS;
472 		IPSTAT_INC(ips_odropped);
473 		ifp->if_snd.ifq_drops += n;
474 		goto bad;
475 	}
476 #endif
477 
478 	/*
479 	 * Look for broadcast address and
480 	 * verify user is allowed to send
481 	 * such a packet.
482 	 */
483 	if (isbroadcast) {
484 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
485 			error = EADDRNOTAVAIL;
486 			goto bad;
487 		}
488 		if ((flags & IP_ALLOWBROADCAST) == 0) {
489 			error = EACCES;
490 			goto bad;
491 		}
492 		/* don't allow broadcast messages to be fragmented */
493 		if (ip_len > mtu) {
494 			error = EMSGSIZE;
495 			goto bad;
496 		}
497 		m->m_flags |= M_BCAST;
498 	} else {
499 		m->m_flags &= ~M_BCAST;
500 	}
501 
502 sendit:
503 #ifdef IPSEC
504 	switch(ip_ipsec_output(&m, inp, &flags, &error)) {
505 	case 1:
506 		goto bad;
507 	case -1:
508 		goto done;
509 	case 0:
510 	default:
511 		break;	/* Continue with packet processing. */
512 	}
513 	/*
514 	 * Check if there was a route for this packet; return error if not.
515 	 */
516 	if (no_route_but_check_spd) {
517 		IPSTAT_INC(ips_noroute);
518 		error = EHOSTUNREACH;
519 		goto bad;
520 	}
521 	/* Update variables that are affected by ipsec4_output(). */
522 	ip = mtod(m, struct ip *);
523 	hlen = ip->ip_hl << 2;
524 #endif /* IPSEC */
525 
526 	/* Jump over all PFIL processing if hooks are not active. */
527 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
528 		goto passout;
529 
530 	/* Run through list of hooks for output packets. */
531 	odst.s_addr = ip->ip_dst.s_addr;
532 	error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
533 	if (error != 0 || m == NULL)
534 		goto done;
535 
536 	ip = mtod(m, struct ip *);
537 
538 	/* See if destination IP address was changed by packet filter. */
539 	if (odst.s_addr != ip->ip_dst.s_addr) {
540 		m->m_flags |= M_SKIP_FIREWALL;
541 		/* If destination is now ourself drop to ip_input(). */
542 		if (in_localip(ip->ip_dst)) {
543 			m->m_flags |= M_FASTFWD_OURS;
544 			if (m->m_pkthdr.rcvif == NULL)
545 				m->m_pkthdr.rcvif = V_loif;
546 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
547 				m->m_pkthdr.csum_flags |=
548 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
549 				m->m_pkthdr.csum_data = 0xffff;
550 			}
551 			m->m_pkthdr.csum_flags |=
552 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
553 #ifdef SCTP
554 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
555 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
556 #endif
557 			error = netisr_queue(NETISR_IP, m);
558 			goto done;
559 		} else {
560 			if (ia != NULL)
561 				ifa_free(&ia->ia_ifa);
562 			goto again;	/* Redo the routing table lookup. */
563 		}
564 	}
565 
566 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
567 	if (m->m_flags & M_FASTFWD_OURS) {
568 		if (m->m_pkthdr.rcvif == NULL)
569 			m->m_pkthdr.rcvif = V_loif;
570 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
571 			m->m_pkthdr.csum_flags |=
572 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
573 			m->m_pkthdr.csum_data = 0xffff;
574 		}
575 #ifdef SCTP
576 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
577 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
578 #endif
579 		m->m_pkthdr.csum_flags |=
580 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
581 
582 		error = netisr_queue(NETISR_IP, m);
583 		goto done;
584 	}
585 	/* Or forward to some other address? */
586 	if ((m->m_flags & M_IP_NEXTHOP) &&
587 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
588 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
589 		m->m_flags |= M_SKIP_FIREWALL;
590 		m->m_flags &= ~M_IP_NEXTHOP;
591 		m_tag_delete(m, fwd_tag);
592 		if (ia != NULL)
593 			ifa_free(&ia->ia_ifa);
594 		goto again;
595 	}
596 
597 passout:
598 	/* 127/8 must not appear on wire - RFC1122. */
599 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
600 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
601 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
602 			IPSTAT_INC(ips_badaddr);
603 			error = EADDRNOTAVAIL;
604 			goto bad;
605 		}
606 	}
607 
608 	m->m_pkthdr.csum_flags |= CSUM_IP;
609 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
610 		in_delayed_cksum(m);
611 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
612 	}
613 #ifdef SCTP
614 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
615 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
616 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
617 	}
618 #endif
619 
620 	/*
621 	 * If small enough for interface, or the interface will take
622 	 * care of the fragmentation for us, we can just send directly.
