xref: /freebsd/sys/netinet/ip_output.c (revision 4f52dfbb8d6c4d446500c5b097e3806ec219fbd4)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1990, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_inet.h"
38 #include "opt_ratelimit.h"
39 #include "opt_ipsec.h"
40 #include "opt_mbuf_stress_test.h"
41 #include "opt_mpath.h"
42 #include "opt_route.h"
43 #include "opt_sctp.h"
44 #include "opt_rss.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
49 #include <sys/lock.h>
50 #include <sys/malloc.h>
51 #include <sys/mbuf.h>
52 #include <sys/priv.h>
53 #include <sys/proc.h>
54 #include <sys/protosw.h>
55 #include <sys/rmlock.h>
56 #include <sys/sdt.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/sysctl.h>
60 #include <sys/ucred.h>
61 
62 #include <net/if.h>
63 #include <net/if_var.h>
64 #include <net/if_llatbl.h>
65 #include <net/netisr.h>
66 #include <net/pfil.h>
67 #include <net/route.h>
68 #ifdef RADIX_MPATH
69 #include <net/radix_mpath.h>
70 #endif
71 #include <net/rss_config.h>
72 #include <net/vnet.h>
73 
74 #include <netinet/in.h>
75 #include <netinet/in_kdtrace.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/ip.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/in_rss.h>
80 #include <netinet/in_var.h>
81 #include <netinet/ip_var.h>
82 #include <netinet/ip_options.h>
83 #ifdef SCTP
84 #include <netinet/sctp.h>
85 #include <netinet/sctp_crc32.h>
86 #endif
87 
88 #include <netipsec/ipsec_support.h>
89 
90 #include <machine/in_cksum.h>
91 
92 #include <security/mac/mac_framework.h>
93 
94 #ifdef MBUF_STRESS_TEST
95 static int mbuf_frag_size = 0;
96 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
97 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
98 #endif
99 
100 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
101 
102 
103 extern int in_mcast_loop;
104 extern	struct protosw inetsw[];
105 
106 static inline int
107 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
108     struct sockaddr_in *dst, int *fibnum, int *error)
109 {
110 	struct m_tag *fwd_tag = NULL;
111 	struct mbuf *m;
112 	struct in_addr odst;
113 	struct ip *ip;
114 
115 	m = *mp;
116 	ip = mtod(m, struct ip *);
117 
118 	/* Run through list of hooks for output packets. */
119 	odst.s_addr = ip->ip_dst.s_addr;
120 	*error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, 0, inp);
121 	m = *mp;
122 	if ((*error) != 0 || m == NULL)
123 		return 1; /* Finished */
124 
125 	ip = mtod(m, struct ip *);
126 
127 	/* See if destination IP address was changed by packet filter. */
128 	if (odst.s_addr != ip->ip_dst.s_addr) {
129 		m->m_flags |= M_SKIP_FIREWALL;
130 		/* If destination is now ourself drop to ip_input(). */
131 		if (in_localip(ip->ip_dst)) {
132 			m->m_flags |= M_FASTFWD_OURS;
133 			if (m->m_pkthdr.rcvif == NULL)
134 				m->m_pkthdr.rcvif = V_loif;
135 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
136 				m->m_pkthdr.csum_flags |=
137 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
138 				m->m_pkthdr.csum_data = 0xffff;
139 			}
140 			m->m_pkthdr.csum_flags |=
141 				CSUM_IP_CHECKED | CSUM_IP_VALID;
142 #ifdef SCTP
143 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
144 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
145 #endif
146 			*error = netisr_queue(NETISR_IP, m);
147 			return 1; /* Finished */
148 		}
149 
150 		bzero(dst, sizeof(*dst));
151 		dst->sin_family = AF_INET;
152 		dst->sin_len = sizeof(*dst);
153 		dst->sin_addr = ip->ip_dst;
154 
155 		return -1; /* Reloop */
156 	}
157 	/* See if fib was changed by packet filter. */
158 	if ((*fibnum) != M_GETFIB(m)) {
159 		m->m_flags |= M_SKIP_FIREWALL;
160 		*fibnum = M_GETFIB(m);
161 		return -1; /* Reloop for FIB change */
162 	}
163 
164 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
165 	if (m->m_flags & M_FASTFWD_OURS) {
166 		if (m->m_pkthdr.rcvif == NULL)
167 			m->m_pkthdr.rcvif = V_loif;
168 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
169 			m->m_pkthdr.csum_flags |=
170 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
171 			m->m_pkthdr.csum_data = 0xffff;
172 		}
173 #ifdef SCTP
174 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
175 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
176 #endif
177 		m->m_pkthdr.csum_flags |=
178 			CSUM_IP_CHECKED | CSUM_IP_VALID;
179 
180 		*error = netisr_queue(NETISR_IP, m);
181 		return 1; /* Finished */
182 	}
183 	/* Or forward to some other address? */
184 	if ((m->m_flags & M_IP_NEXTHOP) &&
185 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
186 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
187 		m->m_flags |= M_SKIP_FIREWALL;
188 		m->m_flags &= ~M_IP_NEXTHOP;
189 		m_tag_delete(m, fwd_tag);
190 
191 		return -1; /* Reloop for CHANGE of dst */
192 	}
193 
194 	return 0;
195 }
196 
197 /*
198  * IP output.  The packet in mbuf chain m contains a skeletal IP
199  * header (with len, off, ttl, proto, tos, src, dst).
200  * The mbuf chain containing the packet will be freed.
201  * The mbuf opt, if present, will not be freed.
202  * If route ro is present and has ro_rt initialized, route lookup would be
203  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
204  * then result of route lookup is stored in ro->ro_rt.
205  *
206  * In the IP forwarding case, the packet will arrive with options already
207  * inserted, so must have a NULL opt pointer.
