xref: /freebsd-13.1/sys/netinet/ip_output.c (revision bf2a85f4)
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_ipsec.h"
39 #include "opt_kern_tls.h"
40 #include "opt_mbuf_stress_test.h"
41 #include "opt_ratelimit.h"
42 #include "opt_route.h"
43 #include "opt_rss.h"
44 #include "opt_sctp.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
49 #include <sys/ktls.h>
50 #include <sys/lock.h>
51 #include <sys/malloc.h>
52 #include <sys/mbuf.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/protosw.h>
56 #include <sys/rmlock.h>
57 #include <sys/sdt.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
60 #include <sys/sysctl.h>
61 #include <sys/ucred.h>
62 
63 #include <net/if.h>
64 #include <net/if_var.h>
65 #include <net/if_vlan_var.h>
66 #include <net/if_llatbl.h>
67 #include <net/ethernet.h>
68 #include <net/netisr.h>
69 #include <net/pfil.h>
70 #include <net/route.h>
71 #include <net/route/nhop.h>
72 #include <net/rss_config.h>
73 #include <net/vnet.h>
74 
75 #include <netinet/in.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_kdtrace.h>
78 #include <netinet/in_systm.h>
79 #include <netinet/ip.h>
80 #include <netinet/in_fib.h>
81 #include <netinet/in_pcb.h>
82 #include <netinet/in_rss.h>
83 #include <netinet/in_var.h>
84 #include <netinet/ip_var.h>
85 #include <netinet/ip_options.h>
86 
87 #include <netinet/udp.h>
88 #include <netinet/udp_var.h>
89 
90 #if defined(SCTP) || defined(SCTP_SUPPORT)
91 #include <netinet/sctp.h>
92 #include <netinet/sctp_crc32.h>
93 #endif
94 
95 #include <netipsec/ipsec_support.h>
96 
97 #include <machine/in_cksum.h>
98 
99 #include <security/mac/mac_framework.h>
100 
101 #ifdef MBUF_STRESS_TEST
102 static int mbuf_frag_size = 0;
103 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
104 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
105 #endif
106 
107 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
108 
109 extern int in_mcast_loop;
110 extern	struct protosw inetsw[];
111 
112 static inline int
ip_output_pfil(struct mbuf ** mp,struct ifnet * ifp,int flags,struct inpcb * inp,struct sockaddr_in * dst,int * fibnum,int * error)113 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
114     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
115 {
116 	struct m_tag *fwd_tag = NULL;
117 	struct mbuf *m;
118 	struct in_addr odst;
119 	struct ip *ip;
120 	int pflags = PFIL_OUT;
121 
122 	if (flags & IP_FORWARDING)
123 		pflags |= PFIL_FWD;
124 
125 	m = *mp;
126 	ip = mtod(m, struct ip *);
127 
128 	/* Run through list of hooks for output packets. */
129 	odst.s_addr = ip->ip_dst.s_addr;
130 	switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
131 	case PFIL_DROPPED:
132 		*error = EACCES;
133 		/* FALLTHROUGH */
134 	case PFIL_CONSUMED:
135 		return 1; /* Finished */
136 	case PFIL_PASS:
137 		*error = 0;
138 	}
139 	m = *mp;
140 	ip = mtod(m, struct ip *);
141 
142 	/* See if destination IP address was changed by packet filter. */
143 	if (odst.s_addr != ip->ip_dst.s_addr) {
144 		m->m_flags |= M_SKIP_FIREWALL;
145 		/* If destination is now ourself drop to ip_input(). */
146 		if (in_localip(ip->ip_dst)) {
147 			m->m_flags |= M_FASTFWD_OURS;
148 			if (m->m_pkthdr.rcvif == NULL)
149 				m->m_pkthdr.rcvif = V_loif;
150 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
151 				m->m_pkthdr.csum_flags |=
152 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
153 				m->m_pkthdr.csum_data = 0xffff;
154 			}
155 			m->m_pkthdr.csum_flags |=
156 				CSUM_IP_CHECKED | CSUM_IP_VALID;
157 #if defined(SCTP) || defined(SCTP_SUPPORT)
158 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
159 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
160 #endif
161 			*error = netisr_queue(NETISR_IP, m);
162 			return 1; /* Finished */
163 		}
164 
165 		bzero(dst, sizeof(*dst));
166 		dst->sin_family = AF_INET;
167 		dst->sin_len = sizeof(*dst);
168 		dst->sin_addr = ip->ip_dst;
169 
170 		return -1; /* Reloop */
171 	}
172 	/* See if fib was changed by packet filter. */
173 	if ((*fibnum) != M_GETFIB(m)) {
174 		m->m_flags |= M_SKIP_FIREWALL;
175 		*fibnum = M_GETFIB(m);
176 		return -1; /* Reloop for FIB change */
177 	}
178 
179 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
180 	if (m->m_flags & M_FASTFWD_OURS) {
181 		if (m->m_pkthdr.rcvif == NULL)
182 			m->m_pkthdr.rcvif = V_loif;
183 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
184 			m->m_pkthdr.csum_flags |=
185 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
186 			m->m_pkthdr.csum_data = 0xffff;
187 		}
188 #if defined(SCTP) || defined(SCTP_SUPPORT)
189 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
190 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
191 #endif
192 		m->m_pkthdr.csum_flags |=
193 			CSUM_IP_CHECKED | CSUM_IP_VALID;
194 
195 		*error = netisr_queue(NETISR_IP, m);
196 		return 1; /* Finished */
197 	}
198 	/* Or forward to some other address? */
199 	if ((m->m_flags & M_IP_NEXTHOP) &&
200 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
201 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
202 		m->m_flags |= M_SKIP_FIREWALL;
203 		m->m_flags &= ~M_IP_NEXTHOP;
204 		m_tag_delete(m, fwd_tag);
205 
206 		return -1; /* Reloop for CHANGE of dst */
207 	}
208 
209 	return 0;
210 }
211 
212 static int
ip_output_send(struct inpcb * inp,struct ifnet * ifp,struct mbuf * m,const struct sockaddr * gw,struct route * ro,bool stamp_tag)213 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
214     const struct sockaddr *gw, struct route *ro, bool stamp_tag)
215 {
216 #ifdef KERN_TLS
217 	struct ktls_session *tls = NULL;
218 #endif
219 	struct m_snd_tag *mst;
220 	int error;
221 
222 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
223 	mst = NULL;
224 
225 #ifdef KERN_TLS
226 	/*
227 	 * If this is an unencrypted TLS record, save a reference to
228 	 * the record.  This local reference is used to call
229 	 * ktls_output_eagain after the mbuf has been freed (thus
230 	 * dropping the mbuf's reference) in if_output.
