xref: /freebsd/sys/netinet/ip_output.c (revision 394605c057ade77775e1f9975e278dbb7693f44f)
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 
32 #include <sys/cdefs.h>
33 #include "opt_inet.h"
34 #include "opt_ipsec.h"
35 #include "opt_kern_tls.h"
36 #include "opt_mbuf_stress_test.h"
37 #include "opt_ratelimit.h"
38 #include "opt_route.h"
39 #include "opt_rss.h"
40 #include "opt_sctp.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/ktls.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mbuf.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/protosw.h>
52 #include <sys/sdt.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sysctl.h>
56 #include <sys/ucred.h>
57 
58 #include <net/if.h>
59 #include <net/if_var.h>
60 #include <net/if_private.h>
61 #include <net/if_vlan_var.h>
62 #include <net/if_llatbl.h>
63 #include <net/ethernet.h>
64 #include <net/netisr.h>
65 #include <net/pfil.h>
66 #include <net/route.h>
67 #include <net/route/nhop.h>
68 #include <net/rss_config.h>
69 #include <net/vnet.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_fib.h>
73 #include <netinet/in_kdtrace.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/ip.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/in_rss.h>
79 #include <netinet/in_var.h>
80 #include <netinet/ip_var.h>
81 #include <netinet/ip_options.h>
82 
83 #include <netinet/udp.h>
84 #include <netinet/udp_var.h>
85 
86 #if defined(SCTP) || defined(SCTP_SUPPORT)
87 #include <netinet/sctp.h>
88 #include <netinet/sctp_crc32.h>
89 #endif
90 
91 #include <netipsec/ipsec_support.h>
92 
93 #include <machine/in_cksum.h>
94 
95 #include <security/mac/mac_framework.h>
96 
97 #ifdef MBUF_STRESS_TEST
98 static int mbuf_frag_size = 0;
99 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
100 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
101 #endif
102 
103 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
104 
105 extern int in_mcast_loop;
106 
107 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)108 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
109     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
110 {
111 	struct m_tag *fwd_tag = NULL;
112 	struct mbuf *m;
113 	struct in_addr odst;
114 	struct ip *ip;
115 	int ret;
116 
117 	m = *mp;
118 	ip = mtod(m, struct ip *);
119 
120 	/* Run through list of hooks for output packets. */
121 	odst.s_addr = ip->ip_dst.s_addr;
122 	if (flags & IP_FORWARDING)
123 		ret = pfil_mbuf_fwd(V_inet_pfil_head, mp, ifp, inp);
124 	else
125 		ret = pfil_mbuf_out(V_inet_pfil_head, mp, ifp, inp);
126 
127 	switch (ret) {
128 	case PFIL_DROPPED:
129 		*error = EACCES;
130 		/* FALLTHROUGH */
131 	case PFIL_CONSUMED:
132 		return 1; /* Finished */
133 	case PFIL_PASS:
134 		*error = 0;
135 	}
136 	m = *mp;
137 	ip = mtod(m, struct ip *);
138 
139 	/* See if destination IP address was changed by packet filter. */
140 	if (odst.s_addr != ip->ip_dst.s_addr) {
141 		m->m_flags |= M_SKIP_FIREWALL;
142 		/* If destination is now ourself drop to ip_input(). */
143 		if (in_localip(ip->ip_dst)) {
144 			m->m_flags |= M_FASTFWD_OURS;
145 			if (m->m_pkthdr.rcvif == NULL)
146 				m->m_pkthdr.rcvif = V_loif;
147 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
148 				m->m_pkthdr.csum_flags |=
149 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
150 				m->m_pkthdr.csum_data = 0xffff;
151 			}
152 			m->m_pkthdr.csum_flags |=
153 				CSUM_IP_CHECKED | CSUM_IP_VALID;
154 #if defined(SCTP) || defined(SCTP_SUPPORT)
155 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
156 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
157 #endif
158 			*error = netisr_queue(NETISR_IP, m);
159 			return 1; /* Finished */
160 		}
161 
162 		bzero(dst, sizeof(*dst));
163 		dst->sin_family = AF_INET;
164 		dst->sin_len = sizeof(*dst);
165 		dst->sin_addr = ip->ip_dst;
166 
167 		return -1; /* Reloop */
168 	}
169 	/* See if fib was changed by packet filter. */
170 	if ((*fibnum) != M_GETFIB(m)) {
171 		m->m_flags |= M_SKIP_FIREWALL;
172 		*fibnum = M_GETFIB(m);
173 		return -1; /* Reloop for FIB change */
174 	}
175 
176 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
177 	if (m->m_flags & M_FASTFWD_OURS) {
178 		if (m->m_pkthdr.rcvif == NULL)
179 			m->m_pkthdr.rcvif = V_loif;
180 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
181 			m->m_pkthdr.csum_flags |=
182 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
183 			m->m_pkthdr.csum_data = 0xffff;
184 		}
185 #if defined(SCTP) || defined(SCTP_SUPPORT)
186 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
187 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
188 #endif
189 		m->m_pkthdr.csum_flags |=
190 			CSUM_IP_CHECKED | CSUM_IP_VALID;
191 
192 		*error = netisr_queue(NETISR_IP, m);
193 		return 1; /* Finished */
194 	}
195 	/* Or forward to some other address? */
196 	if ((m->m_flags & M_IP_NEXTHOP) &&
197 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
198 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
199 		m->m_flags |= M_SKIP_FIREWALL;
200 		m->m_flags &= ~M_IP_NEXTHOP;
201 		m_tag_delete(m, fwd_tag);
202 
203 		return -1; /* Reloop for CHANGE of dst */
204 	}
205 
206 	return 0;
207 }
208 
209 static int
ip_output_send(struct inpcb * inp,struct ifnet * ifp,struct mbuf * m,const struct sockaddr * gw,struct route * ro,bool stamp_tag)210 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
211     const struct sockaddr *gw, struct route *ro, bool stamp_tag)
212 {
213 #ifdef KERN_TLS
214 	struct ktls_session *tls = NULL;
215 #endif
216 	struct m_snd_tag *mst;
217 	int error;
218 
219 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
220 	mst = NULL;
221 
222 #ifdef KERN_TLS
223 	/*
224 	 * If this is an unencrypted TLS record, save a reference to
225 	 * the record.  This local reference is used to call
226 	 * ktls_output_eagain after the mbuf has been freed (thus
227 	 * dropping the mbuf's reference) in if_output.
