xref: /freebsd/sys/netinet/ip_output.c (revision 67ca7330cf34a789afbbff9ae7e4cdc4a4917ae3)
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_fib.h>
76 #include <netinet/in_kdtrace.h>
77 #include <netinet/in_systm.h>
78 #include <netinet/ip.h>
79 #include <netinet/in_pcb.h>
80 #include <netinet/in_rss.h>
81 #include <netinet/in_var.h>
82 #include <netinet/ip_var.h>
83 #include <netinet/ip_options.h>
84 
85 #include <netinet/udp.h>
86 #include <netinet/udp_var.h>
87 
88 #ifdef SCTP
89 #include <netinet/sctp.h>
90 #include <netinet/sctp_crc32.h>
91 #endif
92 
93 #include <netipsec/ipsec_support.h>
94 
95 #include <machine/in_cksum.h>
96 
97 #include <security/mac/mac_framework.h>
98 
99 #ifdef MBUF_STRESS_TEST
100 static int mbuf_frag_size = 0;
101 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
102 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
103 #endif
104 
105 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
106 
107 
108 extern int in_mcast_loop;
109 extern	struct protosw inetsw[];
110 
111 static inline int
112 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
113     struct sockaddr_in *dst, int *fibnum, int *error)
114 {
115 	struct m_tag *fwd_tag = NULL;
116 	struct mbuf *m;
117 	struct in_addr odst;
118 	struct ip *ip;
119 
120 	m = *mp;
121 	ip = mtod(m, struct ip *);
122 
123 	/* Run through list of hooks for output packets. */
124 	odst.s_addr = ip->ip_dst.s_addr;
125 	switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, PFIL_OUT, inp)) {
126 	case PFIL_DROPPED:
127 		*error = EPERM;
128 		/* FALLTHROUGH */
129 	case PFIL_CONSUMED:
130 		return 1; /* Finished */
131 	case PFIL_PASS:
132 		*error = 0;
133 	}
134 	m = *mp;
135 	ip = mtod(m, struct ip *);
136 
137 	/* See if destination IP address was changed by packet filter. */
138 	if (odst.s_addr != ip->ip_dst.s_addr) {
139 		m->m_flags |= M_SKIP_FIREWALL;
140 		/* If destination is now ourself drop to ip_input(). */
141 		if (in_localip(ip->ip_dst)) {
142 			m->m_flags |= M_FASTFWD_OURS;
143 			if (m->m_pkthdr.rcvif == NULL)
144 				m->m_pkthdr.rcvif = V_loif;
145 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
146 				m->m_pkthdr.csum_flags |=
147 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
148 				m->m_pkthdr.csum_data = 0xffff;
149 			}
150 			m->m_pkthdr.csum_flags |=
151 				CSUM_IP_CHECKED | CSUM_IP_VALID;
152 #ifdef SCTP
153 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
154 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
155 #endif
156 			*error = netisr_queue(NETISR_IP, m);
157 			return 1; /* Finished */
158 		}
159 
160 		bzero(dst, sizeof(*dst));
161 		dst->sin_family = AF_INET;
162 		dst->sin_len = sizeof(*dst);
163 		dst->sin_addr = ip->ip_dst;
164 
165 		return -1; /* Reloop */
166 	}
167 	/* See if fib was changed by packet filter. */
168 	if ((*fibnum) != M_GETFIB(m)) {
169 		m->m_flags |= M_SKIP_FIREWALL;
170 		*fibnum = M_GETFIB(m);
171 		return -1; /* Reloop for FIB change */
172 	}
173 
174 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
175 	if (m->m_flags & M_FASTFWD_OURS) {
176 		if (m->m_pkthdr.rcvif == NULL)
177 			m->m_pkthdr.rcvif = V_loif;
178 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
179 			m->m_pkthdr.csum_flags |=
180 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
181 			m->m_pkthdr.csum_data = 0xffff;
182 		}
183 #ifdef SCTP
184 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
185 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
186 #endif
187 		m->m_pkthdr.csum_flags |=
188 			CSUM_IP_CHECKED | CSUM_IP_VALID;
189 
190 		*error = netisr_queue(NETISR_IP, m);
191 		return 1; /* Finished */
192 	}
193 	/* Or forward to some other address? */
194 	if ((m->m_flags & M_IP_NEXTHOP) &&
195 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
196 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
197 		m->m_flags |= M_SKIP_FIREWALL;
198 		m->m_flags &= ~M_IP_NEXTHOP;
199 		m_tag_delete(m, fwd_tag);
200 
201 		return -1; /* Reloop for CHANGE of dst */
202 	}
203 
204 	return 0;
205 }
206 
207 static int
208 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
209     const struct sockaddr_in *gw, struct route *ro)
210 {
211 	struct m_snd_tag *mst;
212 	int error;
213 
214 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
215 	mst = NULL;
216 
217 #ifdef RATELIMIT
218 	if (inp != NULL) {
219 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
220 		    (inp->inp_snd_tag != NULL &&
221 		    inp->inp_snd_tag->ifp != ifp))
222 			in_pcboutput_txrtlmt(inp, ifp, m);
223 
224 		if (inp->inp_snd_tag != NULL)
225 			mst = inp->inp_snd_tag;
226 	}
227 #endif
228 	if (mst != NULL) {
229 		KASSERT(m->m_pkthdr.rcvif == NULL,
230 		    ("trying to add a send tag to a forwarded packet"));
231 		if (mst->ifp != ifp) {
232 			error = EAGAIN;
233 			goto done;
234 		}
235 
236 		/* stamp send tag on mbuf */
237 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
238 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
239 	}
240 
241 	error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
242 
243 done:
244 	/* Check for route change invalidating send tags. */
245 #ifdef RATELIMIT
246 	if (error == EAGAIN)
247 		in_pcboutput_eagain(inp);
248 #endif
249 	return (error);
250 }
251 
252 /*
253  * IP output.  The packet in mbuf chain m contains a skeletal IP
254  * header (with len, off, ttl, proto, tos, src, dst).
255  * The mbuf chain containing the packet will be freed.