231 	 */
232 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
233 		tls = ktls_hold(m->m_next->m_epg_tls);
234 		mst = tls->snd_tag;
235 
236 		/*
237 		 * If a TLS session doesn't have a valid tag, it must
238 		 * have had an earlier ifp mismatch, so drop this
239 		 * packet.
240 		 */
241 		if (mst == NULL) {
242 			m_freem(m);
243 			error = EAGAIN;
244 			goto done;
245 		}
246 		/*
247 		 * Always stamp tags that include NIC ktls.
248 		 */
249 		stamp_tag = true;
250 	}
251 #endif
252 #ifdef RATELIMIT
253 	if (inp != NULL && mst == NULL) {
254 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
255 		    (inp->inp_snd_tag != NULL &&
256 		    inp->inp_snd_tag->ifp != ifp))
257 			in_pcboutput_txrtlmt(inp, ifp, m);
258 
259 		if (inp->inp_snd_tag != NULL)
260 			mst = inp->inp_snd_tag;
261 	}
262 #endif
263 	if (stamp_tag && mst != NULL) {
264 		KASSERT(m->m_pkthdr.rcvif == NULL,
265 		    ("trying to add a send tag to a forwarded packet"));
266 		if (mst->ifp != ifp) {
267 			m_freem(m);
268 			error = EAGAIN;
269 			goto done;
270 		}
271 
272 		/* stamp send tag on mbuf */
273 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
274 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
275 	}
276 
277 	error = (*ifp->if_output)(ifp, m, gw, ro);
278 
279 done:
280 	/* Check for route change invalidating send tags. */
281 #ifdef KERN_TLS
282 	if (tls != NULL) {
283 		if (error == EAGAIN)
284 			error = ktls_output_eagain(inp, tls);
285 		ktls_free(tls);
286 	}
287 #endif
288 #ifdef RATELIMIT
289 	if (error == EAGAIN)
290 		in_pcboutput_eagain(inp);
291 #endif
292 	return (error);
293 }
294 
295 /* rte<>ro_flags translation */
296 static inline void
rt_update_ro_flags(struct route * ro,const struct nhop_object * nh)297 rt_update_ro_flags(struct route *ro, const struct nhop_object *nh)
298 {
299 	int nh_flags = nh->nh_flags;
300 
301 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
302 
303 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
304 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
305 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
306 }
307 
308 /*
309  * IP output.  The packet in mbuf chain m contains a skeletal IP
310  * header (with len, off, ttl, proto, tos, src, dst).
311  * The mbuf chain containing the packet will be freed.
312  * The mbuf opt, if present, will not be freed.
313  * If route ro is present and has ro_rt initialized, route lookup would be
314  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
315  * then result of route lookup is stored in ro->ro_rt.
316  *
317  * In the IP forwarding case, the packet will arrive with options already
318  * inserted, so must have a NULL opt pointer.
319  */
320 int
ip_output(struct mbuf * m,struct mbuf * opt,struct route * ro,int flags,struct ip_moptions * imo,struct inpcb * inp)321 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
322     struct ip_moptions *imo, struct inpcb *inp)
323 {
324 	struct rm_priotracker in_ifa_tracker;
325 	struct ip *ip;
326 	struct ifnet *ifp = NULL;	/* keep compiler happy */
327 	struct mbuf *m0;
328 	int hlen = sizeof (struct ip);
329 	int mtu = 0;
330 	int error = 0;
331 	int vlan_pcp = -1;
332 	struct sockaddr_in *dst;
333 	const struct sockaddr *gw;
334 	struct in_ifaddr *ia = NULL;
335 	struct in_addr src;
336 	int isbroadcast;
337 	uint16_t ip_len, ip_off;
338 	struct route iproute;
339 	uint32_t fibnum;
340 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
341 	int no_route_but_check_spd = 0;
342 #endif
343 
344 	M_ASSERTPKTHDR(m);
345 	NET_EPOCH_ASSERT();
346 
347 	if (inp != NULL) {
348 		INP_LOCK_ASSERT(inp);
349 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
350 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
351 			m->m_pkthdr.flowid = inp->inp_flowid;
352 			M_HASHTYPE_SET(m, inp->inp_flowtype);
353 		}
354 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
355 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
356 			    INP_2PCP_SHIFT;
357 #ifdef NUMA
358 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
359 #endif
360 	}
361 
362 	if (opt) {
363 		int len = 0;
364 		m = ip_insertoptions(m, opt, &len);
365 		if (len != 0)
366 			hlen = len; /* ip->ip_hl is updated above */
367 	}
368 	ip = mtod(m, struct ip *);
369 	ip_len = ntohs(ip->ip_len);
370 	ip_off = ntohs(ip->ip_off);
371 
372 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
373 		ip->ip_v = IPVERSION;
374 		ip->ip_hl = hlen >> 2;
375 		ip_fillid(ip);
376 	} else {
377 		/* Header already set, fetch hlen from there */
378 		hlen = ip->ip_hl << 2;
379 	}
380 	if ((flags & IP_FORWARDING) == 0)
381 		IPSTAT_INC(ips_localout);
382 
383 	/*
384 	 * dst/gw handling:
385 	 *
386 	 * gw is readonly but can point either to dst OR rt_gateway,
387 	 * therefore we need restore gw if we're redoing lookup.
388 	 */
389 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
390 	if (ro == NULL) {
391 		ro = &iproute;
392 		bzero(ro, sizeof (*ro));
393 	}
394 	dst = (struct sockaddr_in *)&ro->ro_dst;
395 	if (ro->ro_nh == NULL) {
396 		dst->sin_family = AF_INET;
397 		dst->sin_len = sizeof(*dst);
398 		dst->sin_addr = ip->ip_dst;
399 	}
400 	gw = (const struct sockaddr *)dst;
401 again:
402 	/*
403 	 * Validate route against routing table additions;
404 	 * a better/more specific route might have been added.
405 	 */
406 	if (inp != NULL && ro->ro_nh != NULL)
407 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
408 	/*
409 	 * If there is a cached route,
410 	 * check that it is to the same destination
411 	 * and is still up.  If not, free it and try again.
412 	 * The address family should also be checked in case of sharing the
413 	 * cache with IPv6.
414 	 * Also check whether routing cache needs invalidation.
415 	 */
416 	if (ro->ro_nh != NULL &&
417 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
418 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
419 		RO_INVALIDATE_CACHE(ro);
420 	ia = NULL;
421 	/*
422 	 * If routing to interface only, short circuit routing lookup.
423 	 * The use of an all-ones broadcast address implies this; an
424 	 * interface is specified by the broadcast address of an interface,
425 	 * or the destination address of a ptp interface.