208  */
209 int
210 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
211     struct ip_moptions *imo, struct inpcb *inp)
212 {
213 	struct rm_priotracker in_ifa_tracker;
214 	struct ip *ip;
215 	struct ifnet *ifp = NULL;	/* keep compiler happy */
216 	struct mbuf *m0;
217 	int hlen = sizeof (struct ip);
218 	int mtu;
219 	int error = 0;
220 	struct sockaddr_in *dst;
221 	const struct sockaddr_in *gw;
222 	struct in_ifaddr *ia;
223 	int isbroadcast;
224 	uint16_t ip_len, ip_off;
225 	struct route iproute;
226 	struct rtentry *rte;	/* cache for ro->ro_rt */
227 	uint32_t fibnum;
228 	int have_ia_ref;
229 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
230 	int no_route_but_check_spd = 0;
231 #endif
232 	M_ASSERTPKTHDR(m);
233 
234 	if (inp != NULL) {
235 		INP_LOCK_ASSERT(inp);
236 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
237 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
238 			m->m_pkthdr.flowid = inp->inp_flowid;
239 			M_HASHTYPE_SET(m, inp->inp_flowtype);
240 		}
241 	}
242 
243 	if (ro == NULL) {
244 		ro = &iproute;
245 		bzero(ro, sizeof (*ro));
246 	}
247 
248 	if (opt) {
249 		int len = 0;
250 		m = ip_insertoptions(m, opt, &len);
251 		if (len != 0)
252 			hlen = len; /* ip->ip_hl is updated above */
253 	}
254 	ip = mtod(m, struct ip *);
255 	ip_len = ntohs(ip->ip_len);
256 	ip_off = ntohs(ip->ip_off);
257 
258 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
259 		ip->ip_v = IPVERSION;
260 		ip->ip_hl = hlen >> 2;
261 		ip_fillid(ip);
262 		IPSTAT_INC(ips_localout);
263 	} else {
264 		/* Header already set, fetch hlen from there */
265 		hlen = ip->ip_hl << 2;
266 	}
267 
268 	/*
269 	 * dst/gw handling:
270 	 *
271 	 * dst can be rewritten but always points to &ro->ro_dst.
272 	 * gw is readonly but can point either to dst OR rt_gateway,
273 	 * therefore we need restore gw if we're redoing lookup.
274 	 */
275 	gw = dst = (struct sockaddr_in *)&ro->ro_dst;
276 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
277 	rte = ro->ro_rt;
278 	if (rte == NULL) {
279 		bzero(dst, sizeof(*dst));
280 		dst->sin_family = AF_INET;
281 		dst->sin_len = sizeof(*dst);
282 		dst->sin_addr = ip->ip_dst;
283 	}
284 again:
285 	/*
286 	 * Validate route against routing table additions;
287 	 * a better/more specific route might have been added.
288 	 */
289 	if (inp)
290 		RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
291 	/*
292 	 * If there is a cached route,
293 	 * check that it is to the same destination
294 	 * and is still up.  If not, free it and try again.
295 	 * The address family should also be checked in case of sharing the
296 	 * cache with IPv6.
297 	 * Also check whether routing cache needs invalidation.
298 	 */
299 	rte = ro->ro_rt;
300 	if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
301 		    rte->rt_ifp == NULL ||
302 		    !RT_LINK_IS_UP(rte->rt_ifp) ||
303 			  dst->sin_family != AF_INET ||
304 			  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
305 		RO_INVALIDATE_CACHE(ro);
306 		rte = NULL;
307 	}
308 	ia = NULL;
309 	have_ia_ref = 0;
310 	/*
311 	 * If routing to interface only, short circuit routing lookup.
312 	 * The use of an all-ones broadcast address implies this; an
313 	 * interface is specified by the broadcast address of an interface,
314 	 * or the destination address of a ptp interface.
315 	 */
316 	if (flags & IP_SENDONES) {
317 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
318 						      M_GETFIB(m)))) == NULL &&
319 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
320 						    M_GETFIB(m)))) == NULL) {
321 			IPSTAT_INC(ips_noroute);
322 			error = ENETUNREACH;
323 			goto bad;
324 		}
325 		have_ia_ref = 1;
326 		ip->ip_dst.s_addr = INADDR_BROADCAST;
327 		dst->sin_addr = ip->ip_dst;
328 		ifp = ia->ia_ifp;
329 		ip->ip_ttl = 1;
330 		isbroadcast = 1;
331 	} else if (flags & IP_ROUTETOIF) {
332 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
333 						    M_GETFIB(m)))) == NULL &&
334 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
335 						M_GETFIB(m)))) == NULL) {
336 			IPSTAT_INC(ips_noroute);
337 			error = ENETUNREACH;
338 			goto bad;
339 		}
340 		have_ia_ref = 1;
341 		ifp = ia->ia_ifp;
342 		ip->ip_ttl = 1;
343 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
344 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
345 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
346 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
347 		/*
348 		 * Bypass the normal routing lookup for multicast
349 		 * packets if the interface is specified.
350 		 */
351 		ifp = imo->imo_multicast_ifp;
352 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
353 		if (ia)
354 			have_ia_ref = 1;
355 		isbroadcast = 0;	/* fool gcc */
356 	} else {
357 		/*
358 		 * We want to do any cloning requested by the link layer,
359 		 * as this is probably required in all cases for correct
360 		 * operation (as it is for ARP).
361 		 */
362 		if (rte == NULL) {
363 #ifdef RADIX_MPATH
364 			rtalloc_mpath_fib(ro,
365 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
366 			    fibnum);
367 #else
368 			in_rtalloc_ign(ro, 0, fibnum);
369 #endif
370 			rte = ro->ro_rt;
371 		}
372 		if (rte == NULL ||
373 		    (rte->rt_flags & RTF_UP) == 0 ||
374 		    rte->rt_ifp == NULL ||
375 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
376 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
377 			/*
378 			 * There is no route for this packet, but it is
379 			 * possible that a matching SPD entry exists.
