xref: /freebsd/sys/netinet/ip_output.c (revision 28f4385e45a2681c14bd04b83fe1796eaefe8265)
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 
84 #include <netinet/udp.h>
85 #include <netinet/udp_var.h>
86 
87 #ifdef SCTP
88 #include <netinet/sctp.h>
89 #include <netinet/sctp_crc32.h>
90 #endif
91 
92 #include <netipsec/ipsec_support.h>
93 
94 #include <machine/in_cksum.h>
95 
96 #include <security/mac/mac_framework.h>
97 
98 #ifdef MBUF_STRESS_TEST
99 static int mbuf_frag_size = 0;
100 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
101 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
102 #endif
103 
104 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
105 
106 
107 extern int in_mcast_loop;
108 extern	struct protosw inetsw[];
109 
110 static inline int
111 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
112     struct sockaddr_in *dst, int *fibnum, int *error)
113 {
114 	struct m_tag *fwd_tag = NULL;
115 	struct mbuf *m;
116 	struct in_addr odst;
117 	struct ip *ip;
118 
119 	m = *mp;
120 	ip = mtod(m, struct ip *);
121 
122 	/* Run through list of hooks for output packets. */
123 	odst.s_addr = ip->ip_dst.s_addr;
124 	*error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, 0, inp);
125 	m = *mp;
126 	if ((*error) != 0 || m == NULL)
127 		return 1; /* Finished */
128 
129 	ip = mtod(m, struct ip *);
130 
131 	/* See if destination IP address was changed by packet filter. */
132 	if (odst.s_addr != ip->ip_dst.s_addr) {
133 		m->m_flags |= M_SKIP_FIREWALL;
134 		/* If destination is now ourself drop to ip_input(). */
135 		if (in_localip(ip->ip_dst)) {
136 			m->m_flags |= M_FASTFWD_OURS;
137 			if (m->m_pkthdr.rcvif == NULL)
138 				m->m_pkthdr.rcvif = V_loif;
139 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
140 				m->m_pkthdr.csum_flags |=
141 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
142 				m->m_pkthdr.csum_data = 0xffff;
143 			}
144 			m->m_pkthdr.csum_flags |=
145 				CSUM_IP_CHECKED | CSUM_IP_VALID;
146 #ifdef SCTP
147 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
148 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
149 #endif
150 			*error = netisr_queue(NETISR_IP, m);
151 			return 1; /* Finished */
152 		}
153 
154 		bzero(dst, sizeof(*dst));
155 		dst->sin_family = AF_INET;
156 		dst->sin_len = sizeof(*dst);
157 		dst->sin_addr = ip->ip_dst;
158 
159 		return -1; /* Reloop */
160 	}
161 	/* See if fib was changed by packet filter. */
162 	if ((*fibnum) != M_GETFIB(m)) {
163 		m->m_flags |= M_SKIP_FIREWALL;
164 		*fibnum = M_GETFIB(m);
165 		return -1; /* Reloop for FIB change */
166 	}
167 
168 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
169 	if (m->m_flags & M_FASTFWD_OURS) {
170 		if (m->m_pkthdr.rcvif == NULL)
171 			m->m_pkthdr.rcvif = V_loif;
172 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
173 			m->m_pkthdr.csum_flags |=
174 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
175 			m->m_pkthdr.csum_data = 0xffff;
176 		}
177 #ifdef SCTP
178 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
179 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
180 #endif
181 		m->m_pkthdr.csum_flags |=
182 			CSUM_IP_CHECKED | CSUM_IP_VALID;
183 
184 		*error = netisr_queue(NETISR_IP, m);
185 		return 1; /* Finished */
186 	}
187 	/* Or forward to some other address? */
188 	if ((m->m_flags & M_IP_NEXTHOP) &&
189 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
190 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
191 		m->m_flags |= M_SKIP_FIREWALL;
192 		m->m_flags &= ~M_IP_NEXTHOP;
193 		m_tag_delete(m, fwd_tag);
194 
195 		return -1; /* Reloop for CHANGE of dst */
196 	}
197 
198 	return 0;
199 }
200 
201 /*
202  * IP output.  The packet in mbuf chain m contains a skeletal IP
203  * header (with len, off, ttl, proto, tos, src, dst).
204  * The mbuf chain containing the packet will be freed.
205  * The mbuf opt, if present, will not be freed.
206  * If route ro is present and has ro_rt initialized, route lookup would be
207  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
208  * then result of route lookup is stored in ro->ro_rt.
209  *
210  * In the IP forwarding case, the packet will arrive with options already
211  * inserted, so must have a NULL opt pointer.
212  */
213 int
214 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
215     struct ip_moptions *imo, struct inpcb *inp)
216 {
217 	struct rm_priotracker in_ifa_tracker;
218 	struct epoch_tracker et;
219 	struct ip *ip;
220 	struct ifnet *ifp = NULL;	/* keep compiler happy */
221 	struct mbuf *m0;
222 	int hlen = sizeof (struct ip);
223 	int mtu;
224 	int error = 0;
225 	struct sockaddr_in *dst;
226 	const struct sockaddr_in *gw;
227 	struct in_ifaddr *ia;
228 	int isbroadcast;
229 	uint16_t ip_len, ip_off;
230 	struct route iproute;
231 	struct rtentry *rte;	/* cache for ro->ro_rt */
232 	uint32_t fibnum;
233 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
234 	int no_route_but_check_spd = 0;
235 #endif
236 	M_ASSERTPKTHDR(m);
237 
238 	if (inp != NULL) {
239 		INP_LOCK_ASSERT(inp);
240 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
241 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
242 			m->m_pkthdr.flowid = inp->inp_flowid;
243 			M_HASHTYPE_SET(m, inp->inp_flowtype);
244 		}
245 	}
246 
247 	if (ro == NULL) {
248 		ro = &iproute;
249 		bzero(ro, sizeof (*ro));
250 	}
251 
252 	if (opt) {
253 		int len = 0;
254 		m = ip_insertoptions(m, opt, &len);
255 		if (len != 0)
256 			hlen = len; /* ip->ip_hl is updated above */
257 	}
258 	ip = mtod(m, struct ip *);
259 	ip_len = ntohs(ip->ip_len);
260 	ip_off = ntohs(ip->ip_off);
261 
262 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
263 		ip->ip_v = IPVERSION;
264 		ip->ip_hl = hlen >> 2;
265 		ip_fillid(ip);
266 	} else {
267 		/* Header already set, fetch hlen from there */
268 		hlen = ip->ip_hl << 2;
269 	}
270 	if ((flags & IP_FORWARDING) == 0)
271 		IPSTAT_INC(ips_localout);
272 
273 	/*
274 	 * dst/gw handling:
275 	 *
276 	 * dst can be rewritten but always points to &ro->ro_dst.