426 	 */
427 	if (flags & IP_SENDONES) {
428 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
429 						      M_GETFIB(m)))) == NULL &&
430 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
431 						    M_GETFIB(m)))) == NULL) {
432 			IPSTAT_INC(ips_noroute);
433 			error = ENETUNREACH;
434 			goto bad;
435 		}
436 		ip->ip_dst.s_addr = INADDR_BROADCAST;
437 		dst->sin_addr = ip->ip_dst;
438 		ifp = ia->ia_ifp;
439 		mtu = ifp->if_mtu;
440 		ip->ip_ttl = 1;
441 		isbroadcast = 1;
442 		src = IA_SIN(ia)->sin_addr;
443 	} else if (flags & IP_ROUTETOIF) {
444 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
445 						    M_GETFIB(m)))) == NULL &&
446 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
447 						M_GETFIB(m)))) == NULL) {
448 			IPSTAT_INC(ips_noroute);
449 			error = ENETUNREACH;
450 			goto bad;
451 		}
452 		ifp = ia->ia_ifp;
453 		mtu = ifp->if_mtu;
454 		ip->ip_ttl = 1;
455 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
456 		    in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
457 		src = IA_SIN(ia)->sin_addr;
458 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
459 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
460 		/*
461 		 * Bypass the normal routing lookup for multicast
462 		 * packets if the interface is specified.
463 		 */
464 		ifp = imo->imo_multicast_ifp;
465 		mtu = ifp->if_mtu;
466 		IFP_TO_IA(ifp, ia, &in_ifa_tracker);
467 		isbroadcast = 0;	/* fool gcc */
468 		/* Interface may have no addresses. */
469 		if (ia != NULL)
470 			src = IA_SIN(ia)->sin_addr;
471 		else
472 			src.s_addr = INADDR_ANY;
473 	} else if (ro != &iproute) {
474 		if (ro->ro_nh == NULL) {
475 			/*
476 			 * We want to do any cloning requested by the link
477 			 * layer, as this is probably required in all cases
478 			 * for correct operation (as it is for ARP).
479 			 */
480 			uint32_t flowid;
481 			flowid = m->m_pkthdr.flowid;
482 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
483 			    NHR_REF, flowid);
484 
485 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
486 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
487 				/*
488 				 * There is no route for this packet, but it is
489 				 * possible that a matching SPD entry exists.
490 				 */
491 				no_route_but_check_spd = 1;
492 				goto sendit;
493 #endif
494 				IPSTAT_INC(ips_noroute);
495 				error = EHOSTUNREACH;
496 				goto bad;
497 			}
498 		}
499 		struct nhop_object *nh = ro->ro_nh;
500 
501 		ia = ifatoia(nh->nh_ifa);
502 		ifp = nh->nh_ifp;
503 		counter_u64_add(nh->nh_pksent, 1);
504 		rt_update_ro_flags(ro, nh);
505 		if (nh->nh_flags & NHF_GATEWAY)
506 			gw = &nh->gw_sa;
507 		if (nh->nh_flags & NHF_HOST)
508 			isbroadcast = (nh->nh_flags & NHF_BROADCAST);
509 		else if ((ifp->if_flags & IFF_BROADCAST) && (gw->sa_family == AF_INET))
510 			isbroadcast = in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia);
511 		else
512 			isbroadcast = 0;
513 		mtu = nh->nh_mtu;
514 		src = IA_SIN(ia)->sin_addr;
515 	} else {
516 		struct nhop_object *nh;
517 
518 		nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE,
519 		    m->m_pkthdr.flowid);
520 		if (nh == NULL) {
521 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
522 			/*
523 			 * There is no route for this packet, but it is
524 			 * possible that a matching SPD entry exists.
525 			 */
526 			no_route_but_check_spd = 1;
527 			goto sendit;
528 #endif
529 			IPSTAT_INC(ips_noroute);
530 			error = EHOSTUNREACH;
531 			goto bad;
532 		}
533 		ifp = nh->nh_ifp;
534 		mtu = nh->nh_mtu;
535 		rt_update_ro_flags(ro, nh);
536 		if (nh->nh_flags & NHF_GATEWAY)
537 			gw = &nh->gw_sa;
538 		ia = ifatoia(nh->nh_ifa);
539 		src = IA_SIN(ia)->sin_addr;
540 		isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
541 		    (NHF_HOST | NHF_BROADCAST)) ||
542 		    ((ifp->if_flags & IFF_BROADCAST) &&
543 		    (gw->sa_family == AF_INET) &&
544 		    in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia)));
545 	}
546 
547 	/* Catch a possible divide by zero later. */
548 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
549 	    __func__, mtu, ro,
550 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
551 
552 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
553 		m->m_flags |= M_MCAST;
554 		/*
555 		 * IP destination address is multicast.  Make sure "gw"
556 		 * still points to the address in "ro".  (It may have been
557 		 * changed to point to a gateway address, above.)
558 		 */
559 		gw = (const struct sockaddr *)dst;
560 		/*
561 		 * See if the caller provided any multicast options
562 		 */
563 		if (imo != NULL) {
564 			ip->ip_ttl = imo->imo_multicast_ttl;
565 			if (imo->imo_multicast_vif != -1)
566 				ip->ip_src.s_addr =
567 				    ip_mcast_src ?
568 				    ip_mcast_src(imo->imo_multicast_vif) :
569 				    INADDR_ANY;
570 		} else
571 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
572 		/*
573 		 * Confirm that the outgoing interface supports multicast.
574 		 */
575 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
576 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
577 				IPSTAT_INC(ips_noroute);
578 				error = ENETUNREACH;
579 				goto bad;
580 			}
581 		}
582 		/*
583 		 * If source address not specified yet, use address
584 		 * of outgoing interface.
585 		 */
586 		if (ip->ip_src.s_addr == INADDR_ANY)
587 			ip->ip_src = src;
588 
589 		if ((imo == NULL && in_mcast_loop) ||
590 		    (imo && imo->imo_multicast_loop)) {
591 			/*
592 			 * Loop back multicast datagram if not expressly
593 			 * forbidden to do so, even if we are not a member
594 			 * of the group; ip_input() will filter it later,
595 			 * thus deferring a hash lookup and mutex acquisition
596 			 * at the expense of a cheap copy using m_copym().
597 			 */
598 			ip_mloopback(ifp, m, hlen);
599 		} else {
600 			/*
601 			 * If we are acting as a multicast router, perform
602 			 * multicast forwarding as if the packet had just
603 			 * arrived on the interface to which we are about
604 			 * to send.  The multicast forwarding function
605 			 * recursively calls this function, using the
606 			 * IP_FORWARDING flag to prevent infinite recursion.