380 			 */
381 			no_route_but_check_spd = 1;
382 			mtu = 0; /* Silence GCC warning. */
383 			goto sendit;
384 #endif
385 			IPSTAT_INC(ips_noroute);
386 			error = EHOSTUNREACH;
387 			goto bad;
388 		}
389 		ia = ifatoia(rte->rt_ifa);
390 		ifp = rte->rt_ifp;
391 		counter_u64_add(rte->rt_pksent, 1);
392 		rt_update_ro_flags(ro);
393 		if (rte->rt_flags & RTF_GATEWAY)
394 			gw = (struct sockaddr_in *)rte->rt_gateway;
395 		if (rte->rt_flags & RTF_HOST)
396 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
397 		else if (ifp->if_flags & IFF_BROADCAST)
398 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
399 		else
400 			isbroadcast = 0;
401 	}
402 
403 	/*
404 	 * Calculate MTU.  If we have a route that is up, use that,
405 	 * otherwise use the interface's MTU.
406 	 */
407 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
408 		mtu = rte->rt_mtu;
409 	else
410 		mtu = ifp->if_mtu;
411 	/* Catch a possible divide by zero later. */
412 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
413 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
414 
415 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
416 		m->m_flags |= M_MCAST;
417 		/*
418 		 * IP destination address is multicast.  Make sure "gw"
419 		 * still points to the address in "ro".  (It may have been
420 		 * changed to point to a gateway address, above.)
421 		 */
422 		gw = dst;
423 		/*
424 		 * See if the caller provided any multicast options
425 		 */
426 		if (imo != NULL) {
427 			ip->ip_ttl = imo->imo_multicast_ttl;
428 			if (imo->imo_multicast_vif != -1)
429 				ip->ip_src.s_addr =
430 				    ip_mcast_src ?
431 				    ip_mcast_src(imo->imo_multicast_vif) :
432 				    INADDR_ANY;
433 		} else
434 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
435 		/*
436 		 * Confirm that the outgoing interface supports multicast.
437 		 */
438 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
439 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
440 				IPSTAT_INC(ips_noroute);
441 				error = ENETUNREACH;
442 				goto bad;
443 			}
444 		}
445 		/*
446 		 * If source address not specified yet, use address
447 		 * of outgoing interface.
448 		 */
449 		if (ip->ip_src.s_addr == INADDR_ANY) {
450 			/* Interface may have no addresses. */
451 			if (ia != NULL)
452 				ip->ip_src = IA_SIN(ia)->sin_addr;
453 		}
454 
455 		if ((imo == NULL && in_mcast_loop) ||
456 		    (imo && imo->imo_multicast_loop)) {
457 			/*
458 			 * Loop back multicast datagram if not expressly
459 			 * forbidden to do so, even if we are not a member
460 			 * of the group; ip_input() will filter it later,
461 			 * thus deferring a hash lookup and mutex acquisition
462 			 * at the expense of a cheap copy using m_copym().
463 			 */
464 			ip_mloopback(ifp, m, hlen);
465 		} else {
466 			/*
467 			 * If we are acting as a multicast router, perform
468 			 * multicast forwarding as if the packet had just
469 			 * arrived on the interface to which we are about
470 			 * to send.  The multicast forwarding function
471 			 * recursively calls this function, using the
472 			 * IP_FORWARDING flag to prevent infinite recursion.
473 			 *
474 			 * Multicasts that are looped back by ip_mloopback(),
475 			 * above, will be forwarded by the ip_input() routine,
476 			 * if necessary.
477 			 */
478 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
479 				/*
480 				 * If rsvp daemon is not running, do not
481 				 * set ip_moptions. This ensures that the packet
482 				 * is multicast and not just sent down one link
483 				 * as prescribed by rsvpd.
484 				 */
485 				if (!V_rsvp_on)
486 					imo = NULL;
487 				if (ip_mforward &&
488 				    ip_mforward(ip, ifp, m, imo) != 0) {
489 					m_freem(m);
490 					goto done;
491 				}
492 			}
493 		}
494 
495 		/*
496 		 * Multicasts with a time-to-live of zero may be looped-
497 		 * back, above, but must not be transmitted on a network.
498 		 * Also, multicasts addressed to the loopback interface
499 		 * are not sent -- the above call to ip_mloopback() will
500 		 * loop back a copy. ip_input() will drop the copy if
501 		 * this host does not belong to the destination group on
502 		 * the loopback interface.
503 		 */
504 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
505 			m_freem(m);
506 			goto done;
507 		}
508 
509 		goto sendit;
510 	}
511 
512 	/*
513 	 * If the source address is not specified yet, use the address
514 	 * of the outoing interface.
515 	 */
516 	if (ip->ip_src.s_addr == INADDR_ANY) {
517 		/* Interface may have no addresses. */
518 		if (ia != NULL) {
519 			ip->ip_src = IA_SIN(ia)->sin_addr;
520 		}
521 	}
522 
523 	/*
524 	 * Look for broadcast address and
525 	 * verify user is allowed to send
526 	 * such a packet.
527 	 */
528 	if (isbroadcast) {
529 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
530 			error = EADDRNOTAVAIL;
531 			goto bad;
532 		}
533 		if ((flags & IP_ALLOWBROADCAST) == 0) {
534 			error = EACCES;
535 			goto bad;
536 		}
537 		/* don't allow broadcast messages to be fragmented */
538 		if (ip_len > mtu) {
539 			error = EMSGSIZE;
540 			goto bad;
541 		}
542 		m->m_flags |= M_BCAST;
543 	} else {
544 		m->m_flags &= ~M_BCAST;
545 	}
546 
547 sendit:
548 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
549 	if (IPSEC_ENABLED(ipv4)) {
550 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
551 			if (error == EINPROGRESS)
552 				error = 0;
553 			goto done;
554 		}
555 	}
556 	/*
557 	 * Check if there was a route for this packet; return error if not.
