xref: /freebsd/sys/netinet/ip_output.c (revision 7b71f57f4e514a2ab7308ce4147e14d90e099ad0)
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 "opt_inet.h"
33 #include "opt_ipsec.h"
34 #include "opt_kern_tls.h"
35 #include "opt_mbuf_stress_test.h"
36 #include "opt_ratelimit.h"
37 #include "opt_route.h"
38 #include "opt_rss.h"
39 #include "opt_sctp.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/ktls.h>
45 #include <sys/lock.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/priv.h>
49 #include <sys/proc.h>
50 #include <sys/protosw.h>
51 #include <sys/sdt.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/sysctl.h>
55 #include <sys/ucred.h>
56 
57 #include <net/if.h>
58 #include <net/if_var.h>
59 #include <net/if_private.h>
60 #include <net/if_vlan_var.h>
61 #include <net/if_llatbl.h>
62 #include <net/ethernet.h>
63 #include <net/netisr.h>
64 #include <net/pfil.h>
65 #include <net/route.h>
66 #include <net/route/nhop.h>
67 #include <net/rss_config.h>
68 #include <net/vnet.h>
69 
70 #include <netinet/in.h>
71 #include <netinet/in_fib.h>
72 #include <netinet/in_kdtrace.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/ip.h>
75 #include <netinet/in_fib.h>
76 #include <netinet/in_pcb.h>
77 #include <netinet/in_rss.h>
78 #include <netinet/in_var.h>
79 #include <netinet/ip_var.h>
80 #include <netinet/ip_options.h>
81 
82 #include <netinet/udp.h>
83 #include <netinet/udp_var.h>
84 
85 #if defined(SCTP) || defined(SCTP_SUPPORT)
86 #include <netinet/sctp.h>
87 #include <netinet/sctp_crc32.h>
88 #endif
89 
90 #include <netipsec/ipsec_support.h>
91 
92 #include <machine/in_cksum.h>
93 
94 #include <security/mac/mac_framework.h>
95 
96 #ifdef MBUF_STRESS_TEST
97 static int mbuf_frag_size = 0;
98 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
99 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
100 #endif
101 
102 static void	ip_mloopback(struct ifnet *, const struct mbuf *, int);
103 
104 extern int in_mcast_loop;
105 
106 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)107 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
108     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
109 {
110 	struct m_tag *fwd_tag = NULL;
111 	struct mbuf *m;
112 	struct in_addr odst;
113 	struct ip *ip;
114 	int ret;
115 
116 	m = *mp;
117 	ip = mtod(m, struct ip *);
118 
119 	/* Run through list of hooks for output packets. */
120 	odst.s_addr = ip->ip_dst.s_addr;
121 	if (flags & IP_FORWARDING)
122 		ret = pfil_mbuf_fwd(V_inet_pfil_head, mp, ifp, inp);
123 	else
124 		ret = pfil_mbuf_out(V_inet_pfil_head, mp, ifp, inp);
125 
126 	switch (ret) {
127 	case PFIL_DROPPED:
128 		*error = EACCES;
129 		/* FALLTHROUGH */
130 	case PFIL_CONSUMED:
131 		return 1; /* Finished */
132 	case PFIL_PASS:
133 		*error = 0;
134 	}
135 	m = *mp;
136 	ip = mtod(m, struct ip *);
137 
138 	/* See if destination IP address was changed by packet filter. */
139 	if (odst.s_addr != ip->ip_dst.s_addr) {
140 		m->m_flags |= M_SKIP_FIREWALL;
141 		/* If destination is now ourself drop to ip_input(). */
142 		if (in_localip(ip->ip_dst)) {
143 			m->m_flags |= M_FASTFWD_OURS;
144 			if (m->m_pkthdr.rcvif == NULL)
145 				m->m_pkthdr.rcvif = V_loif;
146 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
147 				m->m_pkthdr.csum_flags |=
148 					CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
149 				m->m_pkthdr.csum_data = 0xffff;
150 			}
151 			m->m_pkthdr.csum_flags |=
152 				CSUM_IP_CHECKED | CSUM_IP_VALID;
153 #if defined(SCTP) || defined(SCTP_SUPPORT)
154 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
155 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
156 #endif
157 			*error = netisr_queue(NETISR_IP, m);
158 			return 1; /* Finished */
159 		}
160 
161 		bzero(dst, sizeof(*dst));
162 		dst->sin_family = AF_INET;
163 		dst->sin_len = sizeof(*dst);
164 		dst->sin_addr = ip->ip_dst;
165 
166 		return -1; /* Reloop */
167 	}
168 	/* See if fib was changed by packet filter. */
169 	if ((*fibnum) != M_GETFIB(m)) {
170 		m->m_flags |= M_SKIP_FIREWALL;
171 		*fibnum = M_GETFIB(m);
172 		return -1; /* Reloop for FIB change */
173 	}
174 
175 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
176 	if (m->m_flags & M_FASTFWD_OURS) {
177 		if (m->m_pkthdr.rcvif == NULL)
178 			m->m_pkthdr.rcvif = V_loif;
179 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
180 			m->m_pkthdr.csum_flags |=
181 				CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
182 			m->m_pkthdr.csum_data = 0xffff;
183 		}
184 #if defined(SCTP) || defined(SCTP_SUPPORT)
185 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
186 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
187 #endif
188 		m->m_pkthdr.csum_flags |=
189 			CSUM_IP_CHECKED | CSUM_IP_VALID;
190 
191 		*error = netisr_queue(NETISR_IP, m);
192 		return 1; /* Finished */
193 	}
194 	/* Or forward to some other address? */
195 	if ((m->m_flags & M_IP_NEXTHOP) &&
196 	    ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
197 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
198 		m->m_flags |= M_SKIP_FIREWALL;
199 		m->m_flags &= ~M_IP_NEXTHOP;
200 		m_tag_delete(m, fwd_tag);
201 
202 		return -1; /* Reloop for CHANGE of dst */
203 	}
204 
205 	return 0;
206 }
207 
208 static int
ip_output_send(struct inpcb * inp,struct ifnet * ifp,struct mbuf * m,const struct sockaddr * gw,struct route * ro,bool stamp_tag)209 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
210     const struct sockaddr *gw, struct route *ro, bool stamp_tag)
211 {
212 #ifdef KERN_TLS
213 	struct ktls_session *tls = NULL;
214 #endif
215 	struct m_snd_tag *mst;
216 	int error;
217 
218 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
219 	mst = NULL;
220 
221 #ifdef KERN_TLS
222 	/*
223 	 * If this is an unencrypted TLS record, save a reference to
224 	 * the record.  This local reference is used to call
225 	 * ktls_output_eagain after the mbuf has been freed (thus
226 	 * dropping the mbuf's reference) in if_output.