607 			 *
608 			 * Multicasts that are looped back by ip_mloopback(),
609 			 * above, will be forwarded by the ip_input() routine,
610 			 * if necessary.
611 			 */
612 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
613 				/*
614 				 * If rsvp daemon is not running, do not
615 				 * set ip_moptions. This ensures that the packet
616 				 * is multicast and not just sent down one link
617 				 * as prescribed by rsvpd.
618 				 */
619 				if (!V_rsvp_on)
620 					imo = NULL;
621 				if (ip_mforward &&
622 				    ip_mforward(ip, ifp, m, imo) != 0) {
623 					m_freem(m);
624 					goto done;
625 				}
626 			}
627 		}
628 
629 		/*
630 		 * Multicasts with a time-to-live of zero may be looped-
631 		 * back, above, but must not be transmitted on a network.
632 		 * Also, multicasts addressed to the loopback interface
633 		 * are not sent -- the above call to ip_mloopback() will
634 		 * loop back a copy. ip_input() will drop the copy if
635 		 * this host does not belong to the destination group on
636 		 * the loopback interface.
637 		 */
638 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
639 			m_freem(m);
640 			goto done;
641 		}
642 
643 		goto sendit;
644 	}
645 
646 	/*
647 	 * If the source address is not specified yet, use the address
648 	 * of the outoing interface.
649 	 */
650 	if (ip->ip_src.s_addr == INADDR_ANY)
651 		ip->ip_src = src;
652 
653 	/*
654 	 * Look for broadcast address and
655 	 * verify user is allowed to send
656 	 * such a packet.
657 	 */
658 	if (isbroadcast) {
659 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
660 			error = EADDRNOTAVAIL;
661 			goto bad;
662 		}
663 		if ((flags & IP_ALLOWBROADCAST) == 0) {
664 			error = EACCES;
665 			goto bad;
666 		}
667 		/* don't allow broadcast messages to be fragmented */
668 		if (ip_len > mtu) {
669 			error = EMSGSIZE;
670 			goto bad;
671 		}
672 		m->m_flags |= M_BCAST;
673 	} else {
674 		m->m_flags &= ~M_BCAST;
675 	}
676 
677 sendit:
678 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
679 	if (IPSEC_ENABLED(ipv4)) {
680 		if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
681 			if (error == EINPROGRESS)
682 				error = 0;
683 			goto done;
684 		}
685 	}
686 	/*
687 	 * Check if there was a route for this packet; return error if not.
688 	 */
689 	if (no_route_but_check_spd) {
690 		IPSTAT_INC(ips_noroute);
691 		error = EHOSTUNREACH;
692 		goto bad;
693 	}
694 	/* Update variables that are affected by ipsec4_output(). */
695 	ip = mtod(m, struct ip *);
696 	hlen = ip->ip_hl << 2;
697 #endif /* IPSEC */
698 
699 	/* Jump over all PFIL processing if hooks are not active. */
700 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
701 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
702 		    &error)) {
703 		case 1: /* Finished */
704 			goto done;
705 
706 		case 0: /* Continue normally */
707 			ip = mtod(m, struct ip *);
708 			break;
709 
710 		case -1: /* Need to try again */
711 			/* Reset everything for a new round */
712 			if (ro != NULL) {
713 				RO_NHFREE(ro);
714 				ro->ro_prepend = NULL;
715 			}
716 			gw = (const struct sockaddr *)dst;
717 			ip = mtod(m, struct ip *);
718 			goto again;
719 		}
720 	}
721 
722 	if (vlan_pcp > -1)
723 		EVL_APPLY_PRI(m, vlan_pcp);
724 
725 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
726 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
727 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
728 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
729 			IPSTAT_INC(ips_badaddr);
730 			error = EADDRNOTAVAIL;
731 			goto bad;
732 		}
733 	}
734 
735 	/* Ensure the packet data is mapped if the interface requires it. */
736 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
737 		m = mb_unmapped_to_ext(m);
738 		if (m == NULL) {
739 			IPSTAT_INC(ips_odropped);
740 			error = ENOBUFS;
741 			goto bad;
742 		}
743 	}
744 
745 	m->m_pkthdr.csum_flags |= CSUM_IP;
746 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
747 		in_delayed_cksum(m);
748 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
749 	}
750 #if defined(SCTP) || defined(SCTP_SUPPORT)
751 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
752 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
753 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
754 	}
755 #endif
756 
757 	/*
758 	 * If small enough for interface, or the interface will take
759 	 * care of the fragmentation for us, we can just send directly.
760 	 * Note that if_vxlan could have requested TSO even though the outer
761 	 * frame is UDP.  It is correct to not fragment such datagrams and
762 	 * instead just pass them on to the driver.
763 	 */
764 	if (ip_len <= mtu ||
765 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
766 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
767 		ip->ip_sum = 0;
768 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
769 			ip->ip_sum = in_cksum(m, hlen);
770 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
771 		}
772 
773 		/*
774 		 * Record statistics for this interface address.
775 		 * With CSUM_TSO the byte/packet count will be slightly
776 		 * incorrect because we count the IP+TCP headers only
777 		 * once instead of for every generated packet.
778 		 */
779 		if (!(flags & IP_FORWARDING) && ia) {
780 			if (m->m_pkthdr.csum_flags &
781 			    (CSUM_TSO | CSUM_INNER_TSO))
782 				counter_u64_add(ia->ia_ifa.ifa_opackets,
783 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
784 			else
785 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
786 
787 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
788 		}
789 #ifdef MBUF_STRESS_TEST
790 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
791 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
792 #endif
793 		/*
794 		 * Reset layer specific mbuf flags
795 		 * to avoid confusing lower layers.
796 		 */
797 		m_clrprotoflags(m);
798 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
799 		error = ip_output_send(inp, ifp, m, gw, ro,
800 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
801 		goto done;
802 	}
803 
804 	/* Balk when DF bit is set or the interface didn't support TSO. */
805 	if ((ip_off & IP_DF) ||
806 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
807 		error = EMSGSIZE;
808 		IPSTAT_INC(ips_cantfrag);
809 		goto bad;
810 	}
811 
812 	/*
813 	 * Too large for interface; fragment if possible. If successful,
814 	 * on return, m will point to a list of packets to be sent.
815 	 */
816 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
817 	if (error)
818 		goto bad;
819 	for (; m; m = m0) {
820 		m0 = m->m_nextpkt;
821 		m->m_nextpkt = 0;
822 		if (error == 0) {
823 			/* Record statistics for this interface address. */
824 			if (ia != NULL) {
825 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
826 				counter_u64_add(ia->ia_ifa.ifa_obytes,
827 				    m->m_pkthdr.len);
828 			}
829 			/*
830 			 * Reset layer specific mbuf flags
831 			 * to avoid confusing upper layers.