256  * The mbuf opt, if present, will not be freed.
257  * If route ro is present and has ro_rt initialized, route lookup would be
258  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
259  * then result of route lookup is stored in ro->ro_rt.
260  *
261  * In the IP forwarding case, the packet will arrive with options already
262  * inserted, so must have a NULL opt pointer.
263  */
264 int
265 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
266     struct ip_moptions *imo, struct inpcb *inp)
267 {
268 	struct rm_priotracker in_ifa_tracker;
269 	struct epoch_tracker et;
270 	struct ip *ip;
271 	struct ifnet *ifp = NULL;	/* keep compiler happy */
272 	struct mbuf *m0;
273 	int hlen = sizeof (struct ip);
274 	int mtu;
275 	int error = 0;
276 	struct sockaddr_in *dst, sin;
277 	const struct sockaddr_in *gw;
278 	struct in_ifaddr *ia;
279 	struct in_addr src;
280 	int isbroadcast;
281 	uint16_t ip_len, ip_off;
282 	uint32_t fibnum;
283 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
284 	int no_route_but_check_spd = 0;
285 #endif
286 	M_ASSERTPKTHDR(m);
287 
288 	if (inp != NULL) {
289 		INP_LOCK_ASSERT(inp);
290 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
291 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
292 			m->m_pkthdr.flowid = inp->inp_flowid;
293 			M_HASHTYPE_SET(m, inp->inp_flowtype);
294 		}
295 #ifdef NUMA
296 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
297 #endif
298 	}
299 
300 	if (opt) {
301 		int len = 0;
302 		m = ip_insertoptions(m, opt, &len);
303 		if (len != 0)
304 			hlen = len; /* ip->ip_hl is updated above */
305 	}
306 	ip = mtod(m, struct ip *);
307 	ip_len = ntohs(ip->ip_len);
308 	ip_off = ntohs(ip->ip_off);
309 
310 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
311 		ip->ip_v = IPVERSION;
312 		ip->ip_hl = hlen >> 2;
313 		ip_fillid(ip);
314 	} else {
315 		/* Header already set, fetch hlen from there */
316 		hlen = ip->ip_hl << 2;
317 	}
318 	if ((flags & IP_FORWARDING) == 0)
319 		IPSTAT_INC(ips_localout);
320 
321 	/*
322 	 * dst/gw handling:
323 	 *
324 	 * gw is readonly but can point either to dst OR rt_gateway,
325 	 * therefore we need restore gw if we're redoing lookup.
326 	 */
327 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
328 	if (ro != NULL)
329 		dst = (struct sockaddr_in *)&ro->ro_dst;
330 	else
331 		dst = &sin;
332 	if (ro == NULL || ro->ro_rt == NULL) {
333 		bzero(dst, sizeof(*dst));
334 		dst->sin_family = AF_INET;
335 		dst->sin_len = sizeof(*dst);
336 		dst->sin_addr = ip->ip_dst;
337 	}
338 	gw = dst;
339 	NET_EPOCH_ENTER(et);
340 again:
341 	/*
342 	 * Validate route against routing table additions;
343 	 * a better/more specific route might have been added.
344 	 */
345 	if (inp != NULL && ro != NULL && ro->ro_rt != NULL)
346 		RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
347 	/*
348 	 * If there is a cached route,
349 	 * check that it is to the same destination
350 	 * and is still up.  If not, free it and try again.
351 	 * The address family should also be checked in case of sharing the
352 	 * cache with IPv6.
353 	 * Also check whether routing cache needs invalidation.
354 	 */
355 	if (ro != NULL && ro->ro_rt != NULL &&
356 	    ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
357 	    ro->ro_rt->rt_ifp == NULL || !RT_LINK_IS_UP(ro->ro_rt->rt_ifp) ||
358 	    dst->sin_family != AF_INET ||
359 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
360 		RO_INVALIDATE_CACHE(ro);
361 	ia = NULL;
362 	/*
363 	 * If routing to interface only, short circuit routing lookup.
364 	 * The use of an all-ones broadcast address implies this; an
365 	 * interface is specified by the broadcast address of an interface,
366 	 * or the destination address of a ptp interface.
367 	 */
368 	if (flags & IP_SENDONES) {
369 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
370 						      M_GETFIB(m)))) == NULL &&
371 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
372 						    M_GETFIB(m)))) == NULL) {
373 			IPSTAT_INC(ips_noroute);
374 			error = ENETUNREACH;
375 			goto bad;
376 		}
377 		ip->ip_dst.s_addr = INADDR_BROADCAST;
378 		dst->sin_addr = ip->ip_dst;
379 		ifp = ia->ia_ifp;
380 		mtu = ifp->if_mtu;
381 		ip->ip_ttl = 1;
382 		isbroadcast = 1;
383 		src = IA_SIN(ia)->sin_addr;
384 	} else if (flags & IP_ROUTETOIF) {
385 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
386 						    M_GETFIB(m)))) == NULL &&
387 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
388 						M_GETFIB(m)))) == NULL) {
389 			IPSTAT_INC(ips_noroute);
390 			error = ENETUNREACH;
391 			goto bad;
392 		}
393 		ifp = ia->ia_ifp;
394 		mtu = ifp->if_mtu;
395 		ip->ip_ttl = 1;
396 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
397 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
398 		src = IA_SIN(ia)->sin_addr;
399 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
400 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
401 		/*
402 		 * Bypass the normal routing lookup for multicast
403 		 * packets if the interface is specified.
404 		 */
405 		ifp = imo->imo_multicast_ifp;
406 		mtu = ifp->if_mtu;
407 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
408 		isbroadcast = 0;	/* fool gcc */
409 		/* Interface may have no addresses. */
410 		if (ia != NULL)
411 			src = IA_SIN(ia)->sin_addr;
412 		else
413 			src.s_addr = INADDR_ANY;
414 	} else if (ro != NULL) {
415 		if (ro->ro_rt == NULL) {
416 			/*
417 			 * We want to do any cloning requested by the link
418 			 * layer, as this is probably required in all cases
419 			 * for correct operation (as it is for ARP).