277 	 * gw is readonly but can point either to dst OR rt_gateway,
278 	 * therefore we need restore gw if we're redoing lookup.
279 	 */
280 	gw = dst = (struct sockaddr_in *)&ro->ro_dst;
281 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
282 	rte = ro->ro_rt;
283 	if (rte == NULL) {
284 		bzero(dst, sizeof(*dst));
285 		dst->sin_family = AF_INET;
286 		dst->sin_len = sizeof(*dst);
287 		dst->sin_addr = ip->ip_dst;
288 	}
289 	NET_EPOCH_ENTER(et);
290 again:
291 	/*
292 	 * Validate route against routing table additions;
293 	 * a better/more specific route might have been added.
294 	 */
295 	if (inp)
296 		RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
297 	/*
298 	 * If there is a cached route,
299 	 * check that it is to the same destination
300 	 * and is still up.  If not, free it and try again.
301 	 * The address family should also be checked in case of sharing the
302 	 * cache with IPv6.
303 	 * Also check whether routing cache needs invalidation.
304 	 */
305 	rte = ro->ro_rt;
306 	if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
307 		    rte->rt_ifp == NULL ||
308 		    !RT_LINK_IS_UP(rte->rt_ifp) ||
309 			  dst->sin_family != AF_INET ||
310 			  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
311 		RO_INVALIDATE_CACHE(ro);
312 		rte = NULL;
313 	}
314 	ia = NULL;
315 	/*
316 	 * If routing to interface only, short circuit routing lookup.
317 	 * The use of an all-ones broadcast address implies this; an
318 	 * interface is specified by the broadcast address of an interface,
319 	 * or the destination address of a ptp interface.
320 	 */
321 	if (flags & IP_SENDONES) {
322 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
323 						      M_GETFIB(m)))) == NULL &&
324 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
325 						    M_GETFIB(m)))) == NULL) {
326 			IPSTAT_INC(ips_noroute);
327 			error = ENETUNREACH;
328 			goto bad;
329 		}
330 		ip->ip_dst.s_addr = INADDR_BROADCAST;
331 		dst->sin_addr = ip->ip_dst;
332 		ifp = ia->ia_ifp;
333 		ip->ip_ttl = 1;
334 		isbroadcast = 1;
335 	} else if (flags & IP_ROUTETOIF) {
336 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
337 						    M_GETFIB(m)))) == NULL &&
338 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
339 						M_GETFIB(m)))) == NULL) {
340 			IPSTAT_INC(ips_noroute);
341 			error = ENETUNREACH;
342 			goto bad;
343 		}
344 		ifp = ia->ia_ifp;
345 		ip->ip_ttl = 1;
346 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
347 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
348 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
349 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
350 		/*
351 		 * Bypass the normal routing lookup for multicast
352 		 * packets if the interface is specified.
353 		 */
354 		ifp = imo->imo_multicast_ifp;
355 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
356 		isbroadcast = 0;	/* fool gcc */
357 	} else {
358 		/*
359 		 * We want to do any cloning requested by the link layer,
360 		 * as this is probably required in all cases for correct
361 		 * operation (as it is for ARP).
362 		 */
363 		if (rte == NULL) {
364 #ifdef RADIX_MPATH
365 			rtalloc_mpath_fib(ro,
366 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
367 			    fibnum);
368 #else
369 			in_rtalloc_ign(ro, 0, fibnum);
370 #endif
371 			rte = ro->ro_rt;
372 		}
373 		if (rte == NULL ||
374 		    (rte->rt_flags & RTF_UP) == 0 ||
375 		    rte->rt_ifp == NULL ||
376 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
377 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
378 			/*
379 			 * There is no route for this packet, but it is
380 			 * possible that a matching SPD entry exists.
381 			 */
382 			no_route_but_check_spd = 1;
383 			mtu = 0; /* Silence GCC warning. */
384 			goto sendit;
385 #endif
386 			IPSTAT_INC(ips_noroute);
387 			error = EHOSTUNREACH;
388 			goto bad;
389 		}
390 		ia = ifatoia(rte->rt_ifa);
391 		ifp = rte->rt_ifp;
392 		counter_u64_add(rte->rt_pksent, 1);
393 		rt_update_ro_flags(ro);
394 		if (rte->rt_flags & RTF_GATEWAY)
395 			gw = (struct sockaddr_in *)rte->rt_gateway;
396 		if (rte->rt_flags & RTF_HOST)
397 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
398 		else if (ifp->if_flags & IFF_BROADCAST)
399 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
400 		else
401 			isbroadcast = 0;
402 	}
403 
404 	/*
405 	 * Calculate MTU.  If we have a route that is up, use that,
406 	 * otherwise use the interface's MTU.