623 	 */
624 	if (ip_len <= mtu ||
625 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
626 	    ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
627 		ip->ip_sum = 0;
628 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
629 			ip->ip_sum = in_cksum(m, hlen);
630 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
631 		}
632 
633 		/*
634 		 * Record statistics for this interface address.
635 		 * With CSUM_TSO the byte/packet count will be slightly
636 		 * incorrect because we count the IP+TCP headers only
637 		 * once instead of for every generated packet.
638 		 */
639 		if (!(flags & IP_FORWARDING) && ia) {
640 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
641 				counter_u64_add(ia->ia_ifa.ifa_opackets,
642 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
643 			else
644 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
645 
646 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
647 		}
648 #ifdef MBUF_STRESS_TEST
649 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
650 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
651 #endif
652 		/*
653 		 * Reset layer specific mbuf flags
654 		 * to avoid confusing lower layers.
655 		 */
656 		m_clrprotoflags(m);
657 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
658 		error = (*ifp->if_output)(ifp, m,
659 		    (const struct sockaddr *)gw, ro);
660 		goto done;
661 	}
662 
663 	/* Balk when DF bit is set or the interface didn't support TSO. */
664 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
665 		error = EMSGSIZE;
666 		IPSTAT_INC(ips_cantfrag);
667 		goto bad;
668 	}
669 
670 	/*
671 	 * Too large for interface; fragment if possible. If successful,
672 	 * on return, m will point to a list of packets to be sent.
673 	 */
674 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
675 	if (error)
676 		goto bad;
677 	for (; m; m = m0) {
678 		m0 = m->m_nextpkt;
679 		m->m_nextpkt = 0;
680 		if (error == 0) {
681 			/* Record statistics for this interface address. */
682 			if (ia != NULL) {
683 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
684 				counter_u64_add(ia->ia_ifa.ifa_obytes,
685 				    m->m_pkthdr.len);
686 			}
687 			/*
688 			 * Reset layer specific mbuf flags
689 			 * to avoid confusing upper layers.
690 			 */
691 			m_clrprotoflags(m);
692 
693 			IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
694 			error = (*ifp->if_output)(ifp, m,
695 			    (const struct sockaddr *)gw, ro);
696 		} else
697 			m_freem(m);
698 	}
699 
700 	if (error == 0)
701 		IPSTAT_INC(ips_fragmented);
702 
703 done:
704 	if (ro == &iproute)
705 		RO_RTFREE(ro);
706 	if (ia != NULL)
707 		ifa_free(&ia->ia_ifa);
708 	return (error);
709 bad:
710 	m_freem(m);
711 	goto done;
712 }
713 
714 /*
715  * Create a chain of fragments which fit the given mtu. m_frag points to the
716  * mbuf to be fragmented; on return it points to the chain with the fragments.
717  * Return 0 if no error. If error, m_frag may contain a partially built
718  * chain of fragments that should be freed by the caller.
719  *
720  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
721  */
722 int
723 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
724     u_long if_hwassist_flags)
725 {
726 	int error = 0;
727 	int hlen = ip->ip_hl << 2;
728 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
729 	int off;
730 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
731 	int firstlen;
732 	struct mbuf **mnext;
733 	int nfrags;
734 	uint16_t ip_len, ip_off;
735 
736 	ip_len = ntohs(ip->ip_len);
737 	ip_off = ntohs(ip->ip_off);
738 
739 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
740 		IPSTAT_INC(ips_cantfrag);
741 		return EMSGSIZE;
742 	}
743 
744 	/*
745 	 * Must be able to put at least 8 bytes per fragment.