420 			 */
421 #ifdef RADIX_MPATH
422 			rtalloc_mpath_fib(ro,
423 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
424 			    fibnum);
425 #else
426 			in_rtalloc_ign(ro, 0, fibnum);
427 #endif
428 			if (ro->ro_rt == NULL ||
429 			    (ro->ro_rt->rt_flags & RTF_UP) == 0 ||
430 			    ro->ro_rt->rt_ifp == NULL ||
431 			    !RT_LINK_IS_UP(ro->ro_rt->rt_ifp)) {
432 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
433 				/*
434 				 * There is no route for this packet, but it is
435 				 * possible that a matching SPD entry exists.
436 				 */
437 				no_route_but_check_spd = 1;
438 				mtu = 0; /* Silence GCC warning. */
439 				goto sendit;
440 #endif
441 				IPSTAT_INC(ips_noroute);
442 				error = EHOSTUNREACH;
443 				goto bad;
444 			}
445 		}
446 		ia = ifatoia(ro->ro_rt->rt_ifa);
447 		ifp = ro->ro_rt->rt_ifp;
448 		counter_u64_add(ro->ro_rt->rt_pksent, 1);
449 		rt_update_ro_flags(ro);
450 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
451 			gw = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
452 		if (ro->ro_rt->rt_flags & RTF_HOST)
453 			isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
454 		else if (ifp->if_flags & IFF_BROADCAST)
455 			isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
456 		else
457 			isbroadcast = 0;
458 		if (ro->ro_rt->rt_flags & RTF_HOST)
459 			mtu = ro->ro_rt->rt_mtu;
460 		else
461 			mtu = ifp->if_mtu;
462 		src = IA_SIN(ia)->sin_addr;
463 	} else {
464 		struct nhop4_extended nh;
465 
466 		bzero(&nh, sizeof(nh));
467 		if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh) !=
468 		    0) {
469 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
470 			/*
471 			 * There is no route for this packet, but it is
472 			 * possible that a matching SPD entry exists.
473 			 */
474 			no_route_but_check_spd = 1;
475 			mtu = 0; /* Silence GCC warning. */
476 			goto sendit;
477 #endif
478 			IPSTAT_INC(ips_noroute);
479 			error = EHOSTUNREACH;
480 			goto bad;
481 		}
482 		ifp = nh.nh_ifp;
483 		mtu = nh.nh_mtu;
484 		/*
485 		 * We are rewriting here dst to be gw actually, contradicting
486 		 * comment at the beginning of the function. However, in this
487 		 * case we are always dealing with on stack dst.
488 		 * In case if pfil(9) sends us back to beginning of the
489 		 * function, the dst would be rewritten by ip_output_pfil().
490 		 */
491 		MPASS(dst == &sin);
492 		dst->sin_addr = nh.nh_addr;
493 		ia = nh.nh_ia;
494 		src = nh.nh_src;
495 		isbroadcast = (((nh.nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
496 		    (NHF_HOST | NHF_BROADCAST)) ||
497 		    ((ifp->if_flags & IFF_BROADCAST) &&
498 		    in_ifaddr_broadcast(dst->sin_addr, ia)));
499 	}
500 
501 	/* Catch a possible divide by zero later. */
502 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (rt_flags=0x%08x) ifp=%p",
503 	    __func__, mtu, ro,
504 	    (ro != NULL && ro->ro_rt != NULL) ? ro->ro_rt->rt_flags : 0, ifp));
505 
506 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
507 		m->m_flags |= M_MCAST;
508 		/*
509 		 * IP destination address is multicast.  Make sure "gw"
510 		 * still points to the address in "ro".  (It may have been
511 		 * changed to point to a gateway address, above.)
512 		 */
513 		gw = dst;
514 		/*
515 		 * See if the caller provided any multicast options
516 		 */
517 		if (imo != NULL) {
518 			ip->ip_ttl = imo->imo_multicast_ttl;
519 			if (imo->imo_multicast_vif != -1)
520 				ip->ip_src.s_addr =
521 				    ip_mcast_src ?
522 				    ip_mcast_src(imo->imo_multicast_vif) :
523 				    INADDR_ANY;
524 		} else
525 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
526 		/*
527 		 * Confirm that the outgoing interface supports multicast.
528 		 */
529 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
530 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
531 				IPSTAT_INC(ips_noroute);
532 				error = ENETUNREACH;
533 				goto bad;
534 			}
535 		}
536 		/*
537 		 * If source address not specified yet, use address
538 		 * of outgoing interface.
539 		 */
540 		if (ip->ip_src.s_addr == INADDR_ANY)
541 			ip->ip_src = src;
542 
543 		if ((imo == NULL && in_mcast_loop) ||
544 		    (imo && imo->imo_multicast_loop)) {
545 			/*
546 			 * Loop back multicast datagram if not expressly
547 			 * forbidden to do so, even if we are not a member
548 			 * of the group; ip_input() will filter it later,
549 			 * thus deferring a hash lookup and mutex acquisition
550 			 * at the expense of a cheap copy using m_copym().
551 			 */
552 			ip_mloopback(ifp, m, hlen);
553 		} else {
554 			/*
555 			 * If we are acting as a multicast router, perform
556 			 * multicast forwarding as if the packet had just
557 			 * arrived on the interface to which we are about
558 			 * to send.  The multicast forwarding function
559 			 * recursively calls this function, using the
560 			 * IP_FORWARDING flag to prevent infinite recursion.
561 			 *
562 			 * Multicasts that are looped back by ip_mloopback(),
563 			 * above, will be forwarded by the ip_input() routine,
564 			 * if necessary.
565 			 */
566 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
567 				/*
568 				 * If rsvp daemon is not running, do not
569 				 * set ip_moptions. This ensures that the packet
570 				 * is multicast and not just sent down one link
571 				 * as prescribed by rsvpd.
572 				 */
573 				if (!V_rsvp_on)
574 					imo = NULL;
575 				if (ip_mforward &&
576 				    ip_mforward(ip, ifp, m, imo) != 0) {
577 					m_freem(m);
578 					goto done;
579 				}
580 			}
581 		}
582 
583 		/*
584 		 * Multicasts with a time-to-live of zero may be looped-
585 		 * back, above, but must not be transmitted on a network.