558 	 */
559 	if (no_route_but_check_spd) {
560 		IPSTAT_INC(ips_noroute);
561 		error = EHOSTUNREACH;
562 		goto bad;
563 	}
564 	/* Update variables that are affected by ipsec4_output(). */
565 	ip = mtod(m, struct ip *);
566 	hlen = ip->ip_hl << 2;
567 #endif /* IPSEC */
568 
569 	/* Jump over all PFIL processing if hooks are not active. */
570 	if (PFIL_HOOKED(&V_inet_pfil_hook)) {
571 		switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
572 		case 1: /* Finished */
573 			goto done;
574 
575 		case 0: /* Continue normally */
576 			ip = mtod(m, struct ip *);
577 			break;
578 
579 		case -1: /* Need to try again */
580 			/* Reset everything for a new round */
581 			RO_RTFREE(ro);
582 			if (have_ia_ref)
583 				ifa_free(&ia->ia_ifa);
584 			ro->ro_prepend = NULL;
585 			rte = NULL;
586 			gw = dst;
587 			ip = mtod(m, struct ip *);
588 			goto again;
589 
590 		}
591 	}
592 
593 	/* 127/8 must not appear on wire - RFC1122. */
594 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
595 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
596 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
597 			IPSTAT_INC(ips_badaddr);
598 			error = EADDRNOTAVAIL;
599 			goto bad;
600 		}
601 	}
602 
603 	m->m_pkthdr.csum_flags |= CSUM_IP;
604 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
605 		in_delayed_cksum(m);
606 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
607 	}
608 #ifdef SCTP
609 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
610 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
611 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
612 	}
613 #endif
614 
615 	/*
616 	 * If small enough for interface, or the interface will take
617 	 * care of the fragmentation for us, we can just send directly.
618 	 */
619 	if (ip_len <= mtu ||
620 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
621 		ip->ip_sum = 0;
622 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
623 			ip->ip_sum = in_cksum(m, hlen);
624 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
625 		}
626 
627 		/*
628 		 * Record statistics for this interface address.
629 		 * With CSUM_TSO the byte/packet count will be slightly
630 		 * incorrect because we count the IP+TCP headers only
631 		 * once instead of for every generated packet.
632 		 */
633 		if (!(flags & IP_FORWARDING) && ia) {
634 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
635 				counter_u64_add(ia->ia_ifa.ifa_opackets,
636 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
637 			else
638 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
639 
640 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
641 		}
642 #ifdef MBUF_STRESS_TEST
643 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
644 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
645 #endif
646 		/*
647 		 * Reset layer specific mbuf flags
648 		 * to avoid confusing lower layers.
649 		 */
650 		m_clrprotoflags(m);
651 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
652 #ifdef RATELIMIT
653 		if (inp != NULL) {
654 			if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
655 				in_pcboutput_txrtlmt(inp, ifp, m);
656 			/* stamp send tag on mbuf */
657 			m->m_pkthdr.snd_tag = inp->inp_snd_tag;
658 		} else {
659 			m->m_pkthdr.snd_tag = NULL;
660 		}
661 #endif
662 		error = (*ifp->if_output)(ifp, m,
663 		    (const struct sockaddr *)gw, ro);
664 #ifdef RATELIMIT
665 		/* check for route change */
666 		if (error == EAGAIN)
667 			in_pcboutput_eagain(inp);
668 #endif
669 		goto done;
670 	}
671 
672 	/* Balk when DF bit is set or the interface didn't support TSO. */
673 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
674 		error = EMSGSIZE;
675 		IPSTAT_INC(ips_cantfrag);
676 		goto bad;
677 	}
678 
679 	/*
680 	 * Too large for interface; fragment if possible. If successful,
681 	 * on return, m will point to a list of packets to be sent.
682 	 */
683 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
684 	if (error)
685 		goto bad;
686 	for (; m; m = m0) {
687 		m0 = m->m_nextpkt;
688 		m->m_nextpkt = 0;
689 		if (error == 0) {
690 			/* Record statistics for this interface address. */
691 			if (ia != NULL) {
692 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
693 				counter_u64_add(ia->ia_ifa.ifa_obytes,
694 				    m->m_pkthdr.len);
695 			}
696 			/*
697 			 * Reset layer specific mbuf flags
698 			 * to avoid confusing upper layers.
699 			 */
700 			m_clrprotoflags(m);
701 
702 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
703 			    mtod(m, struct ip *), NULL);
704 #ifdef RATELIMIT
705 			if (inp != NULL) {
706 				if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
707 					in_pcboutput_txrtlmt(inp, ifp, m);
708 				/* stamp send tag on mbuf */
709 				m->m_pkthdr.snd_tag = inp->inp_snd_tag;
710 			} else {
711 				m->m_pkthdr.snd_tag = NULL;
712 			}
713 #endif
714 			error = (*ifp->if_output)(ifp, m,
715 			    (const struct sockaddr *)gw, ro);
716 #ifdef RATELIMIT
717 			/* check for route change */
718 			if (error == EAGAIN)
719 				in_pcboutput_eagain(inp);
720 #endif
721 		} else
722 			m_freem(m);
723 	}
724 
725 	if (error == 0)
726 		IPSTAT_INC(ips_fragmented);
727 
728 done:
729 	if (ro == &iproute)
730 		RO_RTFREE(ro);
731 	else if (rte == NULL)
732 		/*
733 		 * If the caller supplied a route but somehow the reference
734 		 * to it has been released need to prevent the caller
735 		 * calling RTFREE on it again.
736 		 */
737 		ro->ro_rt = NULL;
738 	if (have_ia_ref)
739 		ifa_free(&ia->ia_ifa);
740 	return (error);
741 bad:
742 	m_freem(m);
743 	goto done;
744 }
745 
746 /*
747  * Create a chain of fragments which fit the given mtu. m_frag points to the
748  * mbuf to be fragmented; on return it points to the chain with the fragments.