227 	 */
228 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
229 		tls = ktls_hold(m->m_next->m_epg_tls);
230 		mst = tls->snd_tag;
231 
232 		/*
233 		 * If a TLS session doesn't have a valid tag, it must
234 		 * have had an earlier ifp mismatch, so drop this
235 		 * packet.
236 		 */
237 		if (mst == NULL) {
238 			m_freem(m);
239 			error = EAGAIN;
240 			goto done;
241 		}
242 		/*
243 		 * Always stamp tags that include NIC ktls.
244 		 */
245 		stamp_tag = true;
246 	}
247 #endif
248 #ifdef RATELIMIT
249 	if (inp != NULL && mst == NULL) {
250 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
251 		    (inp->inp_snd_tag != NULL &&
252 		    inp->inp_snd_tag->ifp != ifp))
253 			in_pcboutput_txrtlmt(inp, ifp, m);
254 
255 		if (inp->inp_snd_tag != NULL)
256 			mst = inp->inp_snd_tag;
257 	}
258 #endif
259 	if (stamp_tag && mst != NULL) {
260 		KASSERT(m->m_pkthdr.rcvif == NULL,
261 		    ("trying to add a send tag to a forwarded packet"));
262 		if (mst->ifp != ifp) {
263 			m_freem(m);
264 			error = EAGAIN;
265 			goto done;
266 		}
267 
268 		/* stamp send tag on mbuf */
269 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
270 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
271 	}
272 
273 	error = (*ifp->if_output)(ifp, m, gw, ro);
274 
275 done:
276 	/* Check for route change invalidating send tags. */
277 #ifdef KERN_TLS
278 	if (tls != NULL) {
279 		if (error == EAGAIN)
280 			error = ktls_output_eagain(inp, tls);
281 		ktls_free(tls);
282 	}
283 #endif
284 #ifdef RATELIMIT
285 	if (error == EAGAIN)
286 		in_pcboutput_eagain(inp);
287 #endif
288 	return (error);
289 }
290 
291 /* rte<>ro_flags translation */
292 static inline void
rt_update_ro_flags(struct route * ro,const struct nhop_object * nh)293 rt_update_ro_flags(struct route *ro, const struct nhop_object *nh)
294 {
295 	int nh_flags = nh->nh_flags;
296 
297 	ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
298 
299 	ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
300 	ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
301 	ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
302 }
303 
304 /*
305  * IP output.  The packet in mbuf chain m contains a skeletal IP
306  * header (with len, off, ttl, proto, tos, src, dst).
307  * The mbuf chain containing the packet will be freed.
308  * The mbuf opt, if present, will not be freed.
309  * If route ro is present and has ro_rt initialized, route lookup would be
310  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
311  * then result of route lookup is stored in ro->ro_rt.
312  *
313  * In the IP forwarding case, the packet will arrive with options already
314  * inserted, so must have a NULL opt pointer.
315  */
316 int
ip_output(struct mbuf * m,struct mbuf * opt,struct route * ro,int flags,struct ip_moptions * imo,struct inpcb * inp)317 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
318     struct ip_moptions *imo, struct inpcb *inp)
319 {
320 	struct ip *ip;
321 	struct ifnet *ifp = NULL;	/* keep compiler happy */
322 	struct mbuf *m0;
323 	int hlen = sizeof (struct ip);
324 	int mtu = 0;
325 	int error = 0;
326 	int vlan_pcp = -1;
327 	struct sockaddr_in *dst;
328 	const struct sockaddr *gw;
329 	struct in_ifaddr *ia = NULL;
330 	struct in_addr src;
331 	bool isbroadcast;
332 	uint16_t ip_len, ip_off;
333 	struct route iproute;
334 	uint32_t fibnum;
335 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
336 	int no_route_but_check_spd = 0;
337 #endif
338 
339 	M_ASSERTPKTHDR(m);
340 	NET_EPOCH_ASSERT();
341 
342 	if (inp != NULL) {
343 		INP_LOCK_ASSERT(inp);
344 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
345 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
346 			m->m_pkthdr.flowid = inp->inp_flowid;
347 			M_HASHTYPE_SET(m, inp->inp_flowtype);
348 		}
349 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
350 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
351 			    INP_2PCP_SHIFT;
352 #ifdef NUMA
353 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
354 #endif
355 	}
356 
357 	if (opt) {
358 		int len = 0;
359 		m = ip_insertoptions(m, opt, &len);
360 		if (len != 0)
361 			hlen = len; /* ip->ip_hl is updated above */
362 	}
363 	ip = mtod(m, struct ip *);
364 	ip_len = ntohs(ip->ip_len);
365 	ip_off = ntohs(ip->ip_off);
366 
367 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
368 		ip->ip_v = IPVERSION;
369 		ip->ip_hl = hlen >> 2;
370 		ip_fillid(ip, V_ip_random_id);
371 	} else {
372 		/* Header already set, fetch hlen from there */
373 		hlen = ip->ip_hl << 2;
374 	}
375 	if ((flags & IP_FORWARDING) == 0)
376 		IPSTAT_INC(ips_localout);
377 
378 	/*
379 	 * dst/gw handling:
380 	 *
381 	 * gw is readonly but can point either to dst OR rt_gateway,
382 	 * therefore we need restore gw if we're redoing lookup.