832 			 */
833 			m_clrprotoflags(m);
834 
835 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
836 			    mtod(m, struct ip *), NULL);
837 			error = ip_output_send(inp, ifp, m, gw, ro, true);
838 		} else
839 			m_freem(m);
840 	}
841 
842 	if (error == 0)
843 		IPSTAT_INC(ips_fragmented);
844 
845 done:
846 	return (error);
847  bad:
848 	m_freem(m);
849 	goto done;
850 }
851 
852 /*
853  * Create a chain of fragments which fit the given mtu. m_frag points to the
854  * mbuf to be fragmented; on return it points to the chain with the fragments.
855  * Return 0 if no error. If error, m_frag may contain a partially built
856  * chain of fragments that should be freed by the caller.
857  *
858  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
859  */
860 int
ip_fragment(struct ip * ip,struct mbuf ** m_frag,int mtu,u_long if_hwassist_flags)861 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
862     u_long if_hwassist_flags)
863 {
864 	int error = 0;
865 	int hlen = ip->ip_hl << 2;
866 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
867 	int off;
868 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
869 	int firstlen;
870 	struct mbuf **mnext;
871 	int nfrags;
872 	uint16_t ip_len, ip_off;
873 
874 	ip_len = ntohs(ip->ip_len);
875 	ip_off = ntohs(ip->ip_off);
876 
877 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
878 		IPSTAT_INC(ips_cantfrag);
879 		return EMSGSIZE;
880 	}
881 
882 	/*
883 	 * Must be able to put at least 8 bytes per fragment.
884 	 */
885 	if (len < 8)
886 		return EMSGSIZE;
887 
888 	/*
889 	 * If the interface will not calculate checksums on
890 	 * fragmented packets, then do it here.
891 	 */
892 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
893 		in_delayed_cksum(m0);
894 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
895 	}
896 #if defined(SCTP) || defined(SCTP_SUPPORT)
897 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
898 		sctp_delayed_cksum(m0, hlen);
899 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
900 	}
901 #endif
902 	if (len > PAGE_SIZE) {
903 		/*
904 		 * Fragment large datagrams such that each segment
905 		 * contains a multiple of PAGE_SIZE amount of data,
906 		 * plus headers. This enables a receiver to perform
907 		 * page-flipping zero-copy optimizations.
908 		 *
909 		 * XXX When does this help given that sender and receiver
910 		 * could have different page sizes, and also mtu could
911 		 * be less than the receiver's page size ?
912 		 */
913 		int newlen;
914 
915 		off = MIN(mtu, m0->m_pkthdr.len);
916 
917 		/*
918 		 * firstlen (off - hlen) must be aligned on an
919 		 * 8-byte boundary
920 		 */
921 		if (off < hlen)
922 			goto smart_frag_failure;
923 		off = ((off - hlen) & ~7) + hlen;
924 		newlen = (~PAGE_MASK) & mtu;
925 		if ((newlen + sizeof (struct ip)) > mtu) {
926 			/* we failed, go back the default */
927 smart_frag_failure:
928 			newlen = len;
929 			off = hlen + len;
930 		}
931 		len = newlen;
932 
933 	} else {
934 		off = hlen + len;
935 	}
936 
937 	firstlen = off - hlen;
938 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
939 
940 	/*
941 	 * Loop through length of segment after first fragment,
942 	 * make new header and copy data of each part and link onto chain.
943 	 * Here, m0 is the original packet, m is the fragment being created.
944 	 * The fragments are linked off the m_nextpkt of the original
945 	 * packet, which after processing serves as the first fragment.
946 	 */
947 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
948 		struct ip *mhip;	/* ip header on the fragment */
949 		struct mbuf *m;
950 		int mhlen = sizeof (struct ip);
951 
952 		m = m_gethdr(M_NOWAIT, MT_DATA);
953 		if (m == NULL) {
954 			error = ENOBUFS;
955 			IPSTAT_INC(ips_odropped);
956 			goto done;
957 		}
958 		/*
959 		 * Make sure the complete packet header gets copied
960 		 * from the originating mbuf to the newly created
961 		 * mbuf. This also ensures that existing firewall
962 		 * classification(s), VLAN tags and so on get copied
963 		 * to the resulting fragmented packet(s):
964 		 */
965 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
966 			m_free(m);
967 			error = ENOBUFS;
968 			IPSTAT_INC(ips_odropped);
969 			goto done;
970 		}
971 		/*
972 		 * In the first mbuf, leave room for the link header, then
973 		 * copy the original IP header including options. The payload
974 		 * goes into an additional mbuf chain returned by m_copym().
975 		 */
976 		m->m_data += max_linkhdr;
977 		mhip = mtod(m, struct ip *);
978 		*mhip = *ip;
979 		if (hlen > sizeof (struct ip)) {
980 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
981 			mhip->ip_v = IPVERSION;
982 			mhip->ip_hl = mhlen >> 2;
983 		}
984 		m->m_len = mhlen;
985 		/* XXX do we need to add ip_off below ? */
986 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
987 		if (off + len >= ip_len)
988 			len = ip_len - off;
989 		else
990 			mhip->ip_off |= IP_MF;
991 		mhip->ip_len = htons((u_short)(len + mhlen));
992 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
993 		if (m->m_next == NULL) {	/* copy failed */
994 			m_free(m);
995 			error = ENOBUFS;	/* ??? */
996 			IPSTAT_INC(ips_odropped);
997 			goto done;
998 		}
999 		m->m_pkthdr.len = mhlen + len;
1000 #ifdef MAC
1001 		mac_netinet_fragment(m0, m);
1002 #endif
1003 		mhip->ip_off = htons(mhip->ip_off);
1004 		mhip->ip_sum = 0;
1005 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1006 			mhip->ip_sum = in_cksum(m, mhlen);
1007 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1008 		}
1009 		*mnext = m;
1010 		mnext = &m->m_nextpkt;
1011 	}
1012 	IPSTAT_ADD(ips_ofragments, nfrags);
1013 
1014 	/*
1015 	 * Update first fragment by trimming what's been copied out
1016 	 * and updating header.