228 	 */
229 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
230 		tls = ktls_hold(m->m_next->m_epg_tls);
231 		mst = tls->snd_tag;
232 
233 		/*
234 		 * If a TLS session doesn't have a valid tag, it must
235 		 * have had an earlier ifp mismatch, so drop this
236 		 * packet.
237 		 */
238 		if (mst == NULL) {
239 			m_freem(m);
240 			error = EAGAIN;
241 			goto done;
242 		}
243 		/*
244 		 * Always stamp tags that include NIC ktls.
245 		 */
246 		stamp_tag = true;
247 	}
248 #endif
249 #ifdef RATELIMIT
250 	if (inp != NULL && mst == NULL) {
251 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
252 		    (inp->inp_snd_tag != NULL &&
253 		    inp->inp_snd_tag->ifp != ifp))
254 			in_pcboutput_txrtlmt(inp, ifp, m);
255 
256 		if (inp->inp_snd_tag != NULL)
257 			mst = inp->inp_snd_tag;
258 	}
259 #endif
260 	if (stamp_tag && mst != NULL) {
261 		KASSERT(m->m_pkthdr.rcvif == NULL,
262 		    ("trying to add a send tag to a forwarded packet"));
263 		if (mst->ifp != ifp) {
264 			m_freem(m);
265 			error = EAGAIN;
266 			goto done;
267 		}
268 
269 		/* stamp send tag on mbuf */
270 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
271 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
272 	}
273 
274 	error = (*ifp->if_output)(ifp, m, gw, ro);
275 
276 done:
277 	/* Check for route change invalidating send tags. */
278 #ifdef KERN_TLS
279 	if (tls != NULL) {
280 		if (error == EAGAIN)
281 			error = ktls_output_eagain(inp, tls);
282 		ktls_free(tls);
283 	}
284 #endif
285 #ifdef RATELIMIT
286 	if (error == EAGAIN)
287 		in_pcboutput_eagain(inp);
288 #endif
289 	return (error);
290 }
291 
292 /* rte<>ro_flags translation */
293 static inline void
rt_update_ro_flags(struct route * ro,const struct nhop_object * nh)294 rt_update_ro_flags(struct route *ro, const struct nhop_object *nh)
295 {
296 	int nh_flags = nh->nh_flags;
297 
298 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
299 
300 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
301 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
302 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
303 }
304 
305 /*
306  * IP output.  The packet in mbuf chain m contains a skeletal IP
307  * header (with len, off, ttl, proto, tos, src, dst).
308  * The mbuf chain containing the packet will be freed.
309  * The mbuf opt, if present, will not be freed.
310  * If route ro is present and has ro_rt initialized, route lookup would be
311  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
312  * then result of route lookup is stored in ro->ro_rt.
313  *
314  * In the IP forwarding case, the packet will arrive with options already
315  * inserted, so must have a NULL opt pointer.
316  */
317 int
ip_output(struct mbuf * m,struct mbuf * opt,struct route * ro,int flags,struct ip_moptions * imo,struct inpcb * inp)318 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
319     struct ip_moptions *imo, struct inpcb *inp)
320 {
321 	struct ip *ip;
322 	struct ifnet *ifp = NULL;	/* keep compiler happy */
323 	struct mbuf *m0;
324 	int hlen = sizeof (struct ip);
325 	int mtu = 0;
326 	int error = 0;
327 	int vlan_pcp = -1;
328 	struct sockaddr_in *dst;
329 	const struct sockaddr *gw;
330 	struct in_ifaddr *ia = NULL;
331 	struct in_addr src;
332 	bool isbroadcast;
333 	uint16_t ip_len, ip_off;
334 	struct route iproute;
335 	uint32_t fibnum;
336 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
337 	int no_route_but_check_spd = 0;
338 #endif
339 
340 	M_ASSERTPKTHDR(m);
341 	NET_EPOCH_ASSERT();
342 
343 	if (inp != NULL) {
344 		INP_LOCK_ASSERT(inp);
345 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
346 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
347 			m->m_pkthdr.flowid = inp->inp_flowid;
348 			M_HASHTYPE_SET(m, inp->inp_flowtype);
349 		}
350 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
351 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
352 			    INP_2PCP_SHIFT;
353 #ifdef NUMA
354 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
355 #endif
356 	}
357 
358 	if (opt) {
359 		int len = 0;
360 		m = ip_insertoptions(m, opt, &len);
361 		if (len != 0)
362 			hlen = len; /* ip->ip_hl is updated above */
363 	}
364 	ip = mtod(m, struct ip *);
365 	ip_len = ntohs(ip->ip_len);
366 	ip_off = ntohs(ip->ip_off);
367 
368 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
369 		ip->ip_v = IPVERSION;
370 		ip->ip_hl = hlen >> 2;
371 		ip_fillid(ip, V_ip_random_id);
372 	} else {
373 		/* Header already set, fetch hlen from there */
374 		hlen = ip->ip_hl << 2;
375 	}
376 	if ((flags & IP_FORWARDING) == 0)
377 		IPSTAT_INC(ips_localout);
378 
379 	/*
380 	 * dst/gw handling:
381 	 *
382 	 * gw is readonly but can point either to dst OR rt_gateway,
383 	 * therefore we need restore gw if we're redoing lookup.
384 	 */
385 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
386 	if (ro == NULL) {
387 		ro = &iproute;
388 		bzero(ro, sizeof (*ro));
389 	}
390 	dst = (struct sockaddr_in *)&ro->ro_dst;
391 	if (ro->ro_nh == NULL) {
392 		dst->sin_family = AF_INET;
393 		dst->sin_len = sizeof(*dst);
394 		dst->sin_addr = ip->ip_dst;
395 	}
396 	gw = (const struct sockaddr *)dst;
397 again:
398 	/*
399 	 * Validate route against routing table additions;
400 	 * a better/more specific route might have been added.
401 	 */
402 	if (inp != NULL && ro->ro_nh != NULL)
403 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
404 	/*
405 	 * If there is a cached route,
406 	 * check that it is to the same destination
407 	 * and is still up.  If not, free it and try again.
408 	 * The address family should also be checked in case of sharing the
409 	 * cache with IPv6.
410 	 * Also check whether routing cache needs invalidation.
411 	 */
412 	if (ro->ro_nh != NULL &&
413 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
414 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
415 		RO_INVALIDATE_CACHE(ro);
416 	ia = NULL;
417 	/*
418 	 * If routing to interface only, short circuit routing lookup.