383 	 */
384 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
385 	if (ro == NULL) {
386 		ro = &iproute;
387 		bzero(ro, sizeof (*ro));
388 	}
389 	dst = (struct sockaddr_in *)&ro->ro_dst;
390 	if (ro->ro_nh == NULL) {
391 		dst->sin_family = AF_INET;
392 		dst->sin_len = sizeof(*dst);
393 		dst->sin_addr = ip->ip_dst;
394 	}
395 	gw = (const struct sockaddr *)dst;
396 again:
397 	/*
398 	 * Validate route against routing table additions;
399 	 * a better/more specific route might have been added.
400 	 */
401 	if (inp != NULL && ro->ro_nh != NULL)
402 		NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
403 	/*
404 	 * If there is a cached route,
405 	 * check that it is to the same destination
406 	 * and is still up.  If not, free it and try again.
407 	 * The address family should also be checked in case of sharing the
408 	 * cache with IPv6.
409 	 * Also check whether routing cache needs invalidation.
410 	 */
411 	if (ro->ro_nh != NULL &&
412 	    ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
413 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr))
414 		RO_INVALIDATE_CACHE(ro);
415 	ia = NULL;
416 	/*
417 	 * If routing to interface only, short circuit routing lookup.
418 	 * The use of an all-ones broadcast address implies this; an
419 	 * interface is specified by the broadcast address of an interface,
420 	 * or the destination address of a ptp interface.
421 	 */
422 	if (flags & IP_SENDONES) {
423 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
424 						      M_GETFIB(m)))) == NULL &&
425 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
426 						    M_GETFIB(m)))) == NULL) {
427 			IPSTAT_INC(ips_noroute);
428 			error = ENETUNREACH;
429 			goto bad;
430 		}
431 		ip->ip_dst.s_addr = INADDR_BROADCAST;
432 		dst->sin_addr = ip->ip_dst;
433 		ifp = ia->ia_ifp;
434 		mtu = ifp->if_mtu;
435 		ip->ip_ttl = 1;
436 		isbroadcast = true;
437 		src = IA_SIN(ia)->sin_addr;
438 	} else if (flags & IP_ROUTETOIF) {
439 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
440 						    M_GETFIB(m)))) == NULL &&
441 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
442 						M_GETFIB(m)))) == NULL) {
443 			IPSTAT_INC(ips_noroute);
444 			error = ENETUNREACH;
445 			goto bad;
446 		}
447 		ifp = ia->ia_ifp;
448 		mtu = ifp->if_mtu;
449 		ip->ip_ttl = 1;
450 		isbroadcast = ifp->if_flags & IFF_BROADCAST ?
451 		    (in_broadcast(ip->ip_dst) ||
452 		    in_ifaddr_broadcast(dst->sin_addr, ia)) : 0;
453 		src = IA_SIN(ia)->sin_addr;
454 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
455 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
456 		/*
457 		 * Bypass the normal routing lookup for multicast
458 		 * packets if the interface is specified.
459 		 */
460 		ifp = imo->imo_multicast_ifp;
461 		mtu = ifp->if_mtu;
462 		IFP_TO_IA(ifp, ia);
463 		isbroadcast = false;
464 		/* Interface may have no addresses. */
465 		if (ia != NULL)
466 			src = IA_SIN(ia)->sin_addr;
467 		else
468 			src.s_addr = INADDR_ANY;
469 	} else if (ro != &iproute) {
470 		if (ro->ro_nh == NULL) {
471 			/*
472 			 * We want to do any cloning requested by the link
473 			 * layer, as this is probably required in all cases
474 			 * for correct operation (as it is for ARP).
475 			 */
476 			uint32_t flowid;
477 			flowid = m->m_pkthdr.flowid;
478 			ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
479 			    NHR_REF, flowid);
480 
481 			if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
482 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
483 				/*
484 				 * There is no route for this packet, but it is
485 				 * possible that a matching SPD entry exists.
486 				 */
487 				no_route_but_check_spd = 1;
488 				goto sendit;
489 #endif
490 				IPSTAT_INC(ips_noroute);
491 				error = EHOSTUNREACH;
492 				goto bad;
493 			}
494 		}
495 		struct nhop_object *nh = ro->ro_nh;
496 
497 		ia = ifatoia(nh->nh_ifa);
498 		ifp = nh->nh_ifp;
499 		counter_u64_add(nh->nh_pksent, 1);
500 		rt_update_ro_flags(ro, nh);
501 		if (nh->nh_flags & NHF_GATEWAY)
502 			gw = &nh->gw_sa;
503 		if (nh->nh_flags & NHF_HOST)
504 			isbroadcast = (nh->nh_flags & NHF_BROADCAST);
505 		else if ((ifp->if_flags & IFF_BROADCAST) &&
506 		    (gw->sa_family == AF_INET))
507 			isbroadcast = in_broadcast(ip->ip_dst) ||
508 			    in_ifaddr_broadcast(
509 			    ((const struct sockaddr_in *)gw)->sin_addr, ia);
510 		else
511 			isbroadcast = false;
512 		mtu = nh->nh_mtu;
513 		src = IA_SIN(ia)->sin_addr;
514 	} else {
515 		struct nhop_object *nh;
516 
517 		nh = fib4_lookup(M_GETFIB(m), dst->sin_addr, 0, NHR_NONE,
518 		    m->m_pkthdr.flowid);
519 		if (nh == NULL) {
520 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
521 			/*
522 			 * There is no route for this packet, but it is
523 			 * possible that a matching SPD entry exists.