1017 	 */
1018 	m_adj(m0, hlen + firstlen - ip_len);
1019 	m0->m_pkthdr.len = hlen + firstlen;
1020 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1021 	ip->ip_off = htons(ip_off | IP_MF);
1022 	ip->ip_sum = 0;
1023 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1024 		ip->ip_sum = in_cksum(m0, hlen);
1025 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1026 	}
1027 
1028 done:
1029 	*m_frag = m0;
1030 	return error;
1031 }
1032 
1033 void
in_delayed_cksum(struct mbuf * m)1034 in_delayed_cksum(struct mbuf *m)
1035 {
1036 	struct ip *ip;
1037 	struct udphdr *uh;
1038 	uint16_t cklen, csum, offset;
1039 
1040 	ip = mtod(m, struct ip *);
1041 	offset = ip->ip_hl << 2 ;
1042 
1043 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1044 		/* if udp header is not in the first mbuf copy udplen */
1045 		if (offset + sizeof(struct udphdr) > m->m_len) {
1046 			m_copydata(m, offset + offsetof(struct udphdr,
1047 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1048 			cklen = ntohs(cklen);
1049 		} else {
1050 			uh = (struct udphdr *)mtodo(m, offset);
1051 			cklen = ntohs(uh->uh_ulen);
1052 		}
1053 		csum = in_cksum_skip(m, cklen + offset, offset);
1054 		if (csum == 0)
1055 			csum = 0xffff;
1056 	} else {
1057 		cklen = ntohs(ip->ip_len);
1058 		csum = in_cksum_skip(m, cklen, offset);
1059 	}
1060 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1061 
1062 	if (offset + sizeof(csum) > m->m_len)
1063 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1064 	else
1065 		*(u_short *)mtodo(m, offset) = csum;
1066 }
1067 
1068 /*
1069  * IP socket option processing.
1070  */
1071 int
ip_ctloutput(struct socket * so,struct sockopt * sopt)1072 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1073 {
1074 	struct inpcb *inp = sotoinpcb(so);
1075 	int	error, optval;
1076 #ifdef	RSS
1077 	uint32_t rss_bucket;
1078 	int retval;
1079 #endif
1080 
1081 	error = optval = 0;
1082 	if (sopt->sopt_level != IPPROTO_IP) {
1083 		error = EINVAL;
1084 
1085 		if (sopt->sopt_level == SOL_SOCKET &&
1086 		    sopt->sopt_dir == SOPT_SET) {
1087 			switch (sopt->sopt_name) {
1088 			case SO_REUSEADDR:
1089 				INP_WLOCK(inp);
1090 				if ((so->so_options & SO_REUSEADDR) != 0)
1091 					inp->inp_flags2 |= INP_REUSEADDR;
1092 				else
1093 					inp->inp_flags2 &= ~INP_REUSEADDR;
1094 				INP_WUNLOCK(inp);
1095 				error = 0;
1096 				break;
1097 			case SO_REUSEPORT:
1098 				INP_WLOCK(inp);
1099 				if ((so->so_options & SO_REUSEPORT) != 0)
1100 					inp->inp_flags2 |= INP_REUSEPORT;
1101 				else
1102 					inp->inp_flags2 &= ~INP_REUSEPORT;
1103 				INP_WUNLOCK(inp);
1104 				error = 0;
1105 				break;
1106 			case SO_REUSEPORT_LB:
1107 				INP_WLOCK(inp);
1108 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1109 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1110 				else
1111 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1112 				INP_WUNLOCK(inp);
1113 				error = 0;
1114 				break;
1115 			case SO_SETFIB:
1116 				INP_WLOCK(inp);
1117 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1118 				INP_WUNLOCK(inp);
1119 				error = 0;
1120 				break;
1121 			case SO_MAX_PACING_RATE:
1122 #ifdef RATELIMIT
1123 				INP_WLOCK(inp);
1124 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1125 				INP_WUNLOCK(inp);
1126 				error = 0;
1127 #else
1128 				error = EOPNOTSUPP;
1129 #endif
1130 				break;
1131 			default:
1132 				break;
1133 			}
1134 		}
1135 		return (error);
1136 	}
1137 
1138 	switch (sopt->sopt_dir) {
1139 	case SOPT_SET:
1140 		switch (sopt->sopt_name) {
1141 		case IP_OPTIONS:
1142 #ifdef notyet
1143 		case IP_RETOPTS:
1144 #endif
1145 		{
1146 			struct mbuf *m;
1147 			if (sopt->sopt_valsize > MLEN) {
1148 				error = EMSGSIZE;
1149 				break;
1150 			}
1151 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1152 			if (m == NULL) {
1153 				error = ENOBUFS;
1154 				break;
1155 			}
1156 			m->m_len = sopt->sopt_valsize;
1157 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1158 					    m->m_len);
1159 			if (error) {
1160 				m_free(m);
1161 				break;
1162 			}
1163 			INP_WLOCK(inp);
1164 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1165 			INP_WUNLOCK(inp);
1166 			return (error);
1167 		}
1168 
1169 		case IP_BINDANY:
1170 			if (sopt->sopt_td != NULL) {
1171 				error = priv_check(sopt->sopt_td,
1172 				    PRIV_NETINET_BINDANY);
1173 				if (error)
1174 					break;
1175 			}
1176 			/* FALLTHROUGH */
1177 		case IP_BINDMULTI:
1178 #ifdef	RSS
1179 		case IP_RSS_LISTEN_BUCKET:
1180 #endif
1181 		case IP_TOS:
1182 		case IP_TTL:
1183 		case IP_MINTTL:
1184 		case IP_RECVOPTS:
1185 		case IP_RECVRETOPTS:
1186 		case IP_ORIGDSTADDR:
1187 		case IP_RECVDSTADDR:
1188 		case IP_RECVTTL:
1189 		case IP_RECVIF:
1190 		case IP_ONESBCAST:
1191 		case IP_DONTFRAG:
1192 		case IP_RECVTOS:
1193 		case IP_RECVFLOWID:
1194 #ifdef	RSS
1195 		case IP_RECVRSSBUCKETID:
1196 #endif
1197 		case IP_VLAN_PCP:
1198 			error = sooptcopyin(sopt, &optval, sizeof optval,
1199 					    sizeof optval);
1200 			if (error)
1201 				break;
1202 
1203 			switch (sopt->sopt_name) {
1204 			case IP_TOS:
1205 				inp->inp_ip_tos = optval;
1206 				break;
1207 
1208 			case IP_TTL:
1209 				inp->inp_ip_ttl = optval;
1210 				break;
1211 