746 	 */
747 	if (len < 8)
748 		return EMSGSIZE;
749 
750 	/*
751 	 * If the interface will not calculate checksums on
752 	 * fragmented packets, then do it here.
753 	 */
754 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
755 		in_delayed_cksum(m0);
756 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
757 	}
758 #ifdef SCTP
759 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
760 		sctp_delayed_cksum(m0, hlen);
761 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
762 	}
763 #endif
764 	if (len > PAGE_SIZE) {
765 		/*
766 		 * Fragment large datagrams such that each segment
767 		 * contains a multiple of PAGE_SIZE amount of data,
768 		 * plus headers. This enables a receiver to perform
769 		 * page-flipping zero-copy optimizations.
770 		 *
771 		 * XXX When does this help given that sender and receiver
772 		 * could have different page sizes, and also mtu could
773 		 * be less than the receiver's page size ?
774 		 */
775 		int newlen;
776 		struct mbuf *m;
777 
778 		for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
779 			off += m->m_len;
780 
781 		/*
782 		 * firstlen (off - hlen) must be aligned on an
783 		 * 8-byte boundary
784 		 */
785 		if (off < hlen)
786 			goto smart_frag_failure;
787 		off = ((off - hlen) & ~7) + hlen;
788 		newlen = (~PAGE_MASK) & mtu;
789 		if ((newlen + sizeof (struct ip)) > mtu) {
790 			/* we failed, go back the default */
791 smart_frag_failure:
792 			newlen = len;
793 			off = hlen + len;
794 		}
795 		len = newlen;
796 
797 	} else {
798 		off = hlen + len;
799 	}
800 
801 	firstlen = off - hlen;
802 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
803 
804 	/*
805 	 * Loop through length of segment after first fragment,
806 	 * make new header and copy data of each part and link onto chain.
807 	 * Here, m0 is the original packet, m is the fragment being created.
808 	 * The fragments are linked off the m_nextpkt of the original
809 	 * packet, which after processing serves as the first fragment.
810 	 */
811 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
812 		struct ip *mhip;	/* ip header on the fragment */
813 		struct mbuf *m;
814 		int mhlen = sizeof (struct ip);
815 
816 		m = m_gethdr(M_NOWAIT, MT_DATA);
817 		if (m == NULL) {
818 			error = ENOBUFS;
819 			IPSTAT_INC(ips_odropped);
820 			goto done;
821 		}
822 		m->m_flags |= (m0->m_flags & M_MCAST);
823 		/*
824 		 * In the first mbuf, leave room for the link header, then
825 		 * copy the original IP header including options. The payload
826 		 * goes into an additional mbuf chain returned by m_copym().
827 		 */
828 		m->m_data += max_linkhdr;
829 		mhip = mtod(m, struct ip *);
830 		*mhip = *ip;
831 		if (hlen > sizeof (struct ip)) {
832 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
833 			mhip->ip_v = IPVERSION;
834 			mhip->ip_hl = mhlen >> 2;
835 		}
836 		m->m_len = mhlen;
837 		/* XXX do we need to add ip_off below ? */
838 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
839 		if (off + len >= ip_len)
840 			len = ip_len - off;
841 		else
842 			mhip->ip_off |= IP_MF;
843 		mhip->ip_len = htons((u_short)(len + mhlen));
844 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
845 		if (m->m_next == NULL) {	/* copy failed */
846 			m_free(m);
847 			error = ENOBUFS;	/* ??? */
848 			IPSTAT_INC(ips_odropped);
849 			goto done;
850 		}
851 		m->m_pkthdr.len = mhlen + len;
852 		m->m_pkthdr.rcvif = NULL;
853 #ifdef MAC
854 		mac_netinet_fragment(m0, m);
855 #endif
856 		m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
857 		mhip->ip_off = htons(mhip->ip_off);
858 		mhip->ip_sum = 0;
859 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
860 			mhip->ip_sum = in_cksum(m, mhlen);
861 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
862 		}
863 		*mnext = m;
864 		mnext = &m->m_nextpkt;
865 	}
866 	IPSTAT_ADD(ips_ofragments, nfrags);
867 
868 	/*
869 	 * Update first fragment by trimming what's been copied out
870 	 * and updating header.