524 			 */
525 			no_route_but_check_spd = 1;
526 			goto sendit;
527 #endif
528 			IPSTAT_INC(ips_noroute);
529 			error = EHOSTUNREACH;
530 			goto bad;
531 		}
532 		ifp = nh->nh_ifp;
533 		mtu = nh->nh_mtu;
534 		rt_update_ro_flags(ro, nh);
535 		if (nh->nh_flags & NHF_GATEWAY)
536 			gw = &nh->gw_sa;
537 		ia = ifatoia(nh->nh_ifa);
538 		src = IA_SIN(ia)->sin_addr;
539 		isbroadcast = ((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
540 		    (NHF_HOST | NHF_BROADCAST)) ||
541 		    ((ifp->if_flags & IFF_BROADCAST) &&
542 		    (gw->sa_family == AF_INET) &&
543 		    (in_broadcast(ip->ip_dst) || in_ifaddr_broadcast(
544 		    ((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 		struct ip ip_hdr;
681 
682 		if ((error = IPSEC_OUTPUT(ipv4, ifp, m, inp, mtu)) != 0) {
683 			if (error == EINPROGRESS)
684 				error = 0;
685 			goto done;
686 		}
687 
688 		/* Update variables that are affected by ipsec4_output(). */
689 		m_copydata(m, 0, sizeof(ip_hdr), (char *)&ip_hdr);
690 		hlen = ip_hdr.ip_hl << 2;
691 	}
692 
693 	/*
694 	 * Check if there was a route for this packet; return error if not.
695 	 */
696 	if (no_route_but_check_spd) {
697 		IPSTAT_INC(ips_noroute);
698 		error = EHOSTUNREACH;
699 		goto bad;
700 	}
701 #endif /* IPSEC */
702 
703 	/* Jump over all PFIL processing if hooks are not active. */
704 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
705 		switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
706 		    &error)) {
707 		case 1: /* Finished */
708 			goto done;
709 
710 		case 0: /* Continue normally */
711 			ip = mtod(m, struct ip *);
712 			ip_len = ntohs(ip->ip_len);
713 			break;
714 
715 		case -1: /* Need to try again */
716 			/* Reset everything for a new round */
717 			if (ro != NULL) {
718 				RO_NHFREE(ro);
719 				ro->ro_prepend = NULL;
720 			}
721 			gw = (const struct sockaddr *)dst;
722 			ip = mtod(m, struct ip *);
723 			goto again;
724 		}
725 	}
726 
727 	if (vlan_pcp > -1)
728 		EVL_APPLY_PRI(m, vlan_pcp);
729 
730 	/* IN_LOOPBACK must not appear on the wire - RFC1122. */
731 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
732 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
733 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
734 			IPSTAT_INC(ips_badaddr);
735 			error = EADDRNOTAVAIL;
736 			goto bad;
737 		}
738 	}
739 
740 	/* Ensure the packet data is mapped if the interface requires it. */
741 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
742 		struct mbuf *m1;
743 
744 		error = mb_unmapped_to_ext(m, &m1);
745 		if (error != 0) {
746 			if (error == EINVAL) {
747 				if_printf(ifp, "TLS packet\n");
748 				/* XXXKIB */
749 			} else if (error == ENOMEM) {
750 				error = ENOBUFS;
751 			}
752 			IPSTAT_INC(ips_odropped);
753 			goto done;
754 		} else {
755 			m = m1;
756 		}
757 	}
758 
759 	m->m_pkthdr.csum_flags |= CSUM_IP;
760 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
761 		in_delayed_cksum(m);
762 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
763 	}
764 #if defined(SCTP) || defined(SCTP_SUPPORT)
765 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
766 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
767 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
768 	}
769 #endif
770 
771 	/*
772 	 * If small enough for interface, or the interface will take
773 	 * care of the fragmentation for us, we can just send directly.
774 	 * Note that if_vxlan could have requested TSO even though the outer
775 	 * frame is UDP.  It is correct to not fragment such datagrams and
776 	 * instead just pass them on to the driver.
777 	 */
778 	if (ip_len <= mtu ||
779 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist &
780 	    (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
781 		ip->ip_sum = 0;
782 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
783 			ip->ip_sum = in_cksum(m, hlen);
784 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
785 		}
786 
787 		/*
788 		 * Record statistics for this interface address.
789 		 * With CSUM_TSO the byte/packet count will be slightly
790 		 * incorrect because we count the IP+TCP headers only
791 		 * once instead of for every generated packet.
792 		 */
793 		if (!(flags & IP_FORWARDING) && ia) {
794 			if (m->m_pkthdr.csum_flags &
795 			    (CSUM_TSO | CSUM_INNER_TSO))
796 				counter_u64_add(ia->ia_ifa.ifa_opackets,
797 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
798 			else
799 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
800 
801 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
802 		}
803 #ifdef MBUF_STRESS_TEST
804 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
805 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
806 #endif
807 		/*
808 		 * Reset layer specific mbuf flags
809 		 * to avoid confusing lower layers.
810 		 */
811 		m_clrprotoflags(m);
812 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
813 		error = ip_output_send(inp, ifp, m, gw, ro,
814 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
815 		goto done;
816 	}
817 
818 	/* Balk when DF bit is set or the interface didn't support TSO. */
819 	if ((ip_off & IP_DF) ||
820 	    (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
821 		error = EMSGSIZE;
822 		IPSTAT_INC(ips_cantfrag);
823 		goto bad;
824 	}
825 
826 	/*
827 	 * Too large for interface; fragment if possible. If successful,
828 	 * on return, m will point to a list of packets to be sent.