749  * Return 0 if no error. If error, m_frag may contain a partially built
750  * chain of fragments that should be freed by the caller.
751  *
752  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
753  */
754 int
755 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
756     u_long if_hwassist_flags)
757 {
758 	int error = 0;
759 	int hlen = ip->ip_hl << 2;
760 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
761 	int off;
762 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
763 	int firstlen;
764 	struct mbuf **mnext;
765 	int nfrags;
766 	uint16_t ip_len, ip_off;
767 
768 	ip_len = ntohs(ip->ip_len);
769 	ip_off = ntohs(ip->ip_off);
770 
771 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
772 		IPSTAT_INC(ips_cantfrag);
773 		return EMSGSIZE;
774 	}
775 
776 	/*
777 	 * Must be able to put at least 8 bytes per fragment.
778 	 */
779 	if (len < 8)
780 		return EMSGSIZE;
781 
782 	/*
783 	 * If the interface will not calculate checksums on
784 	 * fragmented packets, then do it here.
785 	 */
786 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
787 		in_delayed_cksum(m0);
788 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
789 	}
790 #ifdef SCTP
791 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
792 		sctp_delayed_cksum(m0, hlen);
793 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
794 	}
795 #endif
796 	if (len > PAGE_SIZE) {
797 		/*
798 		 * Fragment large datagrams such that each segment
799 		 * contains a multiple of PAGE_SIZE amount of data,
800 		 * plus headers. This enables a receiver to perform
801 		 * page-flipping zero-copy optimizations.
802 		 *
803 		 * XXX When does this help given that sender and receiver
804 		 * could have different page sizes, and also mtu could
805 		 * be less than the receiver's page size ?
806 		 */
807 		int newlen;
808 
809 		off = MIN(mtu, m0->m_pkthdr.len);
810 
811 		/*
812 		 * firstlen (off - hlen) must be aligned on an
813 		 * 8-byte boundary
814 		 */
815 		if (off < hlen)
816 			goto smart_frag_failure;
817 		off = ((off - hlen) & ~7) + hlen;
818 		newlen = (~PAGE_MASK) & mtu;
819 		if ((newlen + sizeof (struct ip)) > mtu) {
820 			/* we failed, go back the default */
821 smart_frag_failure:
822 			newlen = len;
823 			off = hlen + len;
824 		}
825 		len = newlen;
826 
827 	} else {
828 		off = hlen + len;
829 	}
830 
831 	firstlen = off - hlen;
832 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
833 
834 	/*
835 	 * Loop through length of segment after first fragment,
836 	 * make new header and copy data of each part and link onto chain.
837 	 * Here, m0 is the original packet, m is the fragment being created.
838 	 * The fragments are linked off the m_nextpkt of the original
839 	 * packet, which after processing serves as the first fragment.
840 	 */
841 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
842 		struct ip *mhip;	/* ip header on the fragment */
843 		struct mbuf *m;
844 		int mhlen = sizeof (struct ip);
845 
846 		m = m_gethdr(M_NOWAIT, MT_DATA);
847 		if (m == NULL) {
848 			error = ENOBUFS;
849 			IPSTAT_INC(ips_odropped);
850 			goto done;
851 		}
852 		/*
853 		 * Make sure the complete packet header gets copied
854 		 * from the originating mbuf to the newly created
855 		 * mbuf. This also ensures that existing firewall
856 		 * classification(s), VLAN tags and so on get copied
857 		 * to the resulting fragmented packet(s):
858 		 */
859 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
860 			m_free(m);
861 			error = ENOBUFS;
862 			IPSTAT_INC(ips_odropped);
863 			goto done;
864 		}
865 		/*
866 		 * In the first mbuf, leave room for the link header, then
867 		 * copy the original IP header including options. The payload
868 		 * goes into an additional mbuf chain returned by m_copym().
869 		 */
870 		m->m_data += max_linkhdr;
871 		mhip = mtod(m, struct ip *);
872 		*mhip = *ip;
873 		if (hlen > sizeof (struct ip)) {
874 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
875 			mhip->ip_v = IPVERSION;
876 			mhip->ip_hl = mhlen >> 2;
877 		}
878 		m->m_len = mhlen;
879 		/* XXX do we need to add ip_off below ? */
880 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
881 		if (off + len >= ip_len)
882 			len = ip_len - off;
883 		else
884 			mhip->ip_off |= IP_MF;
885 		mhip->ip_len = htons((u_short)(len + mhlen));
886 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
887 		if (m->m_next == NULL) {	/* copy failed */
888 			m_free(m);
889 			error = ENOBUFS;	/* ??? */
890 			IPSTAT_INC(ips_odropped);
891 			goto done;
892 		}
893 		m->m_pkthdr.len = mhlen + len;
894 #ifdef MAC
895 		mac_netinet_fragment(m0, m);
896 #endif
897 		mhip->ip_off = htons(mhip->ip_off);
898 		mhip->ip_sum = 0;
899 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
900 			mhip->ip_sum = in_cksum(m, mhlen);
901 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
902 		}
903 		*mnext = m;
904 		mnext = &m->m_nextpkt;
905 	}
906 	IPSTAT_ADD(ips_ofragments, nfrags);
907 
908 	/*
909 	 * Update first fragment by trimming what's been copied out
910 	 * and updating header.