871 	 */
872 	m_adj(m0, hlen + firstlen - ip_len);
873 	m0->m_pkthdr.len = hlen + firstlen;
874 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
875 	ip->ip_off = htons(ip_off | IP_MF);
876 	ip->ip_sum = 0;
877 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
878 		ip->ip_sum = in_cksum(m0, hlen);
879 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
880 	}
881 
882 done:
883 	*m_frag = m0;
884 	return error;
885 }
886 
887 void
888 in_delayed_cksum(struct mbuf *m)
889 {
890 	struct ip *ip;
891 	uint16_t csum, offset, ip_len;
892 
893 	ip = mtod(m, struct ip *);
894 	offset = ip->ip_hl << 2 ;
895 	ip_len = ntohs(ip->ip_len);
896 	csum = in_cksum_skip(m, ip_len, offset);
897 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
898 		csum = 0xffff;
899 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
900 
901 	if (offset + sizeof(u_short) > m->m_len) {
902 		printf("delayed m_pullup, m->len: %d  off: %d  p: %d\n",
903 		    m->m_len, offset, ip->ip_p);
904 		/*
905 		 * XXX
906 		 * this shouldn't happen, but if it does, the
907 		 * correct behavior may be to insert the checksum
908 		 * in the appropriate next mbuf in the chain.
909 		 */
910 		return;
911 	}
912 	*(u_short *)(m->m_data + offset) = csum;
913 }
914 
915 /*
916  * IP socket option processing.
917  */
918 int
919 ip_ctloutput(struct socket *so, struct sockopt *sopt)
920 {
921 	struct	inpcb *inp = sotoinpcb(so);
922 	int	error, optval;
923 
924 	error = optval = 0;
925 	if (sopt->sopt_level != IPPROTO_IP) {
926 		error = EINVAL;
927 
928 		if (sopt->sopt_level == SOL_SOCKET &&
929 		    sopt->sopt_dir == SOPT_SET) {
930 			switch (sopt->sopt_name) {
931 			case SO_REUSEADDR:
932 				INP_WLOCK(inp);
933 				if ((so->so_options & SO_REUSEADDR) != 0)
934 					inp->inp_flags2 |= INP_REUSEADDR;
935 				else
936 					inp->inp_flags2 &= ~INP_REUSEADDR;
937 				INP_WUNLOCK(inp);
938 				error = 0;
939 				break;
940 			case SO_REUSEPORT:
941 				INP_WLOCK(inp);
942 				if ((so->so_options & SO_REUSEPORT) != 0)
943 					inp->inp_flags2 |= INP_REUSEPORT;
944 				else
945 					inp->inp_flags2 &= ~INP_REUSEPORT;
946 				INP_WUNLOCK(inp);
947 				error = 0;
948 				break;
949 			case SO_SETFIB:
950 				INP_WLOCK(inp);
951 				inp->inp_inc.inc_fibnum = so->so_fibnum;
952 				INP_WUNLOCK(inp);
953 				error = 0;
954 				break;
955 			default:
956 				break;
957 			}
958 		}
959 		return (error);
960 	}
961 
962 	switch (sopt->sopt_dir) {
963 	case SOPT_SET:
964 		switch (sopt->sopt_name) {
965 		case IP_OPTIONS:
966 #ifdef notyet
967 		case IP_RETOPTS:
968 #endif
969 		{
970 			struct mbuf *m;
971 			if (sopt->sopt_valsize > MLEN) {