829 	 */
830 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
831 	if (error)
832 		goto bad;
833 	for (; m; m = m0) {
834 		m0 = m->m_nextpkt;
835 		m->m_nextpkt = 0;
836 		if (error == 0) {
837 			/* Record statistics for this interface address. */
838 			if (ia != NULL) {
839 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
840 				counter_u64_add(ia->ia_ifa.ifa_obytes,
841 				    m->m_pkthdr.len);
842 			}
843 			/*
844 			 * Reset layer specific mbuf flags
845 			 * to avoid confusing upper layers.
846 			 */
847 			m_clrprotoflags(m);
848 
849 			IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
850 			    mtod(m, struct ip *), NULL);
851 			error = ip_output_send(inp, ifp, m, gw, ro, true);
852 		} else
853 			m_freem(m);
854 	}
855 
856 	if (error == 0)
857 		IPSTAT_INC(ips_fragmented);
858 
859 done:
860 	return (error);
861 bad:
862 	m_freem(m);
863 	goto done;
864 }
865 
866 /*
867  * Create a chain of fragments which fit the given mtu. m_frag points to the
868  * mbuf to be fragmented; on return it points to the chain with the fragments.
869  * Return 0 if no error. If error, m_frag may contain a partially built
870  * chain of fragments that should be freed by the caller.
871  *
872  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
873  */
874 int
ip_fragment(struct ip * ip,struct mbuf ** m_frag,int mtu,u_long if_hwassist_flags)875 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
876     u_long if_hwassist_flags)
877 {
878 	int error = 0;
879 	int hlen = ip->ip_hl << 2;
880 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
881 	int off;
882 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
883 	int firstlen;
884 	struct mbuf **mnext;
885 	int nfrags;
886 	uint16_t ip_len, ip_off;
887 
888 	ip_len = ntohs(ip->ip_len);
889 	ip_off = ntohs(ip->ip_off);
890 
891 	/*
892 	 * Packet shall not have "Don't Fragment" flag and have at least 8
893 	 * bytes of payload.
894 	 */
895 	if (__predict_false((ip_off & IP_DF) || len < 8)) {
896 		IPSTAT_INC(ips_cantfrag);
897 		return (EMSGSIZE);
898 	}
899 
900 	/*
901 	 * If the interface will not calculate checksums on
902 	 * fragmented packets, then do it here.
903 	 */
904 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
905 		in_delayed_cksum(m0);
906 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
907 	}
908 #if defined(SCTP) || defined(SCTP_SUPPORT)
909 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
910 		sctp_delayed_cksum(m0, hlen);
911 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
912 	}
913 #endif
914 	if (len > PAGE_SIZE) {
915 		/*
916 		 * Fragment large datagrams such that each segment
917 		 * contains a multiple of PAGE_SIZE amount of data,
918 		 * plus headers. This enables a receiver to perform
919 		 * page-flipping zero-copy optimizations.
920 		 *
921 		 * XXX When does this help given that sender and receiver
922 		 * could have different page sizes, and also mtu could
923 		 * be less than the receiver's page size ?
924 		 */
925 		int newlen;
926 
927 		off = MIN(mtu, m0->m_pkthdr.len);
928 
929 		/*
930 		 * firstlen (off - hlen) must be aligned on an
931 		 * 8-byte boundary
932 		 */
933 		if (off < hlen)
934 			goto smart_frag_failure;
935 		off = ((off - hlen) & ~7) + hlen;
936 		newlen = (~PAGE_MASK) & mtu;
937 		if ((newlen + sizeof (struct ip)) > mtu) {
938 			/* we failed, go back the default */
939 smart_frag_failure:
940 			newlen = len;
941 			off = hlen + len;
942 		}
943 		len = newlen;
944 
945 	} else {
946 		off = hlen + len;
947 	}
948 
949 	firstlen = off - hlen;
950 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
951 
952 	/*
953 	 * Loop through length of segment after first fragment,
954 	 * make new header and copy data of each part and link onto chain.
955 	 * Here, m0 is the original packet, m is the fragment being created.
956 	 * The fragments are linked off the m_nextpkt of the original
957 	 * packet, which after processing serves as the first fragment.
958 	 */
959 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
960 		struct ip *mhip;	/* ip header on the fragment */
961 		struct mbuf *m;
962 		int mhlen = sizeof (struct ip);
963 
964 		m = m_gethdr(M_NOWAIT, MT_DATA);
965 		if (m == NULL) {
966 			error = ENOBUFS;
967 			IPSTAT_INC(ips_odropped);
968 			goto done;
969 		}
970 		/*
971 		 * Make sure the complete packet header gets copied
972 		 * from the originating mbuf to the newly created
973 		 * mbuf. This also ensures that existing firewall
974 		 * classification(s), VLAN tags and so on get copied
975 		 * to the resulting fragmented packet(s):
976 		 */
977 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
978 			m_free(m);
979 			error = ENOBUFS;
980 			IPSTAT_INC(ips_odropped);
981 			goto done;
982 		}
983 		/*
984 		 * In the first mbuf, leave room for the link header, then
985 		 * copy the original IP header including options. The payload
986 		 * goes into an additional mbuf chain returned by m_copym().
987 		 */
988 		m->m_data += max_linkhdr;
989 		mhip = mtod(m, struct ip *);
990 		*mhip = *ip;
991 		if (hlen > sizeof (struct ip)) {
992 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
993 			mhip->ip_v = IPVERSION;
994 			mhip->ip_hl = mhlen >> 2;
995 		}
996 		m->m_len = mhlen;
997 		/* XXX do we need to add ip_off below ? */
998 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
999 		if (off + len >= ip_len)
1000 			len = ip_len - off;
1001 		else
1002 			mhip->ip_off |= IP_MF;
1003 		mhip->ip_len = htons((u_short)(len + mhlen));
1004 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
1005 		if (m->m_next == NULL) {	/* copy failed */
1006 			m_free(m);
1007 			error = ENOBUFS;	/* ??? */
1008 			IPSTAT_INC(ips_odropped);
1009 			goto done;
1010 		}
1011 		m->m_pkthdr.len = mhlen + len;
1012 #ifdef MAC
1013 		mac_netinet_fragment(m0, m);
1014 #endif
1015 		mhip->ip_off = htons(mhip->ip_off);
1016 		mhip->ip_sum = 0;
1017 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1018 			mhip->ip_sum = in_cksum(m, mhlen);
1019 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
1020 		}
1021 		*mnext = m;
1022 		mnext = &m->m_nextpkt;
1023 	}
1024 	IPSTAT_ADD(ips_ofragments, nfrags);
1025 
1026 	/*
1027 	 * Update first fragment by trimming what's been copied out
1028 	 * and updating header.