419 	 * The use of an all-ones broadcast address implies this; an
420 	 * interface is specified by the broadcast address of an interface,
421 	 * or the destination address of a ptp interface.
422 	 */
423 	if (flags & IP_SENDONES) {
424 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
425 						      M_GETFIB(m)))) == NULL &&
426 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
427 						    M_GETFIB(m)))) == NULL) {
428 			IPSTAT_INC(ips_noroute);
429 			error = ENETUNREACH;
430 			goto bad;
431 		}
432 		ip->ip_dst.s_addr = INADDR_BROADCAST;
433 		dst->sin_addr = ip->ip_dst;
434 		ifp = ia->ia_ifp;
435 		mtu = ifp->if_mtu;
436 		ip->ip_ttl = 1;
437 		isbroadcast = true;
438 		src = IA_SIN(ia)->sin_addr;
439 	} else if (flags & IP_ROUTETOIF) {
440 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
441 						    M_GETFIB(m)))) == NULL &&
442 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
443 						M_GETFIB(m)))) == NULL) {
444 			IPSTAT_INC(ips_noroute);
445 			error = ENETUNREACH;
446 			goto bad;
447 		}
448 		ifp = ia->ia_ifp;
449 		mtu = ifp->if_mtu;
450 		ip->ip_ttl = 1;
451 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
452 		    (in_broadcast(ip->ip_dst) ||
453 		    in_ifaddr_broadcast(dst->sin_addr, ia)) : 0;
454 		src = IA_SIN(ia)->sin_addr;
455 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
456 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
457 		/*
458 		 * Bypass the normal routing lookup for multicast
459 		 * packets if the interface is specified.
460 		 */
461 		ifp = imo->imo_multicast_ifp;
462 		mtu = ifp->if_mtu;
463 		IFP_TO_IA(ifp, ia);
464 		isbroadcast = false;
465 		/* Interface may have no addresses. */
466 		if (ia != NULL)
467 			src = IA_SIN(ia)->sin_addr;
468 		else
469 			src.s_addr = INADDR_ANY;
470 	} else if (ro != &iproute) {
471 		if (ro->ro_nh == NULL) {
472 			/*
473 			 * We want to do any cloning requested by the link
474 			 * layer, as this is probably required in all cases
475 			 * for correct operation (as it is for ARP).
476 			 */
477 			uint32_t flowid;
478 			flowid = m->m_pkthdr.flowid;
479 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
480 			    NHR_REF, flowid);
481 
482 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
483 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
484 				/*
485 				 * There is no route for this packet, but it is
486 				 * possible that a matching SPD entry exists.
487 				 */
488 				no_route_but_check_spd = 1;
489 				goto sendit;
490 #endif
491 				IPSTAT_INC(ips_noroute);
492 				error = EHOSTUNREACH;
493 				goto bad;
494 			}
495 		}
496 		struct nhop_object *nh = ro->ro_nh;
497 
498 		ia = ifatoia(nh->nh_ifa);
499 		ifp = nh->nh_ifp;
500 		counter_u64_add(nh->nh_pksent, 1);
501 		rt_update_ro_flags(ro, nh);
502 		if (nh->nh_flags & NHF_GATEWAY)
503 			gw = &nh->gw_sa;
504 		if (nh->nh_flags & NHF_HOST)
505 			isbroadcast = (nh->nh_flags & NHF_BROADCAST);
506 		else if ((ifp->if_flags & IFF_BROADCAST) &&
507 		    (gw->sa_family == AF_INET))
508 			isbroadcast = in_broadcast(ip->ip_dst) ||
509 			    in_ifaddr_broadcast(
510 			    ((const struct sockaddr_in *)gw)->sin_addr, ia);
511 		else
512 			isbroadcast = false;
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), dst->sin_addr, 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_broadcast(ip->ip_dst) || in_ifaddr_broadcast(
545 		    ((const struct sockaddr_in *)gw)->sin_addr, ia)));
546 	}
547 
548 	/* Catch a possible divide by zero later. */
549 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
550 	    __func__, mtu, ro,
551 	    (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
552 
553 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
554 		m->m_flags |= M_MCAST;
555 		/*
556 		 * IP destination address is multicast.  Make sure "gw"
557 		 * still points to the address in "ro".  (It may have been
558 		 * changed to point to a gateway address, above.)
559 		 */
560 		gw = (const struct sockaddr *)dst;
561 		/*
562 		 * See if the caller provided any multicast options
563 		 */
564 		if (imo != NULL) {
565 			ip->ip_ttl = imo->imo_multicast_ttl;
566 			if (imo->imo_multicast_vif != -1)
567 				ip->ip_src.s_addr =
568 				    ip_mcast_src ?
569 				    ip_mcast_src(imo->imo_multicast_vif) :
570 				    INADDR_ANY;
571 		} else
572 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
573 		/*
574 		 * Confirm that the outgoing interface supports multicast.
575 		 */
576 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
577 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
578 				IPSTAT_INC(ips_noroute);
579 				error = ENETUNREACH;
580 				goto bad;
581 			}
582 		}
583 		/*
584 		 * If source address not specified yet, use address
585 		 * of outgoing interface.
586 		 */
587 		if (ip->ip_src.s_addr == INADDR_ANY)
588 			ip->ip_src = src;
589 
590 		if ((imo == NULL && in_mcast_loop) ||
591 		    (imo && imo->imo_multicast_loop)) {
592 			/*
593 			 * Loop back multicast datagram if not expressly
594 			 * forbidden to do so, even if we are not a member
595 			 * of the group; ip_input() will filter it later,
596 			 * thus deferring a hash lookup and mutex acquisition
597 			 * at the expense of a cheap copy using m_copym().
598 			 */
599 			ip_mloopback(ifp, m, hlen);
600 		} else {
601 			/*
602 			 * If we are acting as a multicast router, perform
603 			 * multicast forwarding as if the packet had just
604 			 * arrived on the interface to which we are about
605 			 * to send.  The multicast forwarding function
606 			 * recursively calls this function, using the
607 			 * IP_FORWARDING flag to prevent infinite recursion.
608 			 *
609 			 * Multicasts that are looped back by ip_mloopback(),
610 			 * above, will be forwarded by the ip_input() routine,
611 			 * if necessary.
612 			 */
613 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
614 				/*
615 				 * If rsvp daemon is not running, do not
616 				 * set ip_moptions. This ensures that the packet
617 				 * is multicast and not just sent down one link
618 				 * as prescribed by rsvpd.