407 	 */
408 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
409 		mtu = rte->rt_mtu;
410 	else
411 		mtu = ifp->if_mtu;
412 	/* Catch a possible divide by zero later. */
413 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
414 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
415 
416 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
417 		m->m_flags |= M_MCAST;
418 		/*
419 		 * IP destination address is multicast.  Make sure "gw"
420 		 * still points to the address in "ro".  (It may have been
421 		 * changed to point to a gateway address, above.)
422 		 */
423 		gw = dst;
424 		/*
425 		 * See if the caller provided any multicast options
426 		 */
427 		if (imo != NULL) {
428 			ip->ip_ttl = imo->imo_multicast_ttl;
429 			if (imo->imo_multicast_vif != -1)
430 				ip->ip_src.s_addr =
431 				    ip_mcast_src ?
432 				    ip_mcast_src(imo->imo_multicast_vif) :
433 				    INADDR_ANY;
434 		} else
435 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
436 		/*
437 		 * Confirm that the outgoing interface supports multicast.
438 		 */
439 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
440 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
441 				IPSTAT_INC(ips_noroute);
442 				error = ENETUNREACH;
443 				goto bad;
444 			}
445 		}
446 		/*
447 		 * If source address not specified yet, use address
448 		 * of outgoing interface.
449 		 */
450 		if (ip->ip_src.s_addr == INADDR_ANY) {
451 			/* Interface may have no addresses. */
452 			if (ia != NULL)
453 				ip->ip_src = IA_SIN(ia)->sin_addr;
454 		}
455 
456 		if ((imo == NULL && in_mcast_loop) ||
457 		    (imo && imo->imo_multicast_loop)) {
458 			/*
459 			 * Loop back multicast datagram if not expressly
460 			 * forbidden to do so, even if we are not a member
461 			 * of the group; ip_input() will filter it later,
462 			 * thus deferring a hash lookup and mutex acquisition
463 			 * at the expense of a cheap copy using m_copym().
464 			 */
465 			ip_mloopback(ifp, m, hlen);
466 		} else {
467 			/*
468 			 * If we are acting as a multicast router, perform
469 			 * multicast forwarding as if the packet had just
470 			 * arrived on the interface to which we are about
471 			 * to send.  The multicast forwarding function
472 			 * recursively calls this function, using the
473 			 * IP_FORWARDING flag to prevent infinite recursion.
474 			 *
475 			 * Multicasts that are looped back by ip_mloopback(),
476 			 * above, will be forwarded by the ip_input() routine,
477 			 * if necessary.
478 			 */
479 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
480 				/*
481 				 * If rsvp daemon is not running, do not
482 				 * set ip_moptions. This ensures that the packet
483 				 * is multicast and not just sent down one link
484 				 * as prescribed by rsvpd.
485 				 */
486 				if (!V_rsvp_on)
487 					imo = NULL;
488 				if (ip_mforward &&
489 				    ip_mforward(ip, ifp, m, imo) != 0) {
490 					m_freem(m);
491 					goto done;
492 				}
493 			}
494 		}
495 
496 		/*
497 		 * Multicasts with a time-to-live of zero may be looped-
498 		 * back, above, but must not be transmitted on a network.
499 		 * Also, multicasts addressed to the loopback interface
500 		 * are not sent -- the above call to ip_mloopback() will
501 		 * loop back a copy. ip_input() will drop the copy if
502 		 * this host does not belong to the destination group on
503 		 * the loopback interface.
504 		 */
505 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
506 			m_freem(m);
507 			goto done;
508 		}
509 
510 		goto sendit;
511 	}
512 
513 	/*
514 	 * If the source address is not specified yet, use the address
515 	 * of the outoing interface.
516 	 */
517 	if (ip->ip_src.s_addr == INADDR_ANY) {
518 		/* Interface may have no addresses. */
519 		if (ia != NULL) {
520 			ip->ip_src = IA_SIN(ia)->sin_addr;
521 		}
522 	}
523 
524 	/*
525 	 * Look for broadcast address and
526 	 * verify user is allowed to send
527 	 * such a packet.
528 	 */
529 	if (isbroadcast) {
530 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
531 			error = EADDRNOTAVAIL;
532 			goto bad;
533 		}
534 		if ((flags & IP_ALLOWBROADCAST) == 0) {
535 			error = EACCES;
536 			goto bad;
537 		}
538 		/* don't allow broadcast messages to be fragmented */
539 		if (ip_len > mtu) {
540 			error = EMSGSIZE;
541 			goto bad;
542 		}
543 		m->m_flags |= M_BCAST;
544 	} else {
545 		m->m_flags &= ~M_BCAST;
546 	}
547 
548 sendit:
549 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
550 	if (IPSEC_ENABLED(ipv4)) {
551 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
552 			if (error == EINPROGRESS)
553 				error = 0;
554 			goto done;
555 		}
556 	}
557 	/*
558 	 * Check if there was a route for this packet; return error if not.
559 	 */
560 	if (no_route_but_check_spd) {
561 		IPSTAT_INC(ips_noroute);
562 		error = EHOSTUNREACH;
563 		goto bad;
564 	}
565 	/* Update variables that are affected by ipsec4_output(). */
566 	ip = mtod(m, struct ip *);
567 	hlen = ip->ip_hl << 2;
568 #endif /* IPSEC */
569 
570 	/* Jump over all PFIL processing if hooks are not active. */
571 	if (PFIL_HOOKED(&V_inet_pfil_hook)) {
572 		switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
573 		case 1: /* Finished */
574 			goto done;
575 
576 		case 0: /* Continue normally */
577 			ip = mtod(m, struct ip *);
578 			break;
579 
580 		case -1: /* Need to try again */
581 			/* Reset everything for a new round */
582 			RO_RTFREE(ro);
583 			ro->ro_prepend = NULL;
584 			rte = NULL;
585 			gw = dst;
586 			ip = mtod(m, struct ip *);
587 			goto again;
588 
589 		}
590 	}
591 
592 	/* 127/8 must not appear on wire - RFC1122. */
593 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
594 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
595 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
596 			IPSTAT_INC(ips_badaddr);
597 			error = EADDRNOTAVAIL;
598 			goto bad;
599 		}
600 	}
601 
602 	m->m_pkthdr.csum_flags |= CSUM_IP;
603 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
604 		in_delayed_cksum(m);
605 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
606 	}
607 #ifdef SCTP
608 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
609 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
610 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
611 	}
612 #endif
613 
614 	/*
615 	 * If small enough for interface, or the interface will take
616 	 * care of the fragmentation for us, we can just send directly.