1029 	 */
1030 	m_adj(m0, hlen + firstlen - ip_len);
1031 	m0->m_pkthdr.len = hlen + firstlen;
1032 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1033 	ip->ip_off = htons(ip_off | IP_MF);
1034 	ip->ip_sum = 0;
1035 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1036 		ip->ip_sum = in_cksum(m0, hlen);
1037 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1038 	}
1039 
1040 done:
1041 	*m_frag = m0;
1042 	return error;
1043 }
1044 
1045 void
in_delayed_cksum_o(struct mbuf * m,uint16_t iph_offset)1046 in_delayed_cksum_o(struct mbuf *m, uint16_t iph_offset)
1047 {
1048 	struct ip *ip;
1049 	struct udphdr *uh;
1050 	uint16_t cklen, csum, offset;
1051 
1052 	ip = (struct ip *)mtodo(m, iph_offset);
1053 	offset = iph_offset + (ip->ip_hl << 2);
1054 
1055 	if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1056 		/* if udp header is not in the first mbuf copy udplen */
1057 		if (offset + sizeof(struct udphdr) > m->m_len) {
1058 			m_copydata(m, offset + offsetof(struct udphdr,
1059 			    uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1060 			cklen = ntohs(cklen);
1061 		} else {
1062 			uh = (struct udphdr *)mtodo(m, offset);
1063 			cklen = ntohs(uh->uh_ulen);
1064 		}
1065 		csum = in_cksum_skip(m, cklen + offset, offset);
1066 		if (csum == 0)
1067 			csum = 0xffff;
1068 	} else {
1069 		cklen = ntohs(ip->ip_len);
1070 		csum = in_cksum_skip(m, cklen, offset);
1071 	}
1072 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1073 
1074 	if (offset + sizeof(csum) > m->m_len)
1075 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1076 	else
1077 		*(u_short *)mtodo(m, offset) = csum;
1078 }
1079 
1080 void
in_delayed_cksum(struct mbuf * m)1081 in_delayed_cksum(struct mbuf *m)
1082 {
1083 
1084 	in_delayed_cksum_o(m, 0);
1085 }
1086 
1087 /*
1088  * IP socket option processing.
1089  */
1090 int
ip_ctloutput(struct socket * so,struct sockopt * sopt)1091 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1092 {
1093 	struct inpcb *inp = sotoinpcb(so);
1094 	int	error, optval;
1095 #ifdef	RSS
1096 	uint32_t rss_bucket;
1097 	int retval;
1098 #endif
1099 
1100 	error = optval = 0;
1101 	if (sopt->sopt_level != IPPROTO_IP) {
1102 		error = EINVAL;
1103 
1104 		if (sopt->sopt_level == SOL_SOCKET &&
1105 		    sopt->sopt_dir == SOPT_SET) {
1106 			switch (sopt->sopt_name) {
1107 			case SO_SETFIB:
1108 				error = sooptcopyin(sopt, &optval,
1109 				    sizeof(optval), sizeof(optval));
1110 				if (error != 0)
1111 					break;
1112 
1113 				INP_WLOCK(inp);
1114 				if ((inp->inp_flags & INP_BOUNDFIB) != 0 &&
1115 				    optval != so->so_fibnum) {
1116 					INP_WUNLOCK(inp);
1117 					error = EISCONN;
1118 					break;
1119 				}
1120 				error = sosetfib(inp->inp_socket, optval);
1121 				if (error == 0)
1122 					inp->inp_inc.inc_fibnum = optval;
1123 				INP_WUNLOCK(inp);
1124 				break;
1125 			case SO_MAX_PACING_RATE:
1126 #ifdef RATELIMIT
1127 				INP_WLOCK(inp);
1128 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1129 				INP_WUNLOCK(inp);
1130 				error = 0;
1131 #else
1132 				error = EOPNOTSUPP;
1133 #endif
1134 				break;
1135 			default:
1136 				break;
1137 			}
1138 		}
1139 		return (error);
1140 	}
1141 
1142 	switch (sopt->sopt_dir) {
1143 	case SOPT_SET:
1144 		switch (sopt->sopt_name) {
1145 		case IP_OPTIONS:
1146 #ifdef notyet
1147 		case IP_RETOPTS:
1148 #endif
1149 		{
1150 			struct mbuf *m;
1151 			if (sopt->sopt_valsize > MLEN) {
1152 				error = EMSGSIZE;
1153 				break;
1154 			}
1155 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1156 			if (m == NULL) {
1157 				error = ENOBUFS;
1158 				break;
1159 			}
1160 			m->m_len = sopt->sopt_valsize;
1161 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1162 					    m->m_len);
1163 			if (error) {
1164 				m_free(m);
1165 				break;
1166 			}
1167 			INP_WLOCK(inp);
1168 			error = ip_pcbopts(inp, sopt->sopt_name, m);
1169 			INP_WUNLOCK(inp);
1170 			return (error);
1171 		}
1172 
1173 		case IP_BINDANY:
1174 			if (sopt->sopt_td != NULL) {
1175 				error = priv_check(sopt->sopt_td,
1176 				    PRIV_NETINET_BINDANY);
1177 				if (error)
1178 					break;
1179 			}
1180 			/* FALLTHROUGH */
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_RECVFLOWID:
1274 				OPTSET2(INP_RECVFLOWID, optval);
1275 				break;
1276 #ifdef RSS
1277 			case IP_RECVRSSBUCKETID:
1278 				OPTSET2(INP_RECVRSSBUCKETID, optval);
1279 				break;
1280 #endif
1281 			case IP_VLAN_PCP:
1282 				if ((optval >= -1) && (optval <=
1283 				    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1284 					if (optval == -1) {
1285 						INP_WLOCK(inp);
1286 						inp->inp_flags2 &=
1287 						    ~(INP_2PCP_SET |
1288 						      INP_2PCP_MASK);
1289 						INP_WUNLOCK(inp);
1290 					} else {
1291 						INP_WLOCK(inp);
1292 						inp->inp_flags2 |=
1293 						    INP_2PCP_SET;
1294 						inp->inp_flags2 &=
1295 						    ~INP_2PCP_MASK;
1296 						inp->inp_flags2 |=
1297 						    optval << INP_2PCP_SHIFT;
1298 						INP_WUNLOCK(inp);
1299 					}
1300 				} else
1301 					error = EINVAL;
1302 				break;
1303 			}
1304 			break;
1305 #undef OPTSET
1306 #undef OPTSET2
1307 
1308 		/*
1309 		 * Multicast socket options are processed by the in_mcast
1310 		 * module.