619 				 */
620 				if (!V_rsvp_on)
621 					imo = NULL;
622 				if (ip_mforward &&
623 				    ip_mforward(ip, ifp, m, imo) != 0) {
624 					m_freem(m);
625 					goto done;
626 				}
627 			}
628 		}
629 
630 		/*
631 		 * Multicasts with a time-to-live of zero may be looped-
632 		 * back, above, but must not be transmitted on a network.
633 		 * Also, multicasts addressed to the loopback interface
634 		 * are not sent -- the above call to ip_mloopback() will
635 		 * loop back a copy. ip_input() will drop the copy if
636 		 * this host does not belong to the destination group on
637 		 * the loopback interface.
638 		 */
639 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
640 			m_freem(m);
641 			goto done;
642 		}
643 
644 		goto sendit;
645 	}
646 
647 	/*
648 	 * If the source address is not specified yet, use the address
649 	 * of the outoing interface.
650 	 */
651 	if (ip->ip_src.s_addr == INADDR_ANY)
652 		ip->ip_src = src;
653 
654 	/*
655 	 * Look for broadcast address and
656 	 * verify user is allowed to send
657 	 * such a packet.
658 	 */
659 	if (isbroadcast) {
660 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
661 			error = EADDRNOTAVAIL;
662 			goto bad;
663 		}
664 		if ((flags & IP_ALLOWBROADCAST) == 0) {
665 			error = EACCES;
666 			goto bad;
667 		}
668 		/* don't allow broadcast messages to be fragmented */
669 		if (ip_len > mtu) {
670 			error = EMSGSIZE;
671 			goto bad;
672 		}
673 		m->m_flags |= M_BCAST;
674 	} else {
675 		m->m_flags &= ~M_BCAST;
676 	}
677 
678 sendit:
679 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
680 	if (IPSEC_ENABLED(ipv4)) {
681 		struct ip ip_hdr;
682 
683 		if ((error = IPSEC_OUTPUT(ipv4, ifp, m, inp, mtu)) != 0) {
684 			if (error == EINPROGRESS)
685 				error = 0;
686 			goto done;
687 		}
688 
689 		/* Update variables that are affected by ipsec4_output(). */
690 		m_copydata(m, 0, sizeof(ip_hdr), (char *)&ip_hdr);
691 		hlen = ip_hdr.ip_hl << 2;
692 	}
693 
694 	/*
695 	 * Check if there was a route for this packet; return error if not.
696 	 */
697 	if (no_route_but_check_spd) {
698 		IPSTAT_INC(ips_noroute);
699 		error = EHOSTUNREACH;
700 		goto bad;
701 	}
702 #endif /* IPSEC */
703 
704 	/* Jump over all PFIL processing if hooks are not active. */
705 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
706 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
707 		    &error)) {
708 		case 1: /* Finished */
709 			goto done;
710 
711 		case 0: /* Continue normally */
712 			ip = mtod(m, struct ip *);
713 			ip_len = ntohs(ip->ip_len);
714 			break;
715 
716 		case -1: /* Need to try again */
717 			/* Reset everything for a new round */
718 			if (ro != NULL) {
719 				RO_NHFREE(ro);
720 				ro->ro_prepend = NULL;
721 			}
722 			gw = (const struct sockaddr *)dst;
723 			ip = mtod(m, struct ip *);
724 			goto again;
725 		}
726 	}
727 
728 	if (vlan_pcp > -1)
729 		EVL_APPLY_PRI(m, vlan_pcp);
730 
731 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
732 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
733 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
734 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
735 			IPSTAT_INC(ips_badaddr);
736 			error = EADDRNOTAVAIL;
737 			goto bad;
738 		}
739 	}
740 
741 	/* Ensure the packet data is mapped if the interface requires it. */
742 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
743 		struct mbuf *m1;
744 
745 		error = mb_unmapped_to_ext(m, &m1);
746 		if (error != 0) {
747 			if (error == EINVAL) {
748 				if_printf(ifp, "TLS packet\n");
749 				/* XXXKIB */
750 			} else if (error == ENOMEM) {
751 				error = ENOBUFS;
752 			}
753 			IPSTAT_INC(ips_odropped);
754 			goto done;
755 		} else {
756 			m = m1;
757 		}
758 	}
759 
760 	m->m_pkthdr.csum_flags |= CSUM_IP;
761 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
762 		in_delayed_cksum(m);
763 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
764 	}
765 #if defined(SCTP) || defined(SCTP_SUPPORT)
766 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
767 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
768 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
769 	}
770 #endif
771 
772 	/*
773 	 * If small enough for interface, or the interface will take
774 	 * care of the fragmentation for us, we can just send directly.
775 	 * Note that if_vxlan could have requested TSO even though the outer
776 	 * frame is UDP.  It is correct to not fragment such datagrams and
777 	 * instead just pass them on to the driver.
778 	 */
779 	if (ip_len <= mtu ||
780 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
781 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
782 		ip->ip_sum = 0;
783 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
784 			ip->ip_sum = in_cksum(m, hlen);
785 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
786 		}
787 
788 		/*
789 		 * Record statistics for this interface address.
790 		 * With CSUM_TSO the byte/packet count will be slightly
791 		 * incorrect because we count the IP+TCP headers only
792 		 * once instead of for every generated packet.
793 		 */
794 		if (!(flags & IP_FORWARDING) && ia) {
795 			if (m->m_pkthdr.csum_flags &
796 			    (CSUM_TSO | CSUM_INNER_TSO))
797 				counter_u64_add(ia->ia_ifa.ifa_opackets,
798 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
799 			else
800 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
801 
802 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
803 		}
804 #ifdef MBUF_STRESS_TEST
805 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
806 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
807 #endif
808 		/*
809 		 * Reset layer specific mbuf flags
810 		 * to avoid confusing lower layers.
811 		 */
812 		m_clrprotoflags(m);
813 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
814 		error = ip_output_send(inp, ifp, m, gw, ro,
815 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
816 		goto done;
817 	}
818 
819 	/* Balk when DF bit is set or the interface didn't support TSO. */
820 	if ((ip_off & IP_DF) ||
821 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
822 		error = EMSGSIZE;
823 		IPSTAT_INC(ips_cantfrag);
824 		goto bad;
825 	}
826 
827 	/*
828 	 * Too large for interface; fragment if possible. If successful,
829 	 * on return, m will point to a list of packets to be sent.