586 		 * Also, multicasts addressed to the loopback interface
587 		 * are not sent -- the above call to ip_mloopback() will
588 		 * loop back a copy. ip_input() will drop the copy if
589 		 * this host does not belong to the destination group on
590 		 * the loopback interface.
591 		 */
592 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
593 			m_freem(m);
594 			goto done;
595 		}
596 
597 		goto sendit;
598 	}
599 
600 	/*
601 	 * If the source address is not specified yet, use the address
602 	 * of the outoing interface.
603 	 */
604 	if (ip->ip_src.s_addr == INADDR_ANY)
605 		ip->ip_src = src;
606 
607 	/*
608 	 * Look for broadcast address and
609 	 * verify user is allowed to send
610 	 * such a packet.
611 	 */
612 	if (isbroadcast) {
613 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
614 			error = EADDRNOTAVAIL;
615 			goto bad;
616 		}
617 		if ((flags & IP_ALLOWBROADCAST) == 0) {
618 			error = EACCES;
619 			goto bad;
620 		}
621 		/* don't allow broadcast messages to be fragmented */
622 		if (ip_len > mtu) {
623 			error = EMSGSIZE;
624 			goto bad;
625 		}
626 		m->m_flags |= M_BCAST;
627 	} else {
628 		m->m_flags &= ~M_BCAST;
629 	}
630 
631 sendit:
632 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
633 	if (IPSEC_ENABLED(ipv4)) {
634 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
635 			if (error == EINPROGRESS)
636 				error = 0;
637 			goto done;
638 		}
639 	}
640 	/*
641 	 * Check if there was a route for this packet; return error if not.
642 	 */
643 	if (no_route_but_check_spd) {
644 		IPSTAT_INC(ips_noroute);
645 		error = EHOSTUNREACH;
646 		goto bad;
647 	}
648 	/* Update variables that are affected by ipsec4_output(). */
649 	ip = mtod(m, struct ip *);
650 	hlen = ip->ip_hl << 2;
651 #endif /* IPSEC */
652 
653 	/* Jump over all PFIL processing if hooks are not active. */
654 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
655 		switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
656 		case 1: /* Finished */
657 			goto done;
658 
659 		case 0: /* Continue normally */
660 			ip = mtod(m, struct ip *);
661 			break;
662 
663 		case -1: /* Need to try again */
664 			/* Reset everything for a new round */
665 			if (ro != NULL) {
666 				RO_RTFREE(ro);
667 				ro->ro_prepend = NULL;
668 			}
669 			gw = dst;
670 			ip = mtod(m, struct ip *);
671 			goto again;
672 
673 		}
674 	}
675 
676 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
677 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
678 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
679 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
680 			IPSTAT_INC(ips_badaddr);
681 			error = EADDRNOTAVAIL;
682 			goto bad;
683 		}
684 	}
685 
686 	m->m_pkthdr.csum_flags |= CSUM_IP;
687 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
688 		in_delayed_cksum(m);
689 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
690 	}
691 #ifdef SCTP
692 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
693 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
694 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
695 	}
696 #endif
697 
698 	/*
699 	 * If small enough for interface, or the interface will take
700 	 * care of the fragmentation for us, we can just send directly.
701 	 */
702 	if (ip_len <= mtu ||
703 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
704 		ip->ip_sum = 0;
705 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
706 			ip->ip_sum = in_cksum(m, hlen);
707 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
708 		}
709 
710 		/*
711 		 * Record statistics for this interface address.
712 		 * With CSUM_TSO the byte/packet count will be slightly
713 		 * incorrect because we count the IP+TCP headers only
714 		 * once instead of for every generated packet.
715 		 */
716 		if (!(flags & IP_FORWARDING) && ia) {
717 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
718 				counter_u64_add(ia->ia_ifa.ifa_opackets,
719 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
720 			else
721 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
722 
723 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
724 		}
725 #ifdef MBUF_STRESS_TEST
726 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
727 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
728 #endif
729 		/*
730 		 * Reset layer specific mbuf flags
731 		 * to avoid confusing lower layers.
732 		 */
733 		m_clrprotoflags(m);
734 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
735 		error = ip_output_send(inp, ifp, m, gw, ro);
736 		goto done;
737 	}
738 
739 	/* Balk when DF bit is set or the interface didn't support TSO. */
740 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
741 		error = EMSGSIZE;
742 		IPSTAT_INC(ips_cantfrag);
743 		goto bad;
744 	}
745 
746 	/*
747 	 * Too large for interface; fragment if possible. If successful,
748 	 * on return, m will point to a list of packets to be sent.
749 	 */
750 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
751 	if (error)
752 		goto bad;
753 	for (; m; m = m0) {
754 		m0 = m->m_nextpkt;
755 		m->m_nextpkt = 0;
756 		if (error == 0) {
757 			/* Record statistics for this interface address. */
758 			if (ia != NULL) {
759 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
760 				counter_u64_add(ia->ia_ifa.ifa_obytes,
761 				    m->m_pkthdr.len);
762 			}
763 			/*
764 			 * Reset layer specific mbuf flags
765 			 * to avoid confusing upper layers.
766 			 */
767 			m_clrprotoflags(m);
768 
769 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
770 			    mtod(m, struct ip *), NULL);
771 			error = ip_output_send(inp, ifp, m, gw, ro);
772 		} else
773 			m_freem(m);
774 	}
775 
776 	if (error == 0)
777 		IPSTAT_INC(ips_fragmented);
778 
779 done:
780 	NET_EPOCH_EXIT(et);
781 	return (error);
782  bad:
783 	m_freem(m);
784 	goto done;
785 }
786 
787 /*
788  * Create a chain of fragments which fit the given mtu. m_frag points to the
789  * mbuf to be fragmented; on return it points to the chain with the fragments.
790  * Return 0 if no error. If error, m_frag may contain a partially built
791  * chain of fragments that should be freed by the caller.