1311 		 */
1312 		case IP_MULTICAST_IF:
1313 		case IP_MULTICAST_VIF:
1314 		case IP_MULTICAST_TTL:
1315 		case IP_MULTICAST_LOOP:
1316 		case IP_ADD_MEMBERSHIP:
1317 		case IP_DROP_MEMBERSHIP:
1318 		case IP_ADD_SOURCE_MEMBERSHIP:
1319 		case IP_DROP_SOURCE_MEMBERSHIP:
1320 		case IP_BLOCK_SOURCE:
1321 		case IP_UNBLOCK_SOURCE:
1322 		case IP_MSFILTER:
1323 		case MCAST_JOIN_GROUP:
1324 		case MCAST_LEAVE_GROUP:
1325 		case MCAST_JOIN_SOURCE_GROUP:
1326 		case MCAST_LEAVE_SOURCE_GROUP:
1327 		case MCAST_BLOCK_SOURCE:
1328 		case MCAST_UNBLOCK_SOURCE:
1329 			error = inp_setmoptions(inp, sopt);
1330 			break;
1331 
1332 		case IP_PORTRANGE:
1333 			error = sooptcopyin(sopt, &optval, sizeof optval,
1334 					    sizeof optval);
1335 			if (error)
1336 				break;
1337 
1338 			INP_WLOCK(inp);
1339 			switch (optval) {
1340 			case IP_PORTRANGE_DEFAULT:
1341 				inp->inp_flags &= ~(INP_LOWPORT);
1342 				inp->inp_flags &= ~(INP_HIGHPORT);
1343 				break;
1344 
1345 			case IP_PORTRANGE_HIGH:
1346 				inp->inp_flags &= ~(INP_LOWPORT);
1347 				inp->inp_flags |= INP_HIGHPORT;
1348 				break;
1349 
1350 			case IP_PORTRANGE_LOW:
1351 				inp->inp_flags &= ~(INP_HIGHPORT);
1352 				inp->inp_flags |= INP_LOWPORT;
1353 				break;
1354 
1355 			default:
1356 				error = EINVAL;
1357 				break;
1358 			}
1359 			INP_WUNLOCK(inp);
1360 			break;
1361 
1362 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1363 		case IP_IPSEC_POLICY:
1364 			if (IPSEC_ENABLED(ipv4)) {
1365 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1366 				break;
1367 			}
1368 			/* FALLTHROUGH */
1369 #endif /* IPSEC */
1370 
1371 		default:
1372 			error = ENOPROTOOPT;
1373 			break;
1374 		}
1375 		break;
1376 
1377 	case SOPT_GET:
1378 		switch (sopt->sopt_name) {
1379 		case IP_OPTIONS:
1380 		case IP_RETOPTS:
1381 			INP_RLOCK(inp);
1382 			if (inp->inp_options) {
1383 				struct mbuf *options;
1384 
1385 				options = m_copym(inp->inp_options, 0,
1386 				    M_COPYALL, M_NOWAIT);
1387 				INP_RUNLOCK(inp);
1388 				if (options != NULL) {
1389 					error = sooptcopyout(sopt,
1390 							     mtod(options, char *),
1391 							     options->m_len);
1392 					m_freem(options);
1393 				} else
1394 					error = ENOMEM;
1395 			} else {
1396 				INP_RUNLOCK(inp);
1397 				sopt->sopt_valsize = 0;
1398 			}
1399 			break;
1400 
1401 		case IP_TOS:
1402 		case IP_TTL:
1403 		case IP_MINTTL:
1404 		case IP_RECVOPTS:
1405 		case IP_RECVRETOPTS:
1406 		case IP_ORIGDSTADDR:
1407 		case IP_RECVDSTADDR:
1408 		case IP_RECVTTL:
1409 		case IP_RECVIF:
1410 		case IP_PORTRANGE:
1411 		case IP_ONESBCAST:
1412 		case IP_DONTFRAG:
1413 		case IP_BINDANY:
1414 		case IP_RECVTOS:
1415 		case IP_FLOWID:
1416 		case IP_FLOWTYPE:
1417 		case IP_RECVFLOWID:
1418 #ifdef	RSS
1419 		case IP_RSSBUCKETID:
1420 		case IP_RECVRSSBUCKETID:
1421 #endif
1422 		case IP_VLAN_PCP:
1423 			switch (sopt->sopt_name) {
1424 			case IP_TOS:
1425 				optval = inp->inp_ip_tos;
1426 				break;
1427 
1428 			case IP_TTL:
1429 				optval = inp->inp_ip_ttl;
1430 				break;
1431 
1432 			case IP_MINTTL:
1433 				optval = inp->inp_ip_minttl;
1434 				break;
1435 
1436 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1437 #define	OPTBIT2(bit)	(inp->inp_flags2 & bit ? 1 : 0)
1438 
1439 			case IP_RECVOPTS:
1440 				optval = OPTBIT(INP_RECVOPTS);
1441 				break;
1442 
1443 			case IP_RECVRETOPTS:
1444 				optval = OPTBIT(INP_RECVRETOPTS);
1445 				break;
1446 
1447 			case IP_RECVDSTADDR:
1448 				optval = OPTBIT(INP_RECVDSTADDR);
1449 				break;
1450 
1451 			case IP_ORIGDSTADDR:
1452 				optval = OPTBIT2(INP_ORIGDSTADDR);
1453 				break;
1454 
1455 			case IP_RECVTTL:
1456 				optval = OPTBIT(INP_RECVTTL);
1457 				break;
1458 
1459 			case IP_RECVIF:
1460 				optval = OPTBIT(INP_RECVIF);