972 				error = EMSGSIZE;
973 				break;
974 			}
975 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
976 			if (m == NULL) {
977 				error = ENOBUFS;
978 				break;
979 			}
980 			m->m_len = sopt->sopt_valsize;
981 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
982 					    m->m_len);
983 			if (error) {
984 				m_free(m);
985 				break;
986 			}
987 			INP_WLOCK(inp);
988 			error = ip_pcbopts(inp, sopt->sopt_name, m);
989 			INP_WUNLOCK(inp);
990 			return (error);
991 		}
992 
993 		case IP_BINDANY:
994 			if (sopt->sopt_td != NULL) {
995 				error = priv_check(sopt->sopt_td,
996 				    PRIV_NETINET_BINDANY);
997 				if (error)
998 					break;
999 			}
1000 			/* FALLTHROUGH */
1001 		case IP_TOS:
1002 		case IP_TTL:
1003 		case IP_MINTTL:
1004 		case IP_RECVOPTS:
1005 		case IP_RECVRETOPTS:
1006 		case IP_RECVDSTADDR:
1007 		case IP_RECVTTL:
1008 		case IP_RECVIF:
1009 		case IP_FAITH:
1010 		case IP_ONESBCAST:
1011 		case IP_DONTFRAG:
1012 		case IP_RECVTOS:
1013 			error = sooptcopyin(sopt, &optval, sizeof optval,
1014 					    sizeof optval);
1015 			if (error)
1016 				break;
1017 
1018 			switch (sopt->sopt_name) {
1019 			case IP_TOS:
1020 				inp->inp_ip_tos = optval;
1021 				break;
1022 
1023 			case IP_TTL:
1024 				inp->inp_ip_ttl = optval;
1025 				break;
1026 
1027 			case IP_MINTTL:
1028 				if (optval >= 0 && optval <= MAXTTL)
1029 					inp->inp_ip_minttl = optval;
1030 				else
1031 					error = EINVAL;
1032 				break;
1033 
1034 #define	OPTSET(bit) do {						\
1035 	INP_WLOCK(inp);							\
1036 	if (optval)							\
1037 		inp->inp_flags |= bit;					\
1038 	else								\
1039 		inp->inp_flags &= ~bit;					\
1040 	INP_WUNLOCK(inp);						\
1041 } while (0)
1042 
1043 			case IP_RECVOPTS:
1044 				OPTSET(INP_RECVOPTS);
1045 				break;
1046 
1047 			case IP_RECVRETOPTS:
1048 				OPTSET(INP_RECVRETOPTS);
1049 				break;
1050 
1051 			case IP_RECVDSTADDR:
1052 				OPTSET(INP_RECVDSTADDR);
1053 				break;
1054 
1055 			case IP_RECVTTL:
1056 				OPTSET(INP_RECVTTL);
1057 				break;
1058 
1059 			case IP_RECVIF:
1060 				OPTSET(INP_RECVIF);
1061 				break;
1062 
1063 			case IP_FAITH:
1064 				OPTSET(INP_FAITH);
1065 				break;
1066 
1067 			case IP_ONESBCAST:
1068 				OPTSET(INP_ONESBCAST);
1069 				break;
1070 			case IP_DONTFRAG:
1071 				OPTSET(INP_DONTFRAG);
1072 				break;
1073 			case IP_BINDANY:
1074 				OPTSET(INP_BINDANY);
1075 				break;
1076 			case IP_RECVTOS:
1077 				OPTSET(INP_RECVTOS);
1078 				break;
1079 			}
1080 			break;
1081 #undef OPTSET
1082 
1083 		/*
1084 		 * Multicast socket options are processed by the in_mcast
1085 		 * module.