617 	 */
618 	if (ip_len <= mtu ||
619 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
620 		ip->ip_sum = 0;
621 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
622 			ip->ip_sum = in_cksum(m, hlen);
623 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
624 		}
625 
626 		/*
627 		 * Record statistics for this interface address.
628 		 * With CSUM_TSO the byte/packet count will be slightly
629 		 * incorrect because we count the IP+TCP headers only
630 		 * once instead of for every generated packet.
631 		 */
632 		if (!(flags & IP_FORWARDING) && ia) {
633 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
634 				counter_u64_add(ia->ia_ifa.ifa_opackets,
635 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
636 			else
637 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
638 
639 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
640 		}
641 #ifdef MBUF_STRESS_TEST
642 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
643 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
644 #endif
645 		/*
646 		 * Reset layer specific mbuf flags
647 		 * to avoid confusing lower layers.
648 		 */
649 		m_clrprotoflags(m);
650 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
651 #ifdef RATELIMIT
652 		if (inp != NULL) {
653 			if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
654 				in_pcboutput_txrtlmt(inp, ifp, m);
655 			/* stamp send tag on mbuf */
656 			m->m_pkthdr.snd_tag = inp->inp_snd_tag;
657 		} else {
658 			m->m_pkthdr.snd_tag = NULL;
659 		}
660 #endif
661 		error = (*ifp->if_output)(ifp, m,
662 		    (const struct sockaddr *)gw, ro);
663 #ifdef RATELIMIT
664 		/* check for route change */
665 		if (error == EAGAIN)
666 			in_pcboutput_eagain(inp);
667 #endif
668 		goto done;
669 	}
670 
671 	/* Balk when DF bit is set or the interface didn't support TSO. */
672 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
673 		error = EMSGSIZE;
674 		IPSTAT_INC(ips_cantfrag);
675 		goto bad;
676 	}
677 
678 	/*
679 	 * Too large for interface; fragment if possible. If successful,
680 	 * on return, m will point to a list of packets to be sent.
681 	 */
682 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
683 	if (error)
684 		goto bad;
685 	for (; m; m = m0) {
686 		m0 = m->m_nextpkt;
687 		m->m_nextpkt = 0;
688 		if (error == 0) {
689 			/* Record statistics for this interface address. */
690 			if (ia != NULL) {
691 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
692 				counter_u64_add(ia->ia_ifa.ifa_obytes,
693 				    m->m_pkthdr.len);
694 			}
695 			/*
696 			 * Reset layer specific mbuf flags
697 			 * to avoid confusing upper layers.
698 			 */
699 			m_clrprotoflags(m);
700 
701 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
702 			    mtod(m, struct ip *), NULL);
703 #ifdef RATELIMIT
704 			if (inp != NULL) {
705 				if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
706 					in_pcboutput_txrtlmt(inp, ifp, m);
707 				/* stamp send tag on mbuf */
708 				m->m_pkthdr.snd_tag = inp->inp_snd_tag;
709 			} else {
710 				m->m_pkthdr.snd_tag = NULL;
711 			}
712 #endif
713 			error = (*ifp->if_output)(ifp, m,
714 			    (const struct sockaddr *)gw, ro);
715 #ifdef RATELIMIT
716 			/* check for route change */
717 			if (error == EAGAIN)
718 				in_pcboutput_eagain(inp);
719 #endif
720 		} else
721 			m_freem(m);
722 	}
723 
724 	if (error == 0)
725 		IPSTAT_INC(ips_fragmented);
726 
727 done:
728 	if (ro == &iproute)
729 		RO_RTFREE(ro);
730 	else if (rte == NULL)
731 		/*
732 		 * If the caller supplied a route but somehow the reference
733 		 * to it has been released need to prevent the caller
734 		 * calling RTFREE on it again.
735 		 */
736 		ro->ro_rt = NULL;
737 	NET_EPOCH_EXIT(et);
738 	return (error);
739  bad:
740 	m_freem(m);
741 	goto done;
742 }
743 
744 /*
745  * Create a chain of fragments which fit the given mtu. m_frag points to the
746  * mbuf to be fragmented; on return it points to the chain with the fragments.
747  * Return 0 if no error. If error, m_frag may contain a partially built
748  * chain of fragments that should be freed by the caller.
749  *
750  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
751  */
752 int
753 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
754     u_long if_hwassist_flags)
755 {
756 	int error = 0;
757 	int hlen = ip->ip_hl << 2;
758 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
759 	int off;
760 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
761 	int firstlen;
762 	struct mbuf **mnext;
763 	int nfrags;
764 	uint16_t ip_len, ip_off;
765 
766 	ip_len = ntohs(ip->ip_len);
767 	ip_off = ntohs(ip->ip_off);
768 
769 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
770 		IPSTAT_INC(ips_cantfrag);
771 		return EMSGSIZE;
772 	}
773 
774 	/*
775 	 * Must be able to put at least 8 bytes per fragment.
776 	 */
777 	if (len < 8)
778 		return EMSGSIZE;
779 
780 	/*
781 	 * If the interface will not calculate checksums on
782 	 * fragmented packets, then do it here.