830 	 */
831 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
832 	if (error)
833 		goto bad;
834 	for (; m; m = m0) {
835 		m0 = m->m_nextpkt;
836 		m->m_nextpkt = 0;
837 		if (error == 0) {
838 			/* Record statistics for this interface address. */
839 			if (ia != NULL) {
840 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
841 				counter_u64_add(ia->ia_ifa.ifa_obytes,
842 				    m->m_pkthdr.len);
843 			}
844 			/*
845 			 * Reset layer specific mbuf flags
846 			 * to avoid confusing upper layers.
847 			 */
848 			m_clrprotoflags(m);
849 
850 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
851 			    mtod(m, struct ip *), NULL);
852 			error = ip_output_send(inp, ifp, m, gw, ro, true);
853 		} else
854 			m_freem(m);
855 	}
856 
857 	if (error == 0)
858 		IPSTAT_INC(ips_fragmented);
859 
860 done:
861 	return (error);
862 bad:
863 	m_freem(m);
864 	goto done;
865 }
866 
867 /*
868  * Create a chain of fragments which fit the given mtu. m_frag points to the
869  * mbuf to be fragmented; on return it points to the chain with the fragments.
870  * Return 0 if no error. If error, m_frag may contain a partially built
871  * chain of fragments that should be freed by the caller.
872  *
873  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
874  */
875 int
ip_fragment(struct ip * ip,struct mbuf ** m_frag,int mtu,u_long if_hwassist_flags)876 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
877     u_long if_hwassist_flags)
878 {
879 	int error = 0;
880 	int hlen = ip->ip_hl << 2;
881 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
882 	int off;
883 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
884 	int firstlen;
885 	struct mbuf **mnext;
886 	int nfrags;
887 	uint16_t ip_len, ip_off;
888 
889 	ip_len = ntohs(ip->ip_len);
890 	ip_off = ntohs(ip->ip_off);
891 
892 	/*
893 	 * Packet shall not have "Don't Fragment" flag and have at least 8
894 	 * bytes of payload.
895 	 */
896 	if (__predict_false((ip_off & IP_DF) || len < 8)) {
897 		IPSTAT_INC(ips_cantfrag);
898 		return (EMSGSIZE);
899 	}
900 
901 	/*
902 	 * If the interface will not calculate checksums on
903 	 * fragmented packets, then do it here.
904 	 */
905 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
906 		in_delayed_cksum(m0);
907 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
908 	}
909 #if defined(SCTP) || defined(SCTP_SUPPORT)
910 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
911 		sctp_delayed_cksum(m0, hlen);
912 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
913 	}
914 #endif
915 	if (len > PAGE_SIZE) {
916 		/*
917 		 * Fragment large datagrams such that each segment
918 		 * contains a multiple of PAGE_SIZE amount of data,
919 		 * plus headers. This enables a receiver to perform
920 		 * page-flipping zero-copy optimizations.
921 		 *
922 		 * XXX When does this help given that sender and receiver
923 		 * could have different page sizes, and also mtu could
924 		 * be less than the receiver's page size ?
925 		 */
926 		int newlen;
927 
928 		off = MIN(mtu, m0->m_pkthdr.len);
929 
930 		/*
931 		 * firstlen (off - hlen) must be aligned on an
932 		 * 8-byte boundary
933 		 */
934 		if (off < hlen)
935 			goto smart_frag_failure;
936 		off = ((off - hlen) & ~7) + hlen;
937 		newlen = (~PAGE_MASK) & mtu;
938 		if ((newlen + sizeof (struct ip)) > mtu) {
939 			/* we failed, go back the default */
940 smart_frag_failure:
941 			newlen = len;
942 			off = hlen + len;
943 		}
944 		len = newlen;
945 
946 	} else {
947 		off = hlen + len;
948 	}
949 
950 	firstlen = off - hlen;
951 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
952 
953 	/*
954 	 * Loop through length of segment after first fragment,
955 	 * make new header and copy data of each part and link onto chain.
956 	 * Here, m0 is the original packet, m is the fragment being created.
957 	 * The fragments are linked off the m_nextpkt of the original
958 	 * packet, which after processing serves as the first fragment.
959 	 */
960 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
961 		struct ip *mhip;	/* ip header on the fragment */
962 		struct mbuf *m;
963 		int mhlen = sizeof (struct ip);
964 
965 		m = m_gethdr(M_NOWAIT, MT_DATA);
966 		if (m == NULL) {
967 			error = ENOBUFS;
968 			IPSTAT_INC(ips_odropped);
969 			goto done;
970 		}
971 		/*
972 		 * Make sure the complete packet header gets copied
973 		 * from the originating mbuf to the newly created
974 		 * mbuf. This also ensures that existing firewall
975 		 * classification(s), VLAN tags and so on get copied
976 		 * to the resulting fragmented packet(s):
977 		 */
978 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
979 			m_free(m);
980 			error = ENOBUFS;
981 			IPSTAT_INC(ips_odropped);
982 			goto done;
983 		}
984 		/*
985 		 * In the first mbuf, leave room for the link header, then
986 		 * copy the original IP header including options. The payload
987 		 * goes into an additional mbuf chain returned by m_copym().
988 		 */
989 		m->m_data += max_linkhdr;
990 		mhip = mtod(m, struct ip *);
991 		*mhip = *ip;
992 		if (hlen > sizeof (struct ip)) {
993 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
994 			mhip->ip_v = IPVERSION;
995 			mhip->ip_hl = mhlen >> 2;
996 		}
997 		m->m_len = mhlen;
998 		/* XXX do we need to add ip_off below ? */
999 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1000 		if (off + len >= ip_len)
1001 			len = ip_len - off;
1002 		else
1003 			mhip->ip_off |= IP_MF;
1004 		mhip->ip_len = htons((u_short)(len + mhlen));
1005 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1006 		if (m->m_next == NULL) {	/* copy failed */
1007 			m_free(m);
1008 			error = ENOBUFS;	/* ??? */
1009 			IPSTAT_INC(ips_odropped);
1010 			goto done;
1011 		}
1012 		m->m_pkthdr.len = mhlen + len;
1013 #ifdef MAC
1014 		mac_netinet_fragment(m0, m);
1015 #endif
1016 		mhip->ip_off = htons(mhip->ip_off);
1017 		mhip->ip_sum = 0;
1018 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1019 			mhip->ip_sum = in_cksum(m, mhlen);
1020 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1021 		}
1022 		*mnext = m;
1023 		mnext = &m->m_nextpkt;
1024 	}
1025 	IPSTAT_ADD(ips_ofragments, nfrags);
1026 
1027 	/*
1028 	 * Update first fragment by trimming what's been copied out
1029 	 * and updating header.