1086 		 */
1087 		case IP_MULTICAST_IF:
1088 		case IP_MULTICAST_VIF:
1089 		case IP_MULTICAST_TTL:
1090 		case IP_MULTICAST_LOOP:
1091 		case IP_ADD_MEMBERSHIP:
1092 		case IP_DROP_MEMBERSHIP:
1093 		case IP_ADD_SOURCE_MEMBERSHIP:
1094 		case IP_DROP_SOURCE_MEMBERSHIP:
1095 		case IP_BLOCK_SOURCE:
1096 		case IP_UNBLOCK_SOURCE:
1097 		case IP_MSFILTER:
1098 		case MCAST_JOIN_GROUP:
1099 		case MCAST_LEAVE_GROUP:
1100 		case MCAST_JOIN_SOURCE_GROUP:
1101 		case MCAST_LEAVE_SOURCE_GROUP:
1102 		case MCAST_BLOCK_SOURCE:
1103 		case MCAST_UNBLOCK_SOURCE:
1104 			error = inp_setmoptions(inp, sopt);
1105 			break;
1106 
1107 		case IP_PORTRANGE:
1108 			error = sooptcopyin(sopt, &optval, sizeof optval,
1109 					    sizeof optval);
1110 			if (error)
1111 				break;
1112 
1113 			INP_WLOCK(inp);
1114 			switch (optval) {
1115 			case IP_PORTRANGE_DEFAULT:
1116 				inp->inp_flags &= ~(INP_LOWPORT);
1117 				inp->inp_flags &= ~(INP_HIGHPORT);
1118 				break;
1119 
1120 			case IP_PORTRANGE_HIGH:
1121 				inp->inp_flags &= ~(INP_LOWPORT);
1122 				inp->inp_flags |= INP_HIGHPORT;
1123 				break;
1124 
1125 			case IP_PORTRANGE_LOW:
1126 				inp->inp_flags &= ~(INP_HIGHPORT);
1127 				inp->inp_flags |= INP_LOWPORT;
1128 				break;
1129 
1130 			default:
1131 				error = EINVAL;
1132 				break;
1133 			}
1134 			INP_WUNLOCK(inp);
1135 			break;
1136 
1137 #ifdef IPSEC
1138 		case IP_IPSEC_POLICY:
1139 		{
1140 			caddr_t req;
1141 			struct mbuf *m;
1142 
1143 			if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1144 				break;
1145 			if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1146 				break;
1147 			req = mtod(m, caddr_t);
1148 			error = ipsec_set_policy(inp, sopt->sopt_name, req,
1149 			    m->m_len, (sopt->sopt_td != NULL) ?
1150 			    sopt->sopt_td->td_ucred : NULL);
1151 			m_freem(m);
1152 			break;
1153 		}
1154 #endif /* IPSEC */
1155 
1156 		default:
1157 			error = ENOPROTOOPT;
1158 			break;
1159 		}
1160 		break;
1161 
1162 	case SOPT_GET:
1163 		switch (sopt->sopt_name) {
1164 		case IP_OPTIONS:
1165 		case IP_RETOPTS:
1166 			if (inp->inp_options)
1167 				error = sooptcopyout(sopt,
1168 						     mtod(inp->inp_options,
1169 							  char *),
1170 						     inp->inp_options->m_len);
1171 			else
1172 				sopt->sopt_valsize = 0;
1173 			break;
1174 
1175 		case IP_TOS:
1176 		case IP_TTL:
1177 		case IP_MINTTL:
1178 		case IP_RECVOPTS:
1179 		case IP_RECVRETOPTS:
1180 		case IP_RECVDSTADDR:
1181 		case IP_RECVTTL:
1182 		case IP_RECVIF:
1183 		case IP_PORTRANGE:
1184 		case IP_FAITH:
1185 		case IP_ONESBCAST:
1186 		case IP_DONTFRAG:
1187 		case IP_BINDANY:
1188 		case IP_RECVTOS:
1189 			switch (sopt->sopt_name) {
1190 
1191 			case IP_TOS:
1192 				optval = inp->inp_ip_tos;
1193 				break;
1194 
1195 			case IP_TTL:
1196 				optval = inp->inp_ip_ttl;
1197 				break;
1198 
1199 			case IP_MINTTL:
1200 				optval = inp->inp_ip_minttl;
1201 				break;
1202 
1203 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1204 
1205 			case IP_RECVOPTS:
1206 				optval = OPTBIT(INP_RECVOPTS);
1207 				break;
1208 
1209 			case IP_RECVRETOPTS:
1210 				optval = OPTBIT(INP_RECVRETOPTS);
1211 				break;
1212 
1213 			case IP_RECVDSTADDR:
1214 				optval = OPTBIT(INP_RECVDSTADDR);
1215 				break;
1216 
1217 			case IP_RECVTTL:
1218 				optval = OPTBIT(INP_RECVTTL);
1219 				break;
1220 
1221 			case IP_RECVIF:
1222 				optval = OPTBIT(INP_RECVIF);
1223 				break;
1224 
1225 			case IP_PORTRANGE:
1226 				if (inp->inp_flags & INP_HIGHPORT)
1227 					optval = IP_PORTRANGE_HIGH;
1228 				else if (inp->inp_flags & INP_LOWPORT)
1229 					optval = IP_PORTRANGE_LOW;
1230 				else
1231 					optval = 0;
1232 				break;
1233 
1234 			case IP_FAITH:
1235 				optval = OPTBIT(INP_FAITH);
1236 				break;
1237 
1238 			case IP_ONESBCAST:
1239 				optval = OPTBIT(INP_ONESBCAST);
1240 				break;
1241 			case IP_DONTFRAG:
1242 				optval = OPTBIT(INP_DONTFRAG);
1243 				break;
1244 			case IP_BINDANY:
1245 				optval = OPTBIT(INP_BINDANY);
1246 				break;
1247 			case IP_RECVTOS:
1248 				optval = OPTBIT(INP_RECVTOS);
1249 				break;
1250 			}
1251 			error = sooptcopyout(sopt, &optval, sizeof optval);
1252 			break;
1253 
1254 		/*
1255 		 * Multicast socket options are processed by the in_mcast
1256 		 * module.