792  *
793  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
794  */
795 int
796 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
797     u_long if_hwassist_flags)
798 {
799 	int error = 0;
800 	int hlen = ip->ip_hl << 2;
801 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
802 	int off;
803 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
804 	int firstlen;
805 	struct mbuf **mnext;
806 	int nfrags;
807 	uint16_t ip_len, ip_off;
808 
809 	ip_len = ntohs(ip->ip_len);
810 	ip_off = ntohs(ip->ip_off);
811 
812 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
813 		IPSTAT_INC(ips_cantfrag);
814 		return EMSGSIZE;
815 	}
816 
817 	/*
818 	 * Must be able to put at least 8 bytes per fragment.
819 	 */
820 	if (len < 8)
821 		return EMSGSIZE;
822 
823 	/*
824 	 * If the interface will not calculate checksums on
825 	 * fragmented packets, then do it here.
826 	 */
827 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
828 		in_delayed_cksum(m0);
829 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
830 	}
831 #ifdef SCTP
832 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
833 		sctp_delayed_cksum(m0, hlen);
834 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
835 	}
836 #endif
837 	if (len > PAGE_SIZE) {
838 		/*
839 		 * Fragment large datagrams such that each segment
840 		 * contains a multiple of PAGE_SIZE amount of data,
841 		 * plus headers. This enables a receiver to perform
842 		 * page-flipping zero-copy optimizations.
843 		 *
844 		 * XXX When does this help given that sender and receiver
845 		 * could have different page sizes, and also mtu could
846 		 * be less than the receiver's page size ?
847 		 */
848 		int newlen;
849 
850 		off = MIN(mtu, m0->m_pkthdr.len);
851 
852 		/*
853 		 * firstlen (off - hlen) must be aligned on an
854 		 * 8-byte boundary
855 		 */
856 		if (off < hlen)
857 			goto smart_frag_failure;
858 		off = ((off - hlen) & ~7) + hlen;
859 		newlen = (~PAGE_MASK) & mtu;
860 		if ((newlen + sizeof (struct ip)) > mtu) {
861 			/* we failed, go back the default */
862 smart_frag_failure:
863 			newlen = len;
864 			off = hlen + len;
865 		}
866 		len = newlen;
867 
868 	} else {
869 		off = hlen + len;
870 	}
871 
872 	firstlen = off - hlen;
873 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
874 
875 	/*
876 	 * Loop through length of segment after first fragment,
877 	 * make new header and copy data of each part and link onto chain.
878 	 * Here, m0 is the original packet, m is the fragment being created.
879 	 * The fragments are linked off the m_nextpkt of the original
880 	 * packet, which after processing serves as the first fragment.
881 	 */
882 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
883 		struct ip *mhip;	/* ip header on the fragment */
884 		struct mbuf *m;
885 		int mhlen = sizeof (struct ip);
886 
887 		m = m_gethdr(M_NOWAIT, MT_DATA);
888 		if (m == NULL) {
889 			error = ENOBUFS;
890 			IPSTAT_INC(ips_odropped);
891 			goto done;
892 		}
893 		/*
894 		 * Make sure the complete packet header gets copied
895 		 * from the originating mbuf to the newly created
896 		 * mbuf. This also ensures that existing firewall
897 		 * classification(s), VLAN tags and so on get copied
898 		 * to the resulting fragmented packet(s):
899 		 */
900 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
901 			m_free(m);
902 			error = ENOBUFS;
903 			IPSTAT_INC(ips_odropped);
904 			goto done;
905 		}
906 		/*
907 		 * In the first mbuf, leave room for the link header, then
908 		 * copy the original IP header including options. The payload
909 		 * goes into an additional mbuf chain returned by m_copym().
910 		 */
911 		m->m_data += max_linkhdr;
912 		mhip = mtod(m, struct ip *);
913 		*mhip = *ip;
914 		if (hlen > sizeof (struct ip)) {
915 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
916 			mhip->ip_v = IPVERSION;
917 			mhip->ip_hl = mhlen >> 2;
918 		}
919 		m->m_len = mhlen;
920 		/* XXX do we need to add ip_off below ? */
921 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
922 		if (off + len >= ip_len)
923 			len = ip_len - off;
924 		else
925 			mhip->ip_off |= IP_MF;
926 		mhip->ip_len = htons((u_short)(len + mhlen));
927 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
928 		if (m->m_next == NULL) {	/* copy failed */
929 			m_free(m);
930 			error = ENOBUFS;	/* ??? */
931 			IPSTAT_INC(ips_odropped);
932 			goto done;
933 		}
934 		m->m_pkthdr.len = mhlen + len;
935 #ifdef MAC
936 		mac_netinet_fragment(m0, m);
937 #endif
938 		mhip->ip_off = htons(mhip->ip_off);
939 		mhip->ip_sum = 0;
940 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
941 			mhip->ip_sum = in_cksum(m, mhlen);
942 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
943 		}
944 		*mnext = m;
945 		mnext = &m->m_nextpkt;
946 	}
947 	IPSTAT_ADD(ips_ofragments, nfrags);
948 
949 	/*
950 	 * Update first fragment by trimming what's been copied out
951 	 * and updating header.
952 	 */
953 	m_adj(m0, hlen + firstlen - ip_len);
954 	m0->m_pkthdr.len = hlen + firstlen;
955 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
956 	ip->ip_off = htons(ip_off | IP_MF);
957 	ip->ip_sum = 0;
958 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
959 		ip->ip_sum = in_cksum(m0, hlen);
960 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
961 	}
962 
963 done:
964 	*m_frag = m0;
965 	return error;
966 }
967 
968 void
969 in_delayed_cksum(struct mbuf *m)
970 {
971 	struct ip *ip;
972 	struct udphdr *uh;
973 	uint16_t cklen, csum, offset;
974 
975 	ip = mtod(m, struct ip *);
976 	offset = ip->ip_hl << 2 ;
977 
978 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
979 		/* if udp header is not in the first mbuf copy udplen */
980 		if (offset + sizeof(struct udphdr) > m->m_len) {
981 			m_copydata(m, offset + offsetof(struct udphdr,
982 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
983 			cklen = ntohs(cklen);
984 		} else {
985 			uh = (struct udphdr *)mtodo(m, offset);
986 			cklen = ntohs(uh->uh_ulen);
987 		}
988 		csum = in_cksum_skip(m, cklen + offset, offset);
989 		if (csum == 0)
990 			csum = 0xffff;
991 	} else {
992 		cklen = ntohs(ip->ip_len);
993 		csum = in_cksum_skip(m, cklen, offset);
994 	}
995 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
996 
997 	if (offset + sizeof(csum) > m->m_len)
998 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
999 	else
1000 		*(u_short *)mtodo(m, offset) = csum;
1001 }
1002 
1003 /*
1004  * IP socket option processing.