911 	 */
912 	m_adj(m0, hlen + firstlen - ip_len);
913 	m0->m_pkthdr.len = hlen + firstlen;
914 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
915 	ip->ip_off = htons(ip_off | IP_MF);
916 	ip->ip_sum = 0;
917 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
918 		ip->ip_sum = in_cksum(m0, hlen);
919 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
920 	}
921 
922 done:
923 	*m_frag = m0;
924 	return error;
925 }
926 
927 void
928 in_delayed_cksum(struct mbuf *m)
929 {
930 	struct ip *ip;
931 	uint16_t csum, offset, ip_len;
932 
933 	ip = mtod(m, struct ip *);
934 	offset = ip->ip_hl << 2 ;
935 	ip_len = ntohs(ip->ip_len);
936 	csum = in_cksum_skip(m, ip_len, offset);
937 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
938 		csum = 0xffff;
939 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
940 
941 	/* find the mbuf in the chain where the checksum starts*/
942 	while ((m != NULL) && (offset >= m->m_len)) {
943 		offset -= m->m_len;
944 		m = m->m_next;
945 	}
946 	KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
947 	KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
948 	*(u_short *)(m->m_data + offset) = csum;
949 }
950 
951 /*
952  * IP socket option processing.
953  */
954 int
955 ip_ctloutput(struct socket *so, struct sockopt *sopt)
956 {
957 	struct	inpcb *inp = sotoinpcb(so);
958 	int	error, optval;
959 #ifdef	RSS
960 	uint32_t rss_bucket;
961 	int retval;
962 #endif
963 
964 	error = optval = 0;
965 	if (sopt->sopt_level != IPPROTO_IP) {
966 		error = EINVAL;
967 
968 		if (sopt->sopt_level == SOL_SOCKET &&
969 		    sopt->sopt_dir == SOPT_SET) {
970 			switch (sopt->sopt_name) {
971 			case SO_REUSEADDR:
972 				INP_WLOCK(inp);
973 				if ((so->so_options & SO_REUSEADDR) != 0)
974 					inp->inp_flags2 |= INP_REUSEADDR;
975 				else
976 					inp->inp_flags2 &= ~INP_REUSEADDR;
977 				INP_WUNLOCK(inp);
978 				error = 0;
979 				break;
980 			case SO_REUSEPORT:
981 				INP_WLOCK(inp);
982 				if ((so->so_options & SO_REUSEPORT) != 0)
983 					inp->inp_flags2 |= INP_REUSEPORT;
984 				else
985 					inp->inp_flags2 &= ~INP_REUSEPORT;
986 				INP_WUNLOCK(inp);
987 				error = 0;
988 				break;
989 			case SO_SETFIB:
990 				INP_WLOCK(inp);
991 				inp->inp_inc.inc_fibnum = so->so_fibnum;
992 				INP_WUNLOCK(inp);
993 				error = 0;
994 				break;
995 			case SO_MAX_PACING_RATE:
996 #ifdef RATELIMIT
997 				INP_WLOCK(inp);
998 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
999 				INP_WUNLOCK(inp);
1000 				error = 0;
1001 #else
1002 				error = EOPNOTSUPP;
1003 #endif
1004 				break;
1005 			default:
1006 				break;
1007 			}
1008 		}
1009 		return (error);
1010 	}
1011 
1012 	switch (sopt->sopt_dir) {
1013 	case SOPT_SET:
1014 		switch (sopt->sopt_name) {
1015 		case IP_OPTIONS:
1016 #ifdef notyet
1017 		case IP_RETOPTS:
1018 #endif
1019 		{
1020 			struct mbuf *m;
1021 			if (sopt->sopt_valsize > MLEN) {
1022 				error = EMSGSIZE;
1023 				break;
1024 			}
1025 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1026 			if (m == NULL) {
1027 				error = ENOBUFS;
1028 				break;
1029 			}
1030 			m->m_len = sopt->sopt_valsize;
1031 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1032 					    m->m_len);
1033 			if (error) {
1034 				m_free(m);
1035 				break;
1036 			}
1037 			INP_WLOCK(inp);
1038 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1039 			INP_WUNLOCK(inp);
1040 			return (error);
1041 		}
1042 
1043 		case IP_BINDANY:
1044 			if (sopt->sopt_td != NULL) {
1045 				error = priv_check(sopt->sopt_td,
1046 				    PRIV_NETINET_BINDANY);
1047 				if (error)
1048 					break;
1049 			}
1050 			/* FALLTHROUGH */
1051 		case IP_BINDMULTI:
1052 #ifdef	RSS
1053 		case IP_RSS_LISTEN_BUCKET:
1054 #endif
1055 		case IP_TOS:
1056 		case IP_TTL:
1057 		case IP_MINTTL:
1058 		case IP_RECVOPTS:
1059 		case IP_RECVRETOPTS:
1060 		case IP_ORIGDSTADDR:
1061 		case IP_RECVDSTADDR:
1062 		case IP_RECVTTL:
1063 		case IP_RECVIF:
1064 		case IP_ONESBCAST:
1065 		case IP_DONTFRAG:
1066 		case IP_RECVTOS:
1067 		case IP_RECVFLOWID:
1068 #ifdef	RSS
1069 		case IP_RECVRSSBUCKETID:
1070 #endif
1071 			error = sooptcopyin(sopt, &optval, sizeof optval,
1072 					    sizeof optval);
1073 			if (error)
1074 				break;
1075 
1076 			switch (sopt->sopt_name) {
1077 			case IP_TOS:
1078 				inp->inp_ip_tos = optval;
1079 				break;
1080 
1081 			case IP_TTL:
1082 				inp->inp_ip_ttl = optval;
1083 				break;
1084 
1085 			case IP_MINTTL:
1086 				if (optval >= 0 && optval <= MAXTTL)
1087 					inp->inp_ip_minttl = optval;
1088 				else
1089 					error = EINVAL;