783 	 */
784 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
785 		in_delayed_cksum(m0);
786 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
787 	}
788 #ifdef SCTP
789 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
790 		sctp_delayed_cksum(m0, hlen);
791 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
792 	}
793 #endif
794 	if (len > PAGE_SIZE) {
795 		/*
796 		 * Fragment large datagrams such that each segment
797 		 * contains a multiple of PAGE_SIZE amount of data,
798 		 * plus headers. This enables a receiver to perform
799 		 * page-flipping zero-copy optimizations.
800 		 *
801 		 * XXX When does this help given that sender and receiver
802 		 * could have different page sizes, and also mtu could
803 		 * be less than the receiver's page size ?
804 		 */
805 		int newlen;
806 
807 		off = MIN(mtu, m0->m_pkthdr.len);
808 
809 		/*
810 		 * firstlen (off - hlen) must be aligned on an
811 		 * 8-byte boundary
812 		 */
813 		if (off < hlen)
814 			goto smart_frag_failure;
815 		off = ((off - hlen) & ~7) + hlen;
816 		newlen = (~PAGE_MASK) & mtu;
817 		if ((newlen + sizeof (struct ip)) > mtu) {
818 			/* we failed, go back the default */
819 smart_frag_failure:
820 			newlen = len;
821 			off = hlen + len;
822 		}
823 		len = newlen;
824 
825 	} else {
826 		off = hlen + len;
827 	}
828 
829 	firstlen = off - hlen;
830 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
831 
832 	/*
833 	 * Loop through length of segment after first fragment,
834 	 * make new header and copy data of each part and link onto chain.
835 	 * Here, m0 is the original packet, m is the fragment being created.
836 	 * The fragments are linked off the m_nextpkt of the original
837 	 * packet, which after processing serves as the first fragment.
838 	 */
839 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
840 		struct ip *mhip;	/* ip header on the fragment */
841 		struct mbuf *m;
842 		int mhlen = sizeof (struct ip);
843 
844 		m = m_gethdr(M_NOWAIT, MT_DATA);
845 		if (m == NULL) {
846 			error = ENOBUFS;
847 			IPSTAT_INC(ips_odropped);
848 			goto done;
849 		}
850 		/*
851 		 * Make sure the complete packet header gets copied
852 		 * from the originating mbuf to the newly created
853 		 * mbuf. This also ensures that existing firewall
854 		 * classification(s), VLAN tags and so on get copied
855 		 * to the resulting fragmented packet(s):
856 		 */
857 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
858 			m_free(m);
859 			error = ENOBUFS;
860 			IPSTAT_INC(ips_odropped);
861 			goto done;
862 		}
863 		/*
864 		 * In the first mbuf, leave room for the link header, then
865 		 * copy the original IP header including options. The payload
866 		 * goes into an additional mbuf chain returned by m_copym().
867 		 */
868 		m->m_data += max_linkhdr;
869 		mhip = mtod(m, struct ip *);
870 		*mhip = *ip;
871 		if (hlen > sizeof (struct ip)) {
872 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
873 			mhip->ip_v = IPVERSION;
874 			mhip->ip_hl = mhlen >> 2;
875 		}
876 		m->m_len = mhlen;
877 		/* XXX do we need to add ip_off below ? */
878 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
879 		if (off + len >= ip_len)
880 			len = ip_len - off;
881 		else
882 			mhip->ip_off |= IP_MF;
883 		mhip->ip_len = htons((u_short)(len + mhlen));
884 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
885 		if (m->m_next == NULL) {	/* copy failed */
886 			m_free(m);
887 			error = ENOBUFS;	/* ??? */
888 			IPSTAT_INC(ips_odropped);
889 			goto done;
890 		}
891 		m->m_pkthdr.len = mhlen + len;
892 #ifdef MAC
893 		mac_netinet_fragment(m0, m);
894 #endif
895 		mhip->ip_off = htons(mhip->ip_off);
896 		mhip->ip_sum = 0;
897 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
898 			mhip->ip_sum = in_cksum(m, mhlen);
899 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
900 		}
901 		*mnext = m;
902 		mnext = &m->m_nextpkt;
903 	}
904 	IPSTAT_ADD(ips_ofragments, nfrags);
905 
906 	/*
907 	 * Update first fragment by trimming what's been copied out
908 	 * and updating header.
909 	 */
910 	m_adj(m0, hlen + firstlen - ip_len);
911 	m0->m_pkthdr.len = hlen + firstlen;
912 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
913 	ip->ip_off = htons(ip_off | IP_MF);
914 	ip->ip_sum = 0;
915 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
916 		ip->ip_sum = in_cksum(m0, hlen);
917 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
918 	}
919 
920 done:
921 	*m_frag = m0;
922 	return error;
923 }
924 
925 void
926 in_delayed_cksum(struct mbuf *m)
927 {
928 	struct ip *ip;
929 	struct udphdr *uh;
930 	uint16_t cklen, csum, offset;
931 
932 	ip = mtod(m, struct ip *);
933 	offset = ip->ip_hl << 2 ;
934 
935 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
936 		/* if udp header is not in the first mbuf copy udplen */
937 		if (offset + sizeof(struct udphdr) > m->m_len) {
938 			m_copydata(m, offset + offsetof(struct udphdr,
939 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
940 			cklen = ntohs(cklen);
941 		} else {
942 			uh = (struct udphdr *)mtodo(m, offset);
943 			cklen = ntohs(uh->uh_ulen);
944 		}
945 		csum = in_cksum_skip(m, cklen + offset, offset);
946 		if (csum == 0)
947 			csum = 0xffff;
948 	} else {
949 		cklen = ntohs(ip->ip_len);
950 		csum = in_cksum_skip(m, cklen, offset);
951 	}
952 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
953 
954 	if (offset + sizeof(csum) > m->m_len)
955 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
956 	else
957 		*(u_short *)mtodo(m, offset) = csum;
958 }
959 
960 /*
961  * IP socket option processing.