1005  */
1006 int
1007 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1008 {
1009 	struct	inpcb *inp = sotoinpcb(so);
1010 	int	error, optval;
1011 #ifdef	RSS
1012 	uint32_t rss_bucket;
1013 	int retval;
1014 #endif
1015 
1016 	error = optval = 0;
1017 	if (sopt->sopt_level != IPPROTO_IP) {
1018 		error = EINVAL;
1019 
1020 		if (sopt->sopt_level == SOL_SOCKET &&
1021 		    sopt->sopt_dir == SOPT_SET) {
1022 			switch (sopt->sopt_name) {
1023 			case SO_REUSEADDR:
1024 				INP_WLOCK(inp);
1025 				if ((so->so_options & SO_REUSEADDR) != 0)
1026 					inp->inp_flags2 |= INP_REUSEADDR;
1027 				else
1028 					inp->inp_flags2 &= ~INP_REUSEADDR;
1029 				INP_WUNLOCK(inp);
1030 				error = 0;
1031 				break;
1032 			case SO_REUSEPORT:
1033 				INP_WLOCK(inp);
1034 				if ((so->so_options & SO_REUSEPORT) != 0)
1035 					inp->inp_flags2 |= INP_REUSEPORT;
1036 				else
1037 					inp->inp_flags2 &= ~INP_REUSEPORT;
1038 				INP_WUNLOCK(inp);
1039 				error = 0;
1040 				break;
1041 			case SO_REUSEPORT_LB:
1042 				INP_WLOCK(inp);
1043 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1044 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1045 				else
1046 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1047 				INP_WUNLOCK(inp);
1048 				error = 0;
1049 				break;
1050 			case SO_SETFIB:
1051 				INP_WLOCK(inp);
1052 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1053 				INP_WUNLOCK(inp);
1054 				error = 0;
1055 				break;
1056 			case SO_MAX_PACING_RATE:
1057 #ifdef RATELIMIT
1058 				INP_WLOCK(inp);
1059 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1060 				INP_WUNLOCK(inp);
1061 				error = 0;
1062 #else
1063 				error = EOPNOTSUPP;
1064 #endif
1065 				break;
1066 			default:
1067 				break;
1068 			}
1069 		}
1070 		return (error);
1071 	}
1072 
1073 	switch (sopt->sopt_dir) {
1074 	case SOPT_SET:
1075 		switch (sopt->sopt_name) {
1076 		case IP_OPTIONS:
1077 #ifdef notyet
1078 		case IP_RETOPTS:
1079 #endif
1080 		{
1081 			struct mbuf *m;
1082 			if (sopt->sopt_valsize > MLEN) {
1083 				error = EMSGSIZE;
1084 				break;
1085 			}
1086 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1087 			if (m == NULL) {
1088 				error = ENOBUFS;
1089 				break;
1090 			}
1091 			m->m_len = sopt->sopt_valsize;
1092 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1093 					    m->m_len);
1094 			if (error) {
1095 				m_free(m);
1096 				break;
1097 			}
1098 			INP_WLOCK(inp);
1099 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1100 			INP_WUNLOCK(inp);
1101 			return (error);
1102 		}
1103 
1104 		case IP_BINDANY:
1105 			if (sopt->sopt_td != NULL) {
1106 				error = priv_check(sopt->sopt_td,
1107 				    PRIV_NETINET_BINDANY);
1108 				if (error)
1109 					break;
1110 			}
1111 			/* FALLTHROUGH */
1112 		case IP_BINDMULTI:
1113 #ifdef	RSS
1114 		case IP_RSS_LISTEN_BUCKET:
1115 #endif
1116 		case IP_TOS:
1117 		case IP_TTL:
1118 		case IP_MINTTL:
1119 		case IP_RECVOPTS:
1120 		case IP_RECVRETOPTS:
1121 		case IP_ORIGDSTADDR:
1122 		case IP_RECVDSTADDR:
1123 		case IP_RECVTTL:
1124 		case IP_RECVIF:
1125 		case IP_ONESBCAST:
1126 		case IP_DONTFRAG:
1127 		case IP_RECVTOS:
1128 		case IP_RECVFLOWID:
1129 #ifdef	RSS
1130 		case IP_RECVRSSBUCKETID:
1131 #endif
1132 			error = sooptcopyin(sopt, &optval, sizeof optval,
1133 					    sizeof optval);
1134 			if (error)
1135 				break;
1136 
1137 			switch (sopt->sopt_name) {
1138 			case IP_TOS:
1139 				inp->inp_ip_tos = optval;
1140 				break;
1141 
1142 			case IP_TTL:
1143 				inp->inp_ip_ttl = optval;
1144 				break;
1145 
1146 			case IP_MINTTL:
1147 				if (optval >= 0 && optval <= MAXTTL)
1148 					inp->inp_ip_minttl = optval;
1149 				else
1150 					error = EINVAL;
1151 				break;
1152 
1153 #define	OPTSET(bit) do {						\
1154 	INP_WLOCK(inp);							\
1155 	if (optval)							\
1156 		inp->inp_flags |= bit;					\
1157 	else								\
1158 		inp->inp_flags &= ~bit;					\
1159 	INP_WUNLOCK(inp);						\
1160 } while (0)
1161 
1162 #define	OPTSET2(bit, val) do {						\
1163 	INP_WLOCK(inp);							\
1164 	if (val)							\
1165 		inp->inp_flags2 |= bit;					\
1166 	else								\
1167 		inp->inp_flags2 &= ~bit;				\
1168 	INP_WUNLOCK(inp);						\
1169 } while (0)
1170 
1171 			case IP_RECVOPTS:
1172 				OPTSET(INP_RECVOPTS);
1173 				break;
1174 
1175 			case IP_RECVRETOPTS:
1176 				OPTSET(INP_RECVRETOPTS);
1177 				break;