1257 		 */
1258 		case IP_MULTICAST_IF:
1259 		case IP_MULTICAST_VIF:
1260 		case IP_MULTICAST_TTL:
1261 		case IP_MULTICAST_LOOP:
1262 		case IP_MSFILTER:
1263 			error = inp_getmoptions(inp, sopt);
1264 			break;
1265 
1266 #ifdef IPSEC
1267 		case IP_IPSEC_POLICY:
1268 		{
1269 			struct mbuf *m = NULL;
1270 			caddr_t req = NULL;
1271 			size_t len = 0;
1272 
1273 			if (m != 0) {
1274 				req = mtod(m, caddr_t);
1275 				len = m->m_len;
1276 			}
1277 			error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1278 			if (error == 0)
1279 				error = soopt_mcopyout(sopt, m); /* XXX */
1280 			if (error == 0)
1281 				m_freem(m);
1282 			break;
1283 		}
1284 #endif /* IPSEC */
1285 
1286 		default:
1287 			error = ENOPROTOOPT;
1288 			break;
1289 		}
1290 		break;
1291 	}
1292 	return (error);
1293 }
1294 
1295 /*
1296  * Routine called from ip_output() to loop back a copy of an IP multicast
1297  * packet to the input queue of a specified interface.  Note that this
1298  * calls the output routine of the loopback "driver", but with an interface
1299  * pointer that might NOT be a loopback interface -- evil, but easier than
1300  * replicating that code here.
1301  */
1302 static void
1303 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1304     int hlen)
1305 {
1306 	register struct ip *ip;
1307 	struct mbuf *copym;
1308 
1309 	/*
1310 	 * Make a deep copy of the packet because we're going to
1311 	 * modify the pack in order to generate checksums.
1312 	 */
1313 	copym = m_dup(m, M_NOWAIT);
1314 	if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1315 		copym = m_pullup(copym, hlen);
1316 	if (copym != NULL) {
1317 		/* If needed, compute the checksum and mark it as valid. */
1318 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1319 			in_delayed_cksum(copym);
1320 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1321 			copym->m_pkthdr.csum_flags |=
1322 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1323 			copym->m_pkthdr.csum_data = 0xffff;
1324 		}
1325 		/*
1326 		 * We don't bother to fragment if the IP length is greater
1327 		 * than the interface's MTU.  Can this possibly matter?
1328 		 */
1329 		ip = mtod(copym, struct ip *);
1330 		ip->ip_sum = 0;
1331 		ip->ip_sum = in_cksum(copym, hlen);
1332 #if 1 /* XXX */
1333 		if (dst->sin_family != AF_INET) {
1334 			printf("ip_mloopback: bad address family %d\n",
1335 						dst->sin_family);
1336 			dst->sin_family = AF_INET;
1337 		}
1338 #endif
1339 		if_simloop(ifp, copym, dst->sin_family, 0);
1340 	}
1341 }
1342