1090 				break;
1091 
1092 #define	OPTSET(bit) do {						\
1093 	INP_WLOCK(inp);							\
1094 	if (optval)							\
1095 		inp->inp_flags |= bit;					\
1096 	else								\
1097 		inp->inp_flags &= ~bit;					\
1098 	INP_WUNLOCK(inp);						\
1099 } while (0)
1100 
1101 #define	OPTSET2(bit, val) do {						\
1102 	INP_WLOCK(inp);							\
1103 	if (val)							\
1104 		inp->inp_flags2 |= bit;					\
1105 	else								\
1106 		inp->inp_flags2 &= ~bit;				\
1107 	INP_WUNLOCK(inp);						\
1108 } while (0)
1109 
1110 			case IP_RECVOPTS:
1111 				OPTSET(INP_RECVOPTS);
1112 				break;
1113 
1114 			case IP_RECVRETOPTS:
1115 				OPTSET(INP_RECVRETOPTS);
1116 				break;
1117 
1118 			case IP_RECVDSTADDR:
1119 				OPTSET(INP_RECVDSTADDR);
1120 				break;
1121 
1122 			case IP_ORIGDSTADDR:
1123 				OPTSET2(INP_ORIGDSTADDR, optval);
1124 				break;
1125 
1126 			case IP_RECVTTL:
1127 				OPTSET(INP_RECVTTL);
1128 				break;
1129 
1130 			case IP_RECVIF:
1131 				OPTSET(INP_RECVIF);
1132 				break;
1133 
1134 			case IP_ONESBCAST:
1135 				OPTSET(INP_ONESBCAST);
1136 				break;
1137 			case IP_DONTFRAG:
1138 				OPTSET(INP_DONTFRAG);
1139 				break;
1140 			case IP_BINDANY:
1141 				OPTSET(INP_BINDANY);
1142 				break;
1143 			case IP_RECVTOS:
1144 				OPTSET(INP_RECVTOS);
1145 				break;
1146 			case IP_BINDMULTI:
1147 				OPTSET2(INP_BINDMULTI, optval);
1148 				break;
1149 			case IP_RECVFLOWID:
1150 				OPTSET2(INP_RECVFLOWID, optval);
1151 				break;
1152 #ifdef	RSS
1153 			case IP_RSS_LISTEN_BUCKET:
1154 				if ((optval >= 0) &&
1155 				    (optval < rss_getnumbuckets())) {
1156 					inp->inp_rss_listen_bucket = optval;
1157 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1158 				} else {
1159 					error = EINVAL;
1160 				}
1161 				break;
1162 			case IP_RECVRSSBUCKETID:
1163 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1164 				break;
1165 #endif
1166 			}
1167 			break;
1168 #undef OPTSET
1169 #undef OPTSET2
1170 
1171 		/*
1172 		 * Multicast socket options are processed by the in_mcast
1173 		 * module.
1174 		 */
1175 		case IP_MULTICAST_IF:
1176 		case IP_MULTICAST_VIF:
1177 		case IP_MULTICAST_TTL:
1178 		case IP_MULTICAST_LOOP:
1179 		case IP_ADD_MEMBERSHIP:
1180 		case IP_DROP_MEMBERSHIP:
1181 		case IP_ADD_SOURCE_MEMBERSHIP:
1182 		case IP_DROP_SOURCE_MEMBERSHIP:
1183 		case IP_BLOCK_SOURCE:
1184 		case IP_UNBLOCK_SOURCE:
1185 		case IP_MSFILTER:
1186 		case MCAST_JOIN_GROUP:
1187 		case MCAST_LEAVE_GROUP:
1188 		case MCAST_JOIN_SOURCE_GROUP:
1189 		case MCAST_LEAVE_SOURCE_GROUP:
1190 		case MCAST_BLOCK_SOURCE:
1191 		case MCAST_UNBLOCK_SOURCE:
1192 			error = inp_setmoptions(inp, sopt);
1193 			break;
1194 
1195 		case IP_PORTRANGE:
1196 			error = sooptcopyin(sopt, &optval, sizeof optval,
1197 					    sizeof optval);
1198 			if (error)
1199 				break;
1200 
1201 			INP_WLOCK(inp);
1202 			switch (optval) {
1203 			case IP_PORTRANGE_DEFAULT:
1204 				inp->inp_flags &= ~(INP_LOWPORT);
1205 				inp->inp_flags &= ~(INP_HIGHPORT);
1206 				break;
1207 
1208 			case IP_PORTRANGE_HIGH:
1209 				inp->inp_flags &= ~(INP_LOWPORT);
1210 				inp->inp_flags |= INP_HIGHPORT;
1211 				break;
1212 
1213 			case IP_PORTRANGE_LOW:
1214 				inp->inp_flags &= ~(INP_HIGHPORT);
1215 				inp->inp_flags |= INP_LOWPORT;
1216 				break;
1217 
1218 			default:
1219 				error = EINVAL;
1220 				break;
1221 			}
1222 			INP_WUNLOCK(inp);
1223 			break;
1224 
1225 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1226 		case IP_IPSEC_POLICY:
1227 			if (IPSEC_ENABLED(ipv4)) {
1228 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1229 				break;
1230 			}
1231 			/* FALLTHROUGH */
1232 #endif /* IPSEC */
1233 
1234 		default:
1235 			error = ENOPROTOOPT;
1236 			break;
1237 		}
1238 		break;
1239 
1240 	case SOPT_GET:
1241 		switch (sopt->sopt_name) {
1242 		case IP_OPTIONS:
1243 		case IP_RETOPTS:
1244 			if (inp->inp_options)
1245 				error = sooptcopyout(sopt,
1246 						     mtod(inp->inp_options,
1247 							  char *),
1248 						     inp->inp_options->m_len);
1249 			else
1250 				sopt->sopt_valsize = 0;
1251 			break;
1252 
1253 		case IP_TOS:
1254 		case IP_TTL:
1255 		case IP_MINTTL:
1256 		case IP_RECVOPTS:
1257 		case IP_RECVRETOPTS:
1258 		case IP_ORIGDSTADDR:
1259 		case IP_RECVDSTADDR:
1260 		case IP_RECVTTL:
1261 		case IP_RECVIF:
1262 		case IP_PORTRANGE:
1263 		case IP_ONESBCAST:
1264 		case IP_DONTFRAG:
1265 		case IP_BINDANY:
1266 		case IP_RECVTOS:
1267 		case IP_BINDMULTI:
1268 		case IP_FLOWID:
1269 		case IP_FLOWTYPE:
1270 		case IP_RECVFLOWID:
1271 #ifdef	RSS