1178 
1179 			case IP_RECVDSTADDR:
1180 				OPTSET(INP_RECVDSTADDR);
1181 				break;
1182 
1183 			case IP_ORIGDSTADDR:
1184 				OPTSET2(INP_ORIGDSTADDR, optval);
1185 				break;
1186 
1187 			case IP_RECVTTL:
1188 				OPTSET(INP_RECVTTL);
1189 				break;
1190 
1191 			case IP_RECVIF:
1192 				OPTSET(INP_RECVIF);
1193 				break;
1194 
1195 			case IP_ONESBCAST:
1196 				OPTSET(INP_ONESBCAST);
1197 				break;
1198 			case IP_DONTFRAG:
1199 				OPTSET(INP_DONTFRAG);
1200 				break;
1201 			case IP_BINDANY:
1202 				OPTSET(INP_BINDANY);
1203 				break;
1204 			case IP_RECVTOS:
1205 				OPTSET(INP_RECVTOS);
1206 				break;
1207 			case IP_BINDMULTI:
1208 				OPTSET2(INP_BINDMULTI, optval);
1209 				break;
1210 			case IP_RECVFLOWID:
1211 				OPTSET2(INP_RECVFLOWID, optval);
1212 				break;
1213 #ifdef	RSS
1214 			case IP_RSS_LISTEN_BUCKET:
1215 				if ((optval >= 0) &&
1216 				    (optval < rss_getnumbuckets())) {
1217 					inp->inp_rss_listen_bucket = optval;
1218 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1219 				} else {
1220 					error = EINVAL;
1221 				}
1222 				break;
1223 			case IP_RECVRSSBUCKETID:
1224 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1225 				break;
1226 #endif
1227 			}
1228 			break;
1229 #undef OPTSET
1230 #undef OPTSET2
1231 
1232 		/*
1233 		 * Multicast socket options are processed by the in_mcast
1234 		 * module.
1235 		 */
1236 		case IP_MULTICAST_IF:
1237 		case IP_MULTICAST_VIF:
1238 		case IP_MULTICAST_TTL:
1239 		case IP_MULTICAST_LOOP:
1240 		case IP_ADD_MEMBERSHIP:
1241 		case IP_DROP_MEMBERSHIP:
1242 		case IP_ADD_SOURCE_MEMBERSHIP:
1243 		case IP_DROP_SOURCE_MEMBERSHIP:
1244 		case IP_BLOCK_SOURCE:
1245 		case IP_UNBLOCK_SOURCE:
1246 		case IP_MSFILTER:
1247 		case MCAST_JOIN_GROUP:
1248 		case MCAST_LEAVE_GROUP:
1249 		case MCAST_JOIN_SOURCE_GROUP:
1250 		case MCAST_LEAVE_SOURCE_GROUP:
1251 		case MCAST_BLOCK_SOURCE:
1252 		case MCAST_UNBLOCK_SOURCE:
1253 			error = inp_setmoptions(inp, sopt);
1254 			break;
1255 
1256 		case IP_PORTRANGE:
1257 			error = sooptcopyin(sopt, &optval, sizeof optval,
1258 					    sizeof optval);
1259 			if (error)
1260 				break;
1261 
1262 			INP_WLOCK(inp);
1263 			switch (optval) {
1264 			case IP_PORTRANGE_DEFAULT:
1265 				inp->inp_flags &= ~(INP_LOWPORT);
1266 				inp->inp_flags &= ~(INP_HIGHPORT);
1267 				break;
1268 
1269 			case IP_PORTRANGE_HIGH:
1270 				inp->inp_flags &= ~(INP_LOWPORT);
1271 				inp->inp_flags |= INP_HIGHPORT;
1272 				break;
1273 
1274 			case IP_PORTRANGE_LOW:
1275 				inp->inp_flags &= ~(INP_HIGHPORT);
1276 				inp->inp_flags |= INP_LOWPORT;
1277 				break;
1278 
1279 			default:
1280 				error = EINVAL;
1281 				break;
1282 			}
1283 			INP_WUNLOCK(inp);
1284 			break;
1285 
1286 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1287 		case IP_IPSEC_POLICY:
1288 			if (IPSEC_ENABLED(ipv4)) {
1289 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1290 				break;
1291 			}
1292 			/* FALLTHROUGH */
1293 #endif /* IPSEC */
1294 
1295 		default:
1296 			error = ENOPROTOOPT;
1297 			break;
1298 		}
1299 		break;
1300 
1301 	case SOPT_GET:
1302 		switch (sopt->sopt_name) {
1303 		case IP_OPTIONS:
1304 		case IP_RETOPTS:
1305 			INP_RLOCK(inp);
1306 			if (inp->inp_options) {
1307 				struct mbuf *options;
1308 
1309 				options = m_copym(inp->inp_options, 0,
1310 				    M_COPYALL, M_NOWAIT);
1311 				INP_RUNLOCK(inp);
1312 				if (options != NULL) {
1313 					error = sooptcopyout(sopt,
1314 							     mtod(options, char *),
1315 							     options->m_len);
1316 					m_freem(options);
1317 				} else
1318 					error = ENOMEM;
1319 			} else {
1320 				INP_RUNLOCK(inp);
1321 				sopt->sopt_valsize = 0;
1322 			}
1323 			break;
1324 
1325 		case IP_TOS:
1326 		case IP_TTL:
1327 		case IP_MINTTL:
1328 		case IP_RECVOPTS:
1329 		case IP_RECVRETOPTS:
1330 		case IP_ORIGDSTADDR:
1331 		case IP_RECVDSTADDR:
1332 		case IP_RECVTTL:
1333 		case IP_RECVIF:
1334 		case IP_PORTRANGE:
1335 		case IP_ONESBCAST:
1336 		case IP_DONTFRAG:
1337 		case IP_BINDANY:
1338 		case IP_RECVTOS:
1339 		case IP_BINDMULTI:
1340 		case IP_FLOWID:
1341 		case IP_FLOWTYPE:
1342 		case IP_RECVFLOWID:
1343 #ifdef	RSS
1344 		case IP_RSSBUCKETID:
1345 		case IP_RECVRSSBUCKETID:
1346 #endif
1347 			switch (sopt->sopt_name) {
1348 
1349 			case IP_TOS:
1350 				optval = inp->inp_ip_tos;
1351 				break;
1352 