1212 			case IP_MINTTL:
1213 				if (optval >= 0 && optval <= MAXTTL)
1214 					inp->inp_ip_minttl = optval;
1215 				else
1216 					error = EINVAL;
1217 				break;
1218 
1219 #define	OPTSET(bit) do {						\
1220 	INP_WLOCK(inp);							\
1221 	if (optval)							\
1222 		inp->inp_flags |= bit;					\
1223 	else								\
1224 		inp->inp_flags &= ~bit;					\
1225 	INP_WUNLOCK(inp);						\
1226 } while (0)
1227 
1228 #define	OPTSET2(bit, val) do {						\
1229 	INP_WLOCK(inp);							\
1230 	if (val)							\
1231 		inp->inp_flags2 |= bit;					\
1232 	else								\
1233 		inp->inp_flags2 &= ~bit;				\
1234 	INP_WUNLOCK(inp);						\
1235 } while (0)
1236 
1237 			case IP_RECVOPTS:
1238 				OPTSET(INP_RECVOPTS);
1239 				break;
1240 
1241 			case IP_RECVRETOPTS:
1242 				OPTSET(INP_RECVRETOPTS);
1243 				break;
1244 
1245 			case IP_RECVDSTADDR:
1246 				OPTSET(INP_RECVDSTADDR);
1247 				break;
1248 
1249 			case IP_ORIGDSTADDR:
1250 				OPTSET2(INP_ORIGDSTADDR, optval);
1251 				break;
1252 
1253 			case IP_RECVTTL:
1254 				OPTSET(INP_RECVTTL);
1255 				break;
1256 
1257 			case IP_RECVIF:
1258 				OPTSET(INP_RECVIF);
1259 				break;
1260 
1261 			case IP_ONESBCAST:
1262 				OPTSET(INP_ONESBCAST);
1263 				break;
1264 			case IP_DONTFRAG:
1265 				OPTSET(INP_DONTFRAG);
1266 				break;
1267 			case IP_BINDANY:
1268 				OPTSET(INP_BINDANY);
1269 				break;
1270 			case IP_RECVTOS:
1271 				OPTSET(INP_RECVTOS);
1272 				break;
1273 			case IP_BINDMULTI:
1274 				OPTSET2(INP_BINDMULTI, optval);
1275 				break;
1276 			case IP_RECVFLOWID:
1277 				OPTSET2(INP_RECVFLOWID, optval);
1278 				break;
1279 #ifdef	RSS
1280 			case IP_RSS_LISTEN_BUCKET:
1281 				if ((optval >= 0) &&
1282 				    (optval < rss_getnumbuckets())) {
1283 					inp->inp_rss_listen_bucket = optval;
1284 					OPTSET2(INP_RSS_BUCKET_SET, 1);
1285 				} else {
1286 					error = EINVAL;
1287 				}
1288 				break;
1289 			case IP_RECVRSSBUCKETID:
1290 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1291 				break;
1292 #endif
1293 			case IP_VLAN_PCP:
1294 				if ((optval >= -1) && (optval <=
1295 				    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1296 					if (optval == -1) {
1297 						INP_WLOCK(inp);
1298 						inp->inp_flags2 &=
1299 						    ~(INP_2PCP_SET |
1300 						      INP_2PCP_MASK);
1301 						INP_WUNLOCK(inp);
1302 					} else {
1303 						INP_WLOCK(inp);
1304 						inp->inp_flags2 |=
1305 						    INP_2PCP_SET;
1306 						inp->inp_flags2 &=
1307 						    ~INP_2PCP_MASK;
1308 						inp->inp_flags2 |=
1309 						    optval << INP_2PCP_SHIFT;
1310 						INP_WUNLOCK(inp);
1311 					}
1312 				} else
1313 					error = EINVAL;
1314 				break;
1315 			}
1316 			break;
1317 #undef OPTSET
1318 #undef OPTSET2
1319 
1320 		/*
1321 		 * Multicast socket options are processed by the in_mcast
1322 		 * module.
1323 		 */
1324 		case IP_MULTICAST_IF:
1325 		case IP_MULTICAST_VIF:
1326 		case IP_MULTICAST_TTL:
1327 		case IP_MULTICAST_LOOP:
1328 		case IP_ADD_MEMBERSHIP:
1329 		case IP_DROP_MEMBERSHIP:
1330 		case IP_ADD_SOURCE_MEMBERSHIP:
1331 		case IP_DROP_SOURCE_MEMBERSHIP:
1332 		case IP_BLOCK_SOURCE:
1333 		case IP_UNBLOCK_SOURCE:
1334 		case IP_MSFILTER:
1335 		case MCAST_JOIN_GROUP:
1336 		case MCAST_LEAVE_GROUP:
1337 		case MCAST_JOIN_SOURCE_GROUP:
1338 		case MCAST_LEAVE_SOURCE_GROUP:
1339 		case MCAST_BLOCK_SOURCE:
1340 		case MCAST_UNBLOCK_SOURCE:
1341 			error = inp_setmoptions(inp, sopt);
1342 			break;
1343 
1344 		case IP_PORTRANGE:
1345 			error = sooptcopyin(sopt, &optval, sizeof optval,
1346 					    sizeof optval);
1347 			if (error)
1348 				break;
1349 
1350 			INP_WLOCK(inp);
1351 			switch (optval) {
1352 			case IP_PORTRANGE_DEFAULT:
1353 				inp->inp_flags &= ~(INP_LOWPORT);
1354 				inp->inp_flags &= ~(INP_HIGHPORT);
1355 				break;
1356 
1357 			case IP_PORTRANGE_HIGH:
1358 				inp->inp_flags &= ~(INP_LOWPORT);
1359 				inp->inp_flags |= INP_HIGHPORT;
1360 				break;
1361 
1362 			case IP_PORTRANGE_LOW:
1363 				inp->inp_flags &= ~(INP_HIGHPORT);
1364 				inp->inp_flags |= INP_LOWPORT;
1365 				break;
1366 
1367 			default:
1368 				error = EINVAL;
1369 				break;
1370 			}
1371 			INP_WUNLOCK(inp);
1372 			break;
1373 
1374 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1375 		case IP_IPSEC_POLICY:
1376 			if (IPSEC_ENABLED(ipv4)) {
1377 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1378 				break;
1379 			}
1380 			/* FALLTHROUGH */
1381 #endif /* IPSEC */
1382 
1383 		default:
1384 			error = ENOPROTOOPT;
1385 			break;
1386 		}
1387 		break;
1388 
1389 	case SOPT_GET:
1390 		switch (sopt->sopt_name) {
1391 		case IP_OPTIONS:
1392 		case IP_RETOPTS:
1393 			INP_RLOCK(inp);
1394 			if (inp->inp_options) {
1395 				struct mbuf *options;
1396 
1397 				options = m_copym(inp->inp_options, 0,
1398 				    M_COPYALL, M_NOWAIT);
1399 				INP_RUNLOCK(inp);
1400 				if (options != NULL) {
1401 					error = sooptcopyout(sopt,
1402 							     mtod(options, char *),
1403 							     options->m_len);
1404 					m_freem(options);
1405 				} else
1406 					error = ENOMEM;
1407 			} else {
1408 				INP_RUNLOCK(inp);
1409 				sopt->sopt_valsize = 0;
1410 			}
1411 			break;
1412 
1413 		case IP_TOS:
1414 		case IP_TTL:
1415 		case IP_MINTTL:
1416 		case IP_RECVOPTS:
1417 		case IP_RECVRETOPTS:
1418 		case IP_ORIGDSTADDR:
1419 		case IP_RECVDSTADDR:
1420 		case IP_RECVTTL:
1421 		case IP_RECVIF:
1422 		case IP_PORTRANGE:
1423 		case IP_ONESBCAST:
1424 		case IP_DONTFRAG:
1425 		case IP_BINDANY:
1426 		case IP_RECVTOS:
1427 		case IP_BINDMULTI:
1428 		case IP_FLOWID:
1429 		case IP_FLOWTYPE:
1430 		case IP_RECVFLOWID:
1431 #ifdef	RSS
1432 		case IP_RSSBUCKETID:
1433 		case IP_RECVRSSBUCKETID:
1434 #endif
1435 		case IP_VLAN_PCP:
1436 			switch (sopt->sopt_name) {
1437 			case IP_TOS:
1438 				optval = inp->inp_ip_tos;
1439 				break;
1440 
1441 			case IP_TTL:
1442 				optval = inp->inp_ip_ttl;
1443 				break;
1444 
1445 			case IP_MINTTL:
1446 				optval = inp->inp_ip_minttl;
1447 				break;
1448 
1449 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1450 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1451 
1452 			case IP_RECVOPTS:
1453 				optval = OPTBIT(INP_RECVOPTS);
1454 				break;
1455 
1456 			case IP_RECVRETOPTS:
1457 				optval = OPTBIT(INP_RECVRETOPTS);
1458 				break;
1459 
1460 			case IP_RECVDSTADDR:
1461 				optval = OPTBIT(INP_RECVDSTADDR);
1462 				break;
1463 
1464 			case IP_ORIGDSTADDR:
1465 				optval = OPTBIT2(INP_ORIGDSTADDR);
1466 				break;
1467 
1468 			case IP_RECVTTL:
1469 				optval = OPTBIT(INP_RECVTTL);
1470 				break;
1471 
1472 			case IP_RECVIF:
1473 				optval = OPTBIT(INP_RECVIF);
1474 				break;
1475 
1476 			case IP_PORTRANGE:
1477 				if (inp->inp_flags & INP_HIGHPORT)
1478 					optval = IP_PORTRANGE_HIGH;
1479 				else if (inp->inp_flags & INP_LOWPORT)
1480 					optval = IP_PORTRANGE_LOW;
1481 				else
1482 					optval = 0;
1483 				break;
1484 
1485 			case IP_ONESBCAST:
1486 				optval = OPTBIT(INP_ONESBCAST);
1487 				break;
1488 			case IP_DONTFRAG:
1489 				optval = OPTBIT(INP_DONTFRAG);
1490 				break;
1491 			case IP_BINDANY:
1492 				optval = OPTBIT(INP_BINDANY);
1493 				break;
1494 			case IP_RECVTOS:
1495 				optval = OPTBIT(INP_RECVTOS);
1496 				break;
1497 			case IP_FLOWID:
1498 				optval = inp->inp_flowid;
1499 				break;
1500 			case IP_FLOWTYPE:
1501 				optval = inp->inp_flowtype;
1502 				break;
1503 			case IP_RECVFLOWID:
1504 				optval = OPTBIT2(INP_RECVFLOWID);
1505 				break;
1506 #ifdef	RSS
1507 			case IP_RSSBUCKETID:
1508 				retval = rss_hash2bucket(inp->inp_flowid,
1509 				    inp->inp_flowtype,
1510 				    &rss_bucket);
1511 				if (retval == 0)
1512 					optval = rss_bucket;
1513 				else
1514 					error = EINVAL;
1515 				break;
1516 			case IP_RECVRSSBUCKETID:
1517 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1518 				break;
1519 #endif
1520 			case IP_BINDMULTI:
1521 				optval = OPTBIT2(INP_BINDMULTI);
1522 				break;
1523 			case IP_VLAN_PCP:
1524 				if (OPTBIT2(INP_2PCP_SET)) {
1525 					optval = (inp->inp_flags2 &
1526 					    INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1527 				} else {
1528 					optval = -1;
1529 				}
1530 				break;
1531 			}
1532 			error = sooptcopyout(sopt, &optval, sizeof optval);
1533 			break;
1534 
1535 		/*
1536 		 * Multicast socket options are processed by the in_mcast
1537 		 * module.
1538 		 */
1539 		case IP_MULTICAST_IF:
1540 		case IP_MULTICAST_VIF:
1541 		case IP_MULTICAST_TTL:
1542 		case IP_MULTICAST_LOOP:
1543 		case IP_MSFILTER:
1544 			error = inp_getmoptions(inp, sopt);
1545 			break;
1546 
1547 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1548 		case IP_IPSEC_POLICY:
1549 			if (IPSEC_ENABLED(ipv4)) {
1550 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1551 				break;
1552 			}
1553 			/* FALLTHROUGH */
1554 #endif /* IPSEC */
1555 
1556 		default:
1557 			error = ENOPROTOOPT;
1558 			break;
1559 		}
1560 		break;
1561 	}
1562 	return (error);
1563 }
1564 
1565 /*
1566  * Routine called from ip_output() to loop back a copy of an IP multicast
1567  * packet to the input queue of a specified interface.  Note that this
1568  * calls the output routine of the loopback "driver", but with an interface
1569  * pointer that might NOT be a loopback interface -- evil, but easier than
1570  * replicating that code here.
1571  */
1572 static void
ip_mloopback(struct ifnet * ifp,const struct mbuf * m,int hlen)1573 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1574 {
1575 	struct ip *ip;
1576 	struct mbuf *copym;
1577 
1578 	/*
1579 	 * Make a deep copy of the packet because we're going to
1580 	 * modify the pack in order to generate checksums.
1581 	 */
1582 	copym = m_dup(m, M_NOWAIT);
1583 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1584 		copym = m_pullup(copym, hlen);
1585 	if (copym != NULL) {
1586 		/* If needed, compute the checksum and mark it as valid. */
1587 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1588 			in_delayed_cksum(copym);
1589 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1590 			copym->m_pkthdr.csum_flags |=
1591 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1592 			copym->m_pkthdr.csum_data = 0xffff;
1593 		}
1594 		/*
1595 		 * We don't bother to fragment if the IP length is greater
1596 		 * than the interface's MTU.  Can this possibly matter?
1597 		 */
1598 		ip = mtod(copym, struct ip *);
1599 		ip->ip_sum = 0;
1600 		ip->ip_sum = in_cksum(copym, hlen);
1601 		if_simloop(ifp, copym, AF_INET, 0);
1602 	}
1603 }
1604