1030 	 */
1031 	m_adj(m0, hlen + firstlen - ip_len);
1032 	m0->m_pkthdr.len = hlen + firstlen;
1033 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1034 	ip->ip_off = htons(ip_off | IP_MF);
1035 	ip->ip_sum = 0;
1036 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1037 		ip->ip_sum = in_cksum(m0, hlen);
1038 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1039 	}
1040 
1041 done:
1042 	*m_frag = m0;
1043 	return error;
1044 }
1045 
1046 void
in_delayed_cksum(struct mbuf * m)1047 in_delayed_cksum(struct mbuf *m)
1048 {
1049 	struct ip *ip;
1050 	struct udphdr *uh;
1051 	uint16_t cklen, csum, offset;
1052 
1053 	ip = mtod(m, struct ip *);
1054 	offset = ip->ip_hl << 2 ;
1055 
1056 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1057 		/* if udp header is not in the first mbuf copy udplen */
1058 		if (offset + sizeof(struct udphdr) > m->m_len) {
1059 			m_copydata(m, offset + offsetof(struct udphdr,
1060 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1061 			cklen = ntohs(cklen);
1062 		} else {
1063 			uh = (struct udphdr *)mtodo(m, offset);
1064 			cklen = ntohs(uh->uh_ulen);
1065 		}
1066 		csum = in_cksum_skip(m, cklen + offset, offset);
1067 		if (csum == 0)
1068 			csum = 0xffff;
1069 	} else {
1070 		cklen = ntohs(ip->ip_len);
1071 		csum = in_cksum_skip(m, cklen, offset);
1072 	}
1073 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1074 
1075 	if (offset + sizeof(csum) > m->m_len)
1076 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1077 	else
1078 		*(u_short *)mtodo(m, offset) = csum;
1079 }
1080 
1081 /*
1082  * IP socket option processing.
1083  */
1084 int
ip_ctloutput(struct socket * so,struct sockopt * sopt)1085 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1086 {
1087 	struct inpcb *inp = sotoinpcb(so);
1088 	int	error, optval;
1089 #ifdef	RSS
1090 	uint32_t rss_bucket;
1091 	int retval;
1092 #endif
1093 
1094 	error = optval = 0;
1095 	if (sopt->sopt_level != IPPROTO_IP) {
1096 		error = EINVAL;
1097 
1098 		if (sopt->sopt_level == SOL_SOCKET &&
1099 		    sopt->sopt_dir == SOPT_SET) {
1100 			switch (sopt->sopt_name) {
1101 			case SO_SETFIB:
1102 				error = sooptcopyin(sopt, &optval,
1103 				    sizeof(optval), sizeof(optval));
1104 				if (error != 0)
1105 					break;
1106 
1107 				INP_WLOCK(inp);
1108 				if ((inp->inp_flags & INP_BOUNDFIB) != 0 &&
1109 				    optval != so->so_fibnum) {
1110 					INP_WUNLOCK(inp);
1111 					error = EISCONN;
1112 					break;
1113 				}
1114 				error = sosetfib(inp->inp_socket, optval);
1115 				if (error == 0)
1116 					inp->inp_inc.inc_fibnum = optval;
1117 				INP_WUNLOCK(inp);
1118 				break;
1119 			case SO_MAX_PACING_RATE:
1120 #ifdef RATELIMIT
1121 				INP_WLOCK(inp);
1122 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1123 				INP_WUNLOCK(inp);
1124 				error = 0;
1125 #else
1126 				error = EOPNOTSUPP;
1127 #endif
1128 				break;
1129 			default:
1130 				break;
1131 			}
1132 		}
1133 		return (error);
1134 	}
1135 
1136 	switch (sopt->sopt_dir) {
1137 	case SOPT_SET:
1138 		switch (sopt->sopt_name) {
1139 		case IP_OPTIONS:
1140 #ifdef notyet
1141 		case IP_RETOPTS:
1142 #endif
1143 		{
1144 			struct mbuf *m;
1145 			if (sopt->sopt_valsize > MLEN) {
1146 				error = EMSGSIZE;
1147 				break;
1148 			}
1149 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1150 			if (m == NULL) {
1151 				error = ENOBUFS;
1152 				break;
1153 			}
1154 			m->m_len = sopt->sopt_valsize;
1155 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1156 					    m->m_len);
1157 			if (error) {
1158 				m_free(m);
1159 				break;
1160 			}
1161 			INP_WLOCK(inp);
1162 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1163 			INP_WUNLOCK(inp);
1164 			return (error);
1165 		}
1166 
1167 		case IP_BINDANY:
1168 			if (sopt->sopt_td != NULL) {
1169 				error = priv_check(sopt->sopt_td,
1170 				    PRIV_NETINET_BINDANY);
1171 				if (error)
1172 					break;
1173 			}
1174 			/* FALLTHROUGH */
1175 		case IP_TOS:
1176 		case IP_TTL:
1177 		case IP_MINTTL:
1178 		case IP_RECVOPTS:
1179 		case IP_RECVRETOPTS:
1180 		case IP_ORIGDSTADDR:
1181 		case IP_RECVDSTADDR:
1182 		case IP_RECVTTL:
1183 		case IP_RECVIF:
1184 		case IP_ONESBCAST:
1185 		case IP_DONTFRAG:
1186 		case IP_RECVTOS:
1187 		case IP_RECVFLOWID:
1188 #ifdef	RSS
1189 		case IP_RECVRSSBUCKETID:
1190 #endif
1191 		case IP_VLAN_PCP:
1192 			error = sooptcopyin(sopt, &optval, sizeof optval,
1193 					    sizeof optval);
1194 			if (error)
1195 				break;
1196 
1197 			switch (sopt->sopt_name) {
1198 			case IP_TOS:
1199 				inp->inp_ip_tos = optval;
1200 				break;
1201 
1202 			case IP_TTL:
1203 				inp->inp_ip_ttl = optval;
1204 				break;
1205 
1206 			case IP_MINTTL:
1207 				if (optval >= 0 && optval <= MAXTTL)
1208 					inp->inp_ip_minttl = optval;
1209 				else
1210 					error = EINVAL;
1211 				break;
1212 
1213 #define	OPTSET(bit) do {						\
1214 	INP_WLOCK(inp);							\