1461 				break;
1462 
1463 			case IP_PORTRANGE:
1464 				if (inp->inp_flags & INP_HIGHPORT)
1465 					optval = IP_PORTRANGE_HIGH;
1466 				else if (inp->inp_flags & INP_LOWPORT)
1467 					optval = IP_PORTRANGE_LOW;
1468 				else
1469 					optval = 0;
1470 				break;
1471 
1472 			case IP_ONESBCAST:
1473 				optval = OPTBIT(INP_ONESBCAST);
1474 				break;
1475 			case IP_DONTFRAG:
1476 				optval = OPTBIT(INP_DONTFRAG);
1477 				break;
1478 			case IP_BINDANY:
1479 				optval = OPTBIT(INP_BINDANY);
1480 				break;
1481 			case IP_RECVTOS:
1482 				optval = OPTBIT(INP_RECVTOS);
1483 				break;
1484 			case IP_FLOWID:
1485 				optval = inp->inp_flowid;
1486 				break;
1487 			case IP_FLOWTYPE:
1488 				optval = inp->inp_flowtype;
1489 				break;
1490 			case IP_RECVFLOWID:
1491 				optval = OPTBIT2(INP_RECVFLOWID);
1492 				break;
1493 #ifdef	RSS
1494 			case IP_RSSBUCKETID:
1495 				retval = rss_hash2bucket(inp->inp_flowid,
1496 				    inp->inp_flowtype,
1497 				    &rss_bucket);
1498 				if (retval == 0)
1499 					optval = rss_bucket;
1500 				else
1501 					error = EINVAL;
1502 				break;
1503 			case IP_RECVRSSBUCKETID:
1504 				optval = OPTBIT2(INP_RECVRSSBUCKETID);
1505 				break;
1506 #endif
1507 			case IP_VLAN_PCP:
1508 				if (OPTBIT2(INP_2PCP_SET)) {
1509 					optval = (inp->inp_flags2 &
1510 					    INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1511 				} else {
1512 					optval = -1;
1513 				}
1514 				break;
1515 			}
1516 			error = sooptcopyout(sopt, &optval, sizeof optval);
1517 			break;
1518 
1519 		/*
1520 		 * Multicast socket options are processed by the in_mcast
1521 		 * module.
1522 		 */
1523 		case IP_MULTICAST_IF:
1524 		case IP_MULTICAST_VIF:
1525 		case IP_MULTICAST_TTL:
1526 		case IP_MULTICAST_LOOP:
1527 		case IP_MSFILTER:
1528 			error = inp_getmoptions(inp, sopt);
1529 			break;
1530 
1531 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1532 		case IP_IPSEC_POLICY:
1533 			if (IPSEC_ENABLED(ipv4)) {
1534 				error = IPSEC_PCBCTL(ipv4, inp, sopt);
1535 				break;
1536 			}
1537 			/* FALLTHROUGH */
1538 #endif /* IPSEC */
1539 
1540 		default:
1541 			error = ENOPROTOOPT;
1542 			break;
1543 		}
1544 		break;
1545 	}
1546 	return (error);
1547 }
1548 
1549 /*
1550  * Routine called from ip_output() to loop back a copy of an IP multicast
1551  * packet to the input queue of a specified interface.  Note that this
1552  * calls the output routine of the loopback "driver", but with an interface
1553  * pointer that might NOT be a loopback interface -- evil, but easier than
1554  * replicating that code here.
1555  */
1556 static void
ip_mloopback(struct ifnet * ifp,const struct mbuf * m,int hlen)1557 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1558 {
1559 	struct ip *ip;
1560 	struct mbuf *copym;
1561 
1562 	/*
1563 	 * Make a deep copy of the packet because we're going to
1564 	 * modify the pack in order to generate checksums.
1565 	 */
1566 	copym = m_dup(m, M_NOWAIT);
1567 	if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1568 		copym = m_pullup(copym, hlen);
1569 	if (copym != NULL) {
1570 		/* If needed, compute the checksum and mark it as valid. */
1571 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1572 			in_delayed_cksum(copym);
1573 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1574 			copym->m_pkthdr.csum_flags |=
1575 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1576 			copym->m_pkthdr.csum_data = 0xffff;
1577 		}
1578 		/*
1579 		 * We don't bother to fragment if the IP length is greater
1580 		 * than the interface's MTU.  Can this possibly matter?
1581 		 */
1582 		ip = mtod(copym, struct ip *);
1583 		ip->ip_sum = 0;
1584 		ip->ip_sum = in_cksum(copym, hlen);
1585 		if_simloop(ifp, copym, AF_INET, 0);
1586 	}
1587 }
1588