1272 		case IP_RSSBUCKETID:
1273 		case IP_RECVRSSBUCKETID:
1274 #endif
1275 			switch (sopt->sopt_name) {
1276 
1277 			case IP_TOS:
1278 				optval = inp->inp_ip_tos;
1279 				break;
1280 
1281 			case IP_TTL:
1282 				optval = inp->inp_ip_ttl;
1283 				break;
1284 
1285 			case IP_MINTTL:
1286 				optval = inp->inp_ip_minttl;
1287 				break;
1288 
1289 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1290 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1291 
1292 			case IP_RECVOPTS:
1293 				optval = OPTBIT(INP_RECVOPTS);
1294 				break;
1295 
1296 			case IP_RECVRETOPTS:
1297 				optval = OPTBIT(INP_RECVRETOPTS);
1298 				break;
1299 
1300 			case IP_RECVDSTADDR:
1301 				optval = OPTBIT(INP_RECVDSTADDR);
1302 				break;
1303 
1304 			case IP_ORIGDSTADDR:
1305 				optval = OPTBIT2(INP_ORIGDSTADDR);
1306 				break;
1307 
1308 			case IP_RECVTTL:
1309 				optval = OPTBIT(INP_RECVTTL);
1310 				break;
1311 
1312 			case IP_RECVIF:
1313 				optval = OPTBIT(INP_RECVIF);
1314 				break;
1315 
1316 			case IP_PORTRANGE:
1317 				if (inp->inp_flags & INP_HIGHPORT)
1318 					optval = IP_PORTRANGE_HIGH;
1319 				else if (inp->inp_flags & INP_LOWPORT)
1320 					optval = IP_PORTRANGE_LOW;
1321 				else
1322 					optval = 0;
1323 				break;
1324 
1325 			case IP_ONESBCAST:
1326 				optval = OPTBIT(INP_ONESBCAST);
1327 				break;
1328 			case IP_DONTFRAG:
1329 				optval = OPTBIT(INP_DONTFRAG);
1330 				break;
1331 			case IP_BINDANY:
1332 				optval = OPTBIT(INP_BINDANY);
1333 				break;
1334 			case IP_RECVTOS:
1335 				optval = OPTBIT(INP_RECVTOS);
1336 				break;
1337 			case IP_FLOWID:
1338 				optval = inp->inp_flowid;
1339 				break;
1340 			case IP_FLOWTYPE:
1341 				optval = inp->inp_flowtype;
1342 				break;
1343 			case IP_RECVFLOWID:
1344 				optval = OPTBIT2(INP_RECVFLOWID);
1345 				break;
1346 #ifdef	RSS
1347 			case IP_RSSBUCKETID:
1348 				retval = rss_hash2bucket(inp->inp_flowid,
1349 				    inp->inp_flowtype,
1350 				    &rss_bucket);
1351 				if (retval == 0)
1352 					optval = rss_bucket;
1353 				else
1354 					error = EINVAL;
1355 				break;
1356 			case IP_RECVRSSBUCKETID:
1357 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1358 				break;
1359 #endif
1360 			case IP_BINDMULTI:
1361 				optval = OPTBIT2(INP_BINDMULTI);
1362 				break;
1363 			}
1364 			error = sooptcopyout(sopt, &optval, sizeof optval);
1365 			break;
1366 
1367 		/*
1368 		 * Multicast socket options are processed by the in_mcast
1369 		 * module.
1370 		 */
1371 		case IP_MULTICAST_IF:
1372 		case IP_MULTICAST_VIF:
1373 		case IP_MULTICAST_TTL:
1374 		case IP_MULTICAST_LOOP:
1375 		case IP_MSFILTER:
1376 			error = inp_getmoptions(inp, sopt);
1377 			break;
1378 
1379 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1380 		case IP_IPSEC_POLICY:
1381 			if (IPSEC_ENABLED(ipv4)) {
1382 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1383 				break;
1384 			}
1385 			/* FALLTHROUGH */
1386 #endif /* IPSEC */
1387 
1388 		default:
1389 			error = ENOPROTOOPT;
1390 			break;
1391 		}
1392 		break;
1393 	}
1394 	return (error);
1395 }
1396 
1397 /*
1398  * Routine called from ip_output() to loop back a copy of an IP multicast
1399  * packet to the input queue of a specified interface.  Note that this
1400  * calls the output routine of the loopback "driver", but with an interface
1401  * pointer that might NOT be a loopback interface -- evil, but easier than
1402  * replicating that code here.
1403  */
1404 static void
1405 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1406 {
1407 	struct ip *ip;
1408 	struct mbuf *copym;
1409 
1410 	/*
1411 	 * Make a deep copy of the packet because we're going to
1412 	 * modify the pack in order to generate checksums.
1413 	 */
1414 	copym = m_dup(m, M_NOWAIT);
1415 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1416 		copym = m_pullup(copym, hlen);
1417 	if (copym != NULL) {
1418 		/* If needed, compute the checksum and mark it as valid. */
1419 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1420 			in_delayed_cksum(copym);
1421 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1422 			copym->m_pkthdr.csum_flags |=
1423 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1424 			copym->m_pkthdr.csum_data = 0xffff;
1425 		}
1426 		/*
1427 		 * We don't bother to fragment if the IP length is greater
1428 		 * than the interface's MTU.  Can this possibly matter?
1429 		 */
1430 		ip = mtod(copym, struct ip *);
1431 		ip->ip_sum = 0;
1432 		ip->ip_sum = in_cksum(copym, hlen);
1433 		if_simloop(ifp, copym, AF_INET, 0);
1434 	}
1435 }
1436