962  */
963 int
964 ip_ctloutput(struct socket *so, struct sockopt *sopt)
965 {
966 	struct	inpcb *inp = sotoinpcb(so);
967 	int	error, optval;
968 #ifdef	RSS
969 	uint32_t rss_bucket;
970 	int retval;
971 #endif
972 
973 	error = optval = 0;
974 	if (sopt->sopt_level != IPPROTO_IP) {
975 		error = EINVAL;
976 
977 		if (sopt->sopt_level == SOL_SOCKET &&
978 		    sopt->sopt_dir == SOPT_SET) {
979 			switch (sopt->sopt_name) {
980 			case SO_REUSEADDR:
981 				INP_WLOCK(inp);
982 				if ((so->so_options & SO_REUSEADDR) != 0)
983 					inp->inp_flags2 |= INP_REUSEADDR;
984 				else
985 					inp->inp_flags2 &= ~INP_REUSEADDR;
986 				INP_WUNLOCK(inp);
987 				error = 0;
988 				break;
989 			case SO_REUSEPORT:
990 				INP_WLOCK(inp);
991 				if ((so->so_options & SO_REUSEPORT) != 0)
992 					inp->inp_flags2 |= INP_REUSEPORT;
993 				else
994 					inp->inp_flags2 &= ~INP_REUSEPORT;
995 				INP_WUNLOCK(inp);
996 				error = 0;
997 				break;
998 			case SO_REUSEPORT_LB:
999 				INP_WLOCK(inp);
1000 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1001 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1002 				else
1003 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1004 				INP_WUNLOCK(inp);
1005 				error = 0;
1006 				break;
1007 			case SO_SETFIB:
1008 				INP_WLOCK(inp);
1009 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1010 				INP_WUNLOCK(inp);
1011 				error = 0;
1012 				break;
1013 			case SO_MAX_PACING_RATE:
1014 #ifdef RATELIMIT
1015 				INP_WLOCK(inp);
1016 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1017 				INP_WUNLOCK(inp);
1018 				error = 0;
1019 #else
1020 				error = EOPNOTSUPP;
1021 #endif
1022 				break;
1023 			default:
1024 				break;
1025 			}
1026 		}
1027 		return (error);
1028 	}
1029 
1030 	switch (sopt->sopt_dir) {
1031 	case SOPT_SET:
1032 		switch (sopt->sopt_name) {
1033 		case IP_OPTIONS:
1034 #ifdef notyet
1035 		case IP_RETOPTS:
1036 #endif
1037 		{
1038 			struct mbuf *m;
1039 			if (sopt->sopt_valsize > MLEN) {
1040 				error = EMSGSIZE;
1041 				break;
1042 			}
1043 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1044 			if (m == NULL) {
1045 				error = ENOBUFS;
1046 				break;
1047 			}
1048 			m->m_len = sopt->sopt_valsize;
1049 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1050 					    m->m_len);
1051 			if (error) {
1052 				m_free(m);
1053 				break;
1054 			}
1055 			INP_WLOCK(inp);
1056 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1057 			INP_WUNLOCK(inp);
1058 			return (error);
1059 		}
1060 
1061 		case IP_BINDANY:
1062 			if (sopt->sopt_td != NULL) {
1063 				error = priv_check(sopt->sopt_td,
1064 				    PRIV_NETINET_BINDANY);
1065 				if (error)
1066 					break;
1067 			}
1068 			/* FALLTHROUGH */
1069 		case IP_BINDMULTI:
1070 #ifdef	RSS
1071 		case IP_RSS_LISTEN_BUCKET:
1072 #endif
1073 		case IP_TOS:
1074 		case IP_TTL:
1075 		case IP_MINTTL:
1076 		case IP_RECVOPTS:
1077 		case IP_RECVRETOPTS:
1078 		case IP_ORIGDSTADDR:
1079 		case IP_RECVDSTADDR:
1080 		case IP_RECVTTL:
1081 		case IP_RECVIF:
1082 		case IP_ONESBCAST:
1083 		case IP_DONTFRAG:
1084 		case IP_RECVTOS:
1085 		case IP_RECVFLOWID:
1086 #ifdef	RSS
1087 		case IP_RECVRSSBUCKETID:
1088 #endif
1089 			error = sooptcopyin(sopt, &optval, sizeof optval,
1090 					    sizeof optval);
1091 			if (error)
1092 				break;
1093 
1094 			switch (sopt->sopt_name) {
1095 			case IP_TOS:
1096 				inp->inp_ip_tos = optval;
1097 				break;
1098 
1099 			case IP_TTL:
1100 				inp->inp_ip_ttl = optval;
1101 				break;
1102 
1103 			case IP_MINTTL:
1104 				if (optval >= 0 && optval <= MAXTTL)
1105 					inp->inp_ip_minttl = optval;
1106 				else
1107 					error = EINVAL;
1108 				break;
1109 
1110 #define	OPTSET(bit) do {						\
1111 	INP_WLOCK(inp);							\
1112 	if (optval)							\
1113 		inp->inp_flags |= bit;					\
1114 	else								\
1115 		inp->inp_flags &= ~bit;					\
1116 	INP_WUNLOCK(inp);						\
1117 } while (0)
1118 
1119 #define	OPTSET2(bit, val) do {						\
1120 	INP_WLOCK(inp);							\
1121 	if (val)							\
1122 		inp->inp_flags2 |= bit;					\
1123 	else								\
1124 		inp->inp_flags2 &= ~bit;				\
1125 	INP_WUNLOCK(inp);						\
1126 } while (0)
1127 
1128 			case IP_RECVOPTS:
1129 				OPTSET(INP_RECVOPTS);
1130 				break;
1131 
1132 			case IP_RECVRETOPTS:
1133 				OPTSET(INP_RECVRETOPTS);
1134 				break;
1135 
1136 			case IP_RECVDSTADDR:
1137 				OPTSET(INP_RECVDSTADDR);
1138 				break;
1139 
1140 			case IP_ORIGDSTADDR:
1141 				OPTSET2(INP_ORIGDSTADDR, optval);
1142 				break;
1143 
1144 			case IP_RECVTTL:
1145 				OPTSET(INP_RECVTTL);
1146 				break;
1147 
1148 			case IP_RECVIF:
1149 				OPTSET(INP_RECVIF);
1150 				break;
1151 
1152 			case IP_ONESBCAST:
1153 				OPTSET(INP_ONESBCAST);
1154 				break;
1155 			case IP_DONTFRAG:
1156 				OPTSET(INP_DONTFRAG);
1157 				break;
1158 			case IP_BINDANY:
1159 				OPTSET(INP_BINDANY);
1160 				break;
1161 			case IP_RECVTOS:
1162 				OPTSET(INP_RECVTOS);
1163 				break;
1164 			case IP_BINDMULTI:
1165 				OPTSET2(INP_BINDMULTI, optval);
1166 				break;
1167 			case IP_RECVFLOWID:
1168 				OPTSET2(INP_RECVFLOWID, optval);
1169 				break;
1170 #ifdef	RSS
1171 			case IP_RSS_LISTEN_BUCKET:
1172 				if ((optval >= 0) &&
1173 				    (optval < rss_getnumbuckets())) {
1174 					inp->inp_rss_listen_bucket = optval;
1175 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1176 				} else {
1177 					error = EINVAL;
1178 				}
1179 				break;
1180 			case IP_RECVRSSBUCKETID:
1181 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1182 				break;
1183 #endif
1184 			}
1185 			break;
1186 #undef OPTSET
1187 #undef OPTSET2
1188 
1189 		/*
1190 		 * Multicast socket options are processed by the in_mcast
1191 		 * module.
1192 		 */
1193 		case IP_MULTICAST_IF:
1194 		case IP_MULTICAST_VIF:
1195 		case IP_MULTICAST_TTL:
1196 		case IP_MULTICAST_LOOP:
1197 		case IP_ADD_MEMBERSHIP:
1198 		case IP_DROP_MEMBERSHIP:
1199 		case IP_ADD_SOURCE_MEMBERSHIP:
1200 		case IP_DROP_SOURCE_MEMBERSHIP:
1201 		case IP_BLOCK_SOURCE:
1202 		case IP_UNBLOCK_SOURCE:
1203 		case IP_MSFILTER:
1204 		case MCAST_JOIN_GROUP:
1205 		case MCAST_LEAVE_GROUP:
1206 		case MCAST_JOIN_SOURCE_GROUP:
1207 		case MCAST_LEAVE_SOURCE_GROUP:
1208 		case MCAST_BLOCK_SOURCE:
1209 		case MCAST_UNBLOCK_SOURCE:
1210 			error = inp_setmoptions(inp, sopt);
1211 			break;
1212 
1213 		case IP_PORTRANGE:
1214 			error = sooptcopyin(sopt, &optval, sizeof optval,
1215 					    sizeof optval);
1216 			if (error)
1217 				break;
1218 
1219 			INP_WLOCK(inp);
1220 			switch (optval) {
1221 			case IP_PORTRANGE_DEFAULT:
1222 				inp->inp_flags &= ~(INP_LOWPORT);
1223 				inp->inp_flags &= ~(INP_HIGHPORT);
1224 				break;
1225 
1226 			case IP_PORTRANGE_HIGH:
1227 				inp->inp_flags &= ~(INP_LOWPORT);
1228 				inp->inp_flags |= INP_HIGHPORT;
1229 				break;
1230 
1231 			case IP_PORTRANGE_LOW:
1232 				inp->inp_flags &= ~(INP_HIGHPORT);
1233 				inp->inp_flags |= INP_LOWPORT;
1234 				break;
1235 
1236 			default:
1237 				error = EINVAL;
1238 				break;
1239 			}
1240 			INP_WUNLOCK(inp);
1241 			break;
1242 
1243 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1244 		case IP_IPSEC_POLICY:
1245 			if (IPSEC_ENABLED(ipv4)) {
1246 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1247 				break;
1248 			}
1249 			/* FALLTHROUGH */
1250 #endif /* IPSEC */
1251 
1252 		default:
1253 			error = ENOPROTOOPT;
1254 			break;
1255 		}
1256 		break;
1257 
1258 	case SOPT_GET:
1259 		switch (sopt->sopt_name) {
1260 		case IP_OPTIONS:
1261 		case IP_RETOPTS:
1262 			INP_RLOCK(inp);
1263 			if (inp->inp_options) {
1264 				struct mbuf *options;
1265 
1266 				options = m_copym(inp->inp_options, 0,
1267 				    M_COPYALL, M_NOWAIT);
1268 				INP_RUNLOCK(inp);
1269 				if (options != NULL) {
1270 					error = sooptcopyout(sopt,
1271 							     mtod(options, char *),
1272 							     options->m_len);
1273 					m_freem(options);
1274 				} else
1275 					error = ENOMEM;
1276 			} else {
1277 				INP_RUNLOCK(inp);
1278 				sopt->sopt_valsize = 0;
1279 			}
1280 			break;
1281 
1282 		case IP_TOS:
1283 		case IP_TTL:
1284 		case IP_MINTTL:
1285 		case IP_RECVOPTS:
1286 		case IP_RECVRETOPTS:
1287 		case IP_ORIGDSTADDR:
1288 		case IP_RECVDSTADDR:
1289 		case IP_RECVTTL:
1290 		case IP_RECVIF:
1291 		case IP_PORTRANGE:
1292 		case IP_ONESBCAST:
1293 		case IP_DONTFRAG:
1294 		case IP_BINDANY:
1295 		case IP_RECVTOS:
1296 		case IP_BINDMULTI:
1297 		case IP_FLOWID:
1298 		case IP_FLOWTYPE:
1299 		case IP_RECVFLOWID:
1300 #ifdef	RSS
1301 		case IP_RSSBUCKETID:
1302 		case IP_RECVRSSBUCKETID:
1303 #endif
1304 			switch (sopt->sopt_name) {
1305 
1306 			case IP_TOS:
1307 				optval = inp->inp_ip_tos;
1308 				break;
1309 
1310 			case IP_TTL:
1311 				optval = inp->inp_ip_ttl;
1312 				break;
1313 
1314 			case IP_MINTTL:
1315 				optval = inp->inp_ip_minttl;
1316 				break;
1317 
1318 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1319 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1320 
1321 			case IP_RECVOPTS:
1322 				optval = OPTBIT(INP_RECVOPTS);
1323 				break;
1324 
1325 			case IP_RECVRETOPTS:
1326 				optval = OPTBIT(INP_RECVRETOPTS);
1327 				break;
1328 
1329 			case IP_RECVDSTADDR:
1330 				optval = OPTBIT(INP_RECVDSTADDR);
1331 				break;
1332 
1333 			case IP_ORIGDSTADDR:
1334 				optval = OPTBIT2(INP_ORIGDSTADDR);
1335 				break;
1336 
1337 			case IP_RECVTTL:
1338 				optval = OPTBIT(INP_RECVTTL);
1339 				break;
1340 
1341 			case IP_RECVIF:
1342 				optval = OPTBIT(INP_RECVIF);
1343 				break;
1344 
1345 			case IP_PORTRANGE:
1346 				if (inp->inp_flags & INP_HIGHPORT)
1347 					optval = IP_PORTRANGE_HIGH;
1348 				else if (inp->inp_flags & INP_LOWPORT)
1349 					optval = IP_PORTRANGE_LOW;
1350 				else
1351 					optval = 0;
1352 				break;
1353 
1354 			case IP_ONESBCAST:
1355 				optval = OPTBIT(INP_ONESBCAST);
1356 				break;
1357 			case IP_DONTFRAG:
1358 				optval = OPTBIT(INP_DONTFRAG);
1359 				break;
1360 			case IP_BINDANY:
1361 				optval = OPTBIT(INP_BINDANY);
1362 				break;
1363 			case IP_RECVTOS:
1364 				optval = OPTBIT(INP_RECVTOS);
1365 				break;
1366 			case IP_FLOWID:
1367 				optval = inp->inp_flowid;
1368 				break;
1369 			case IP_FLOWTYPE:
1370 				optval = inp->inp_flowtype;
1371 				break;
1372 			case IP_RECVFLOWID:
1373 				optval = OPTBIT2(INP_RECVFLOWID);
1374 				break;
1375 #ifdef	RSS
1376 			case IP_RSSBUCKETID:
1377 				retval = rss_hash2bucket(inp->inp_flowid,
1378 				    inp->inp_flowtype,
1379 				    &rss_bucket);
1380 				if (retval == 0)
1381 					optval = rss_bucket;
1382 				else
1383 					error = EINVAL;
1384 				break;
1385 			case IP_RECVRSSBUCKETID:
1386 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1387 				break;
1388 #endif
1389 			case IP_BINDMULTI:
1390 				optval = OPTBIT2(INP_BINDMULTI);
1391 				break;
1392 			}
1393 			error = sooptcopyout(sopt, &optval, sizeof optval);
1394 			break;
1395 
1396 		/*
1397 		 * Multicast socket options are processed by the in_mcast
1398 		 * module.
1399 		 */
1400 		case IP_MULTICAST_IF:
1401 		case IP_MULTICAST_VIF:
1402 		case IP_MULTICAST_TTL:
1403 		case IP_MULTICAST_LOOP:
1404 		case IP_MSFILTER:
1405 			error = inp_getmoptions(inp, sopt);
1406 			break;
1407 
1408 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1409 		case IP_IPSEC_POLICY:
1410 			if (IPSEC_ENABLED(ipv4)) {
1411 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1412 				break;
1413 			}
1414 			/* FALLTHROUGH */
1415 #endif /* IPSEC */
1416 
1417 		default:
1418 			error = ENOPROTOOPT;
1419 			break;
1420 		}
1421 		break;
1422 	}
1423 	return (error);
1424 }
1425 
1426 /*
1427  * Routine called from ip_output() to loop back a copy of an IP multicast
1428  * packet to the input queue of a specified interface.  Note that this
1429  * calls the output routine of the loopback "driver", but with an interface
1430  * pointer that might NOT be a loopback interface -- evil, but easier than
1431  * replicating that code here.
1432  */
1433 static void
1434 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1435 {
1436 	struct ip *ip;
1437 	struct mbuf *copym;
1438 
1439 	/*
1440 	 * Make a deep copy of the packet because we're going to
1441 	 * modify the pack in order to generate checksums.
1442 	 */
1443 	copym = m_dup(m, M_NOWAIT);
1444 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1445 		copym = m_pullup(copym, hlen);
1446 	if (copym != NULL) {
1447 		/* If needed, compute the checksum and mark it as valid. */
1448 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1449 			in_delayed_cksum(copym);
1450 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1451 			copym->m_pkthdr.csum_flags |=
1452 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1453 			copym->m_pkthdr.csum_data = 0xffff;
1454 		}
1455 		/*
1456 		 * We don't bother to fragment if the IP length is greater
1457 		 * than the interface's MTU.  Can this possibly matter?
1458 		 */
1459 		ip = mtod(copym, struct ip *);
1460 		ip->ip_sum = 0;
1461 		ip->ip_sum = in_cksum(copym, hlen);
1462 		if_simloop(ifp, copym, AF_INET, 0);
1463 	}
1464 }
1465