1215 	if (optval)							\
1216 		inp->inp_flags |= bit;					\
1217 	else								\
1218 		inp->inp_flags &= ~bit;					\
1219 	INP_WUNLOCK(inp);						\
1220 } while (0)
1221 
1222 #define	OPTSET2(bit, val) do {						\
1223 	INP_WLOCK(inp);							\
1224 	if (val)							\
1225 		inp->inp_flags2 |= bit;					\
1226 	else								\
1227 		inp->inp_flags2 &= ~bit;				\
1228 	INP_WUNLOCK(inp);						\
1229 } while (0)
1230 
1231 			case IP_RECVOPTS:
1232 				OPTSET(INP_RECVOPTS);
1233 				break;
1234 
1235 			case IP_RECVRETOPTS:
1236 				OPTSET(INP_RECVRETOPTS);
1237 				break;
1238 
1239 			case IP_RECVDSTADDR:
1240 				OPTSET(INP_RECVDSTADDR);
1241 				break;
1242 
1243 			case IP_ORIGDSTADDR:
1244 				OPTSET2(INP_ORIGDSTADDR, optval);
1245 				break;
1246 
1247 			case IP_RECVTTL:
1248 				OPTSET(INP_RECVTTL);
1249 				break;
1250 
1251 			case IP_RECVIF:
1252 				OPTSET(INP_RECVIF);
1253 				break;
1254 
1255 			case IP_ONESBCAST:
1256 				OPTSET(INP_ONESBCAST);
1257 				break;
1258 			case IP_DONTFRAG:
1259 				OPTSET(INP_DONTFRAG);
1260 				break;
1261 			case IP_BINDANY:
1262 				OPTSET(INP_BINDANY);
1263 				break;
1264 			case IP_RECVTOS:
1265 				OPTSET(INP_RECVTOS);
1266 				break;
1267 			case IP_RECVFLOWID:
1268 				OPTSET2(INP_RECVFLOWID, optval);
1269 				break;
1270 #ifdef RSS
1271 			case IP_RECVRSSBUCKETID:
1272 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1273 				break;
1274 #endif
1275 			case IP_VLAN_PCP:
1276 				if ((optval >= -1) && (optval <=
1277 				    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1278 					if (optval == -1) {
1279 						INP_WLOCK(inp);
1280 						inp->inp_flags2 &=
1281 						    ~(INP_2PCP_SET |
1282 						      INP_2PCP_MASK);
1283 						INP_WUNLOCK(inp);
1284 					} else {
1285 						INP_WLOCK(inp);
1286 						inp->inp_flags2 |=
1287 						    INP_2PCP_SET;
1288 						inp->inp_flags2 &=
1289 						    ~INP_2PCP_MASK;
1290 						inp->inp_flags2 |=
1291 						    optval << INP_2PCP_SHIFT;
1292 						INP_WUNLOCK(inp);
1293 					}
1294 				} else
1295 					error = EINVAL;
1296 				break;
1297 			}
1298 			break;
1299 #undef OPTSET
1300 #undef OPTSET2
1301 
1302 		/*
1303 		 * Multicast socket options are processed by the in_mcast
1304 		 * module.
1305 		 */
1306 		case IP_MULTICAST_IF:
1307 		case IP_MULTICAST_VIF:
1308 		case IP_MULTICAST_TTL:
1309 		case IP_MULTICAST_LOOP:
1310 		case IP_ADD_MEMBERSHIP:
1311 		case IP_DROP_MEMBERSHIP:
1312 		case IP_ADD_SOURCE_MEMBERSHIP:
1313 		case IP_DROP_SOURCE_MEMBERSHIP:
1314 		case IP_BLOCK_SOURCE:
1315 		case IP_UNBLOCK_SOURCE:
1316 		case IP_MSFILTER:
1317 		case MCAST_JOIN_GROUP:
1318 		case MCAST_LEAVE_GROUP:
1319 		case MCAST_JOIN_SOURCE_GROUP:
1320 		case MCAST_LEAVE_SOURCE_GROUP:
1321 		case MCAST_BLOCK_SOURCE:
1322 		case MCAST_UNBLOCK_SOURCE:
1323 			error = inp_setmoptions(inp, sopt);
1324 			break;
1325 
1326 		case IP_PORTRANGE:
1327 			error = sooptcopyin(sopt, &optval, sizeof optval,
1328 					    sizeof optval);
1329 			if (error)
1330 				break;
1331 
1332 			INP_WLOCK(inp);
1333 			switch (optval) {
1334 			case IP_PORTRANGE_DEFAULT:
1335 				inp->inp_flags &= ~(INP_LOWPORT);
1336 				inp->inp_flags &= ~(INP_HIGHPORT);
1337 				break;
1338 
1339 			case IP_PORTRANGE_HIGH:
1340 				inp->inp_flags &= ~(INP_LOWPORT);
1341 				inp->inp_flags |= INP_HIGHPORT;
1342 				break;
1343 
1344 			case IP_PORTRANGE_LOW:
1345 				inp->inp_flags &= ~(INP_HIGHPORT);
1346 				inp->inp_flags |= INP_LOWPORT;
1347 				break;
1348 
1349 			default:
1350 				error = EINVAL;
1351 				break;
1352 			}
1353 			INP_WUNLOCK(inp);
1354 			break;
1355 
1356 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1357 		case IP_IPSEC_POLICY:
1358 			if (IPSEC_ENABLED(ipv4)) {
1359 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1360 				break;
1361 			}
1362 			/* FALLTHROUGH */
1363 #endif /* IPSEC */
1364 
1365 		default:
1366 			error = ENOPROTOOPT;
1367 			break;
1368 		}
1369 		break;
1370 
1371 	case SOPT_GET:
1372 		switch (sopt->sopt_name) {
1373 		case IP_OPTIONS:
1374 		case IP_RETOPTS:
1375 			INP_RLOCK(inp);
1376 			if (inp->inp_options) {
1377 				struct mbuf *options;
1378 
1379 				options = m_copym(inp->inp_options, 0,
1380 				    M_COPYALL, M_NOWAIT);
1381 				INP_RUNLOCK(inp);
1382 				if (options != NULL) {
1383 					error = sooptcopyout(sopt,
1384 							     mtod(options, char *),
1385 							     options->m_len);
1386 					m_freem(options);
1387 				} else
1388 					error = ENOMEM;
1389 			} else {
1390 				INP_RUNLOCK(inp);
1391 				sopt->sopt_valsize = 0;
1392 			}
1393 			break;
1394 
1395 		case IP_TOS:
1396 		case IP_TTL:
1397 		case IP_MINTTL:
1398 		case IP_RECVOPTS:
1399 		case IP_RECVRETOPTS:
1400 		case IP_ORIGDSTADDR:
1401 		case IP_RECVDSTADDR:
1402 		case IP_RECVTTL:
1403 		case IP_RECVIF:
1404 		case IP_PORTRANGE:
1405 		case IP_ONESBCAST:
1406 		case IP_DONTFRAG:
1407 		case IP_BINDANY:
1408 		case IP_RECVTOS:
1409 		case IP_FLOWID:
1410 		case IP_FLOWTYPE:
1411 		case IP_RECVFLOWID:
1412 #ifdef	RSS
1413 		case IP_RSSBUCKETID:
1414 		case IP_RECVRSSBUCKETID:
1415 #endif
1416 		case IP_VLAN_PCP:
1417 			switch (sopt->sopt_name) {
1418 			case IP_TOS:
1419 				optval = inp->inp_ip_tos;
1420 				break;
1421 
1422 			case IP_TTL:
1423 				optval = inp->inp_ip_ttl;
1424 				break;
1425 
1426 			case IP_MINTTL:
1427 				optval = inp->inp_ip_minttl;
1428 				break;
1429 
1430 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1431 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1432 
1433 			case IP_RECVOPTS:
1434 				optval = OPTBIT(INP_RECVOPTS);
1435 				break;
1436 
1437 			case IP_RECVRETOPTS:
1438 				optval = OPTBIT(INP_RECVRETOPTS);
1439 				break;
1440 
1441 			case IP_RECVDSTADDR:
1442 				optval = OPTBIT(INP_RECVDSTADDR);
1443 				break;
1444 
1445 			case IP_ORIGDSTADDR:
1446 				optval = OPTBIT2(INP_ORIGDSTADDR);
1447 				break;
1448 
1449 			case IP_RECVTTL:
1450 				optval = OPTBIT(INP_RECVTTL);
1451 				break;
1452 
1453 			case IP_RECVIF:
1454 				optval = OPTBIT(INP_RECVIF);
1455 				break;
1456 
1457 			case IP_PORTRANGE:
1458 				if (inp->inp_flags & INP_HIGHPORT)
1459 					optval = IP_PORTRANGE_HIGH;
1460 				else if (inp->inp_flags & INP_LOWPORT)
1461 					optval = IP_PORTRANGE_LOW;
1462 				else
1463 					optval = 0;
1464 				break;
1465 
1466 			case IP_ONESBCAST:
1467 				optval = OPTBIT(INP_ONESBCAST);
1468 				break;
1469 			case IP_DONTFRAG:
1470 				optval = OPTBIT(INP_DONTFRAG);
1471 				break;
1472 			case IP_BINDANY:
1473 				optval = OPTBIT(INP_BINDANY);
1474 				break;
1475 			case IP_RECVTOS:
1476 				optval = OPTBIT(INP_RECVTOS);
1477 				break;
1478 			case IP_FLOWID:
1479 				optval = inp->inp_flowid;
1480 				break;
1481 			case IP_FLOWTYPE:
1482 				optval = inp->inp_flowtype;
1483 				break;
1484 			case IP_RECVFLOWID:
1485 				optval = OPTBIT2(INP_RECVFLOWID);
1486 				break;
1487 #ifdef	RSS
1488 			case IP_RSSBUCKETID:
1489 				retval = rss_hash2bucket(inp->inp_flowid,
1490 				    inp->inp_flowtype,
1491 				    &rss_bucket);
1492 				if (retval == 0)
1493 					optval = rss_bucket;
1494 				else
1495 					error = EINVAL;
1496 				break;
1497 			case IP_RECVRSSBUCKETID:
1498 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1499 				break;
1500 #endif
1501 			case IP_VLAN_PCP:
1502 				if (OPTBIT2(INP_2PCP_SET)) {
1503 					optval = (inp->inp_flags2 &
1504 					    INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1505 				} else {
1506 					optval = -1;
1507 				}
1508 				break;
1509 			}
1510 			error = sooptcopyout(sopt, &optval, sizeof optval);
1511 			break;
1512 
1513 		/*
1514 		 * Multicast socket options are processed by the in_mcast
1515 		 * module.
1516 		 */
1517 		case IP_MULTICAST_IF:
1518 		case IP_MULTICAST_VIF:
1519 		case IP_MULTICAST_TTL:
1520 		case IP_MULTICAST_LOOP:
1521 		case IP_MSFILTER:
1522 			error = inp_getmoptions(inp, sopt);
1523 			break;
1524 
1525 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1526 		case IP_IPSEC_POLICY:
1527 			if (IPSEC_ENABLED(ipv4)) {
1528 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1529 				break;
1530 			}
1531 			/* FALLTHROUGH */
1532 #endif /* IPSEC */
1533 
1534 		default:
1535 			error = ENOPROTOOPT;
1536 			break;
1537 		}
1538 		break;
1539 	}
1540 	return (error);
1541 }
1542 
1543 /*
1544  * Routine called from ip_output() to loop back a copy of an IP multicast
1545  * packet to the input queue of a specified interface.  Note that this
1546  * calls the output routine of the loopback "driver", but with an interface
1547  * pointer that might NOT be a loopback interface -- evil, but easier than
1548  * replicating that code here.
1549  */
1550 static void
ip_mloopback(struct ifnet * ifp,const struct mbuf * m,int hlen)1551 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1552 {
1553 	struct ip *ip;
1554 	struct mbuf *copym;
1555 
1556 	/*
1557 	 * Make a deep copy of the packet because we're going to
1558 	 * modify the pack in order to generate checksums.
1559 	 */
1560 	copym = m_dup(m, M_NOWAIT);
1561 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1562 		copym = m_pullup(copym, hlen);
1563 	if (copym != NULL) {
1564 		/* If needed, compute the checksum and mark it as valid. */
1565 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1566 			in_delayed_cksum(copym);
1567 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1568 			copym->m_pkthdr.csum_flags |=
1569 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1570 			copym->m_pkthdr.csum_data = 0xffff;
1571 		}
1572 		/*
1573 		 * We don't bother to fragment if the IP length is greater
1574 		 * than the interface's MTU.  Can this possibly matter?
1575 		 */
1576 		ip = mtod(copym, struct ip *);
1577 		ip->ip_sum = 0;
1578 		ip->ip_sum = in_cksum(copym, hlen);
1579 		if_simloop(ifp, copym, AF_INET, 0);
1580 	}
1581 }
1582