1353 			case IP_TTL:
1354 				optval = inp->inp_ip_ttl;
1355 				break;
1356 
1357 			case IP_MINTTL:
1358 				optval = inp->inp_ip_minttl;
1359 				break;
1360 
1361 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1362 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1363 
1364 			case IP_RECVOPTS:
1365 				optval = OPTBIT(INP_RECVOPTS);
1366 				break;
1367 
1368 			case IP_RECVRETOPTS:
1369 				optval = OPTBIT(INP_RECVRETOPTS);
1370 				break;
1371 
1372 			case IP_RECVDSTADDR:
1373 				optval = OPTBIT(INP_RECVDSTADDR);
1374 				break;
1375 
1376 			case IP_ORIGDSTADDR:
1377 				optval = OPTBIT2(INP_ORIGDSTADDR);
1378 				break;
1379 
1380 			case IP_RECVTTL:
1381 				optval = OPTBIT(INP_RECVTTL);
1382 				break;
1383 
1384 			case IP_RECVIF:
1385 				optval = OPTBIT(INP_RECVIF);
1386 				break;
1387 
1388 			case IP_PORTRANGE:
1389 				if (inp->inp_flags & INP_HIGHPORT)
1390 					optval = IP_PORTRANGE_HIGH;
1391 				else if (inp->inp_flags & INP_LOWPORT)
1392 					optval = IP_PORTRANGE_LOW;
1393 				else
1394 					optval = 0;
1395 				break;
1396 
1397 			case IP_ONESBCAST:
1398 				optval = OPTBIT(INP_ONESBCAST);
1399 				break;
1400 			case IP_DONTFRAG:
1401 				optval = OPTBIT(INP_DONTFRAG);
1402 				break;
1403 			case IP_BINDANY:
1404 				optval = OPTBIT(INP_BINDANY);
1405 				break;
1406 			case IP_RECVTOS:
1407 				optval = OPTBIT(INP_RECVTOS);
1408 				break;
1409 			case IP_FLOWID:
1410 				optval = inp->inp_flowid;
1411 				break;
1412 			case IP_FLOWTYPE:
1413 				optval = inp->inp_flowtype;
1414 				break;
1415 			case IP_RECVFLOWID:
1416 				optval = OPTBIT2(INP_RECVFLOWID);
1417 				break;
1418 #ifdef	RSS
1419 			case IP_RSSBUCKETID:
1420 				retval = rss_hash2bucket(inp->inp_flowid,
1421 				    inp->inp_flowtype,
1422 				    &rss_bucket);
1423 				if (retval == 0)
1424 					optval = rss_bucket;
1425 				else
1426 					error = EINVAL;
1427 				break;
1428 			case IP_RECVRSSBUCKETID:
1429 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1430 				break;
1431 #endif
1432 			case IP_BINDMULTI:
1433 				optval = OPTBIT2(INP_BINDMULTI);
1434 				break;
1435 			}
1436 			error = sooptcopyout(sopt, &optval, sizeof optval);
1437 			break;
1438 
1439 		/*
1440 		 * Multicast socket options are processed by the in_mcast
1441 		 * module.
1442 		 */
1443 		case IP_MULTICAST_IF:
1444 		case IP_MULTICAST_VIF:
1445 		case IP_MULTICAST_TTL:
1446 		case IP_MULTICAST_LOOP:
1447 		case IP_MSFILTER:
1448 			error = inp_getmoptions(inp, sopt);
1449 			break;
1450 
1451 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1452 		case IP_IPSEC_POLICY:
1453 			if (IPSEC_ENABLED(ipv4)) {
1454 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1455 				break;
1456 			}
1457 			/* FALLTHROUGH */
1458 #endif /* IPSEC */
1459 
1460 		default:
1461 			error = ENOPROTOOPT;
1462 			break;
1463 		}
1464 		break;
1465 	}
1466 	return (error);
1467 }
1468 
1469 /*
1470  * Routine called from ip_output() to loop back a copy of an IP multicast
1471  * packet to the input queue of a specified interface.  Note that this
1472  * calls the output routine of the loopback "driver", but with an interface
1473  * pointer that might NOT be a loopback interface -- evil, but easier than
1474  * replicating that code here.
1475  */
1476 static void
1477 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1478 {
1479 	struct ip *ip;
1480 	struct mbuf *copym;
1481 
1482 	/*
1483 	 * Make a deep copy of the packet because we're going to
1484 	 * modify the pack in order to generate checksums.
1485 	 */
1486 	copym = m_dup(m, M_NOWAIT);
1487 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1488 		copym = m_pullup(copym, hlen);
1489 	if (copym != NULL) {
1490 		/* If needed, compute the checksum and mark it as valid. */
1491 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1492 			in_delayed_cksum(copym);
1493 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1494 			copym->m_pkthdr.csum_flags |=
1495 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1496 			copym->m_pkthdr.csum_data = 0xffff;
1497 		}
1498 		/*
1499 		 * We don't bother to fragment if the IP length is greater
1500 		 * than the interface's MTU.  Can this possibly matter?
1501 		 */
1502 		ip = mtod(copym, struct ip *);
1503 		ip->ip_sum = 0;
1504 		ip->ip_sum = in_cksum(copym, hlen);
1505 		if_simloop(ifp, copym, AF_INET, 0);
1506 	}
1507 }
1508