xref: /freebsd/sys/netinet/raw_ip.c (revision 3078531de10dcae44b253a35125c949ff4235284)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1993
5  *	The Regents of the University of California.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)raw_ip.c	8.7 (Berkeley) 5/15/95
33  */
34 
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37 
38 #include "opt_inet.h"
39 #include "opt_inet6.h"
40 #include "opt_ipsec.h"
41 #include "opt_route.h"
42 
43 #include <sys/param.h>
44 #include <sys/jail.h>
45 #include <sys/kernel.h>
46 #include <sys/eventhandler.h>
47 #include <sys/lock.h>
48 #include <sys/malloc.h>
49 #include <sys/mbuf.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/protosw.h>
53 #include <sys/rwlock.h>
54 #include <sys/signalvar.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sx.h>
58 #include <sys/sysctl.h>
59 #include <sys/systm.h>
60 
61 #include <vm/uma.h>
62 
63 #include <net/if.h>
64 #include <net/if_var.h>
65 #include <net/route.h>
66 #include <net/route/route_ctl.h>
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/in_fib.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/in_var.h>
74 #include <netinet/if_ether.h>
75 #include <netinet/ip.h>
76 #include <netinet/ip_var.h>
77 #include <netinet/ip_mroute.h>
78 #include <netinet/ip_icmp.h>
79 
80 #include <netipsec/ipsec_support.h>
81 
82 #include <machine/stdarg.h>
83 #include <security/mac/mac_framework.h>
84 
85 extern ipproto_input_t *ip_protox[];
86 
87 VNET_DEFINE(int, ip_defttl) = IPDEFTTL;
88 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_VNET | CTLFLAG_RW,
89     &VNET_NAME(ip_defttl), 0,
90     "Maximum TTL on IP packets");
91 
92 VNET_DEFINE(struct inpcbinfo, ripcbinfo);
93 #define	V_ripcbinfo		VNET(ripcbinfo)
94 
95 /*
96  * Control and data hooks for ipfw, dummynet, divert and so on.
97  * The data hooks are not used here but it is convenient
98  * to keep them all in one place.
99  */
100 VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL;
101 VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL;
102 
103 int	(*ip_dn_ctl_ptr)(struct sockopt *);
104 int	(*ip_dn_io_ptr)(struct mbuf **, struct ip_fw_args *);
105 void	(*ip_divert_ptr)(struct mbuf *, bool);
106 int	(*ng_ipfw_input_p)(struct mbuf **, struct ip_fw_args *, bool);
107 
108 #ifdef INET
109 /*
110  * Hooks for multicast routing. They all default to NULL, so leave them not
111  * initialized and rely on BSS being set to 0.
112  */
113 
114 /*
115  * The socket used to communicate with the multicast routing daemon.
116  */
117 VNET_DEFINE(struct socket *, ip_mrouter);
118 
119 /*
120  * The various mrouter and rsvp functions.
121  */
122 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
123 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
124 int (*ip_mrouter_done)(void);
125 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
126 		   struct ip_moptions *);
127 int (*mrt_ioctl)(u_long, caddr_t, int);
128 int (*legal_vif_num)(int);
129 u_long (*ip_mcast_src)(int);
130 
131 int (*rsvp_input_p)(struct mbuf **, int *, int);
132 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
133 void (*ip_rsvp_force_done)(struct socket *);
134 #endif /* INET */
135 
136 u_long	rip_sendspace = 9216;
137 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
138     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
139 
140 u_long	rip_recvspace = 9216;
141 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
142     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
143 
144 /*
145  * Hash functions
146  */
147 
148 #define INP_PCBHASH_RAW_SIZE	256
149 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \
150         (((proto) + (laddr) + (faddr)) % (mask) + 1)
151 
152 #ifdef INET
153 static void
154 rip_inshash(struct inpcb *inp)
155 {
156 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
157 	struct inpcbhead *pcbhash;
158 	int hash;
159 
160 	INP_HASH_WLOCK_ASSERT(pcbinfo);
161 	INP_WLOCK_ASSERT(inp);
162 
163 	if (inp->inp_ip_p != 0 &&
164 	    inp->inp_laddr.s_addr != INADDR_ANY &&
165 	    inp->inp_faddr.s_addr != INADDR_ANY) {
166 		hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr,
167 		    inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask);
168 	} else
169 		hash = 0;
170 	pcbhash = &pcbinfo->ipi_hashbase[hash];
171 	CK_LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
172 }
173 
174 static void
175 rip_delhash(struct inpcb *inp)
176 {
177 
178 	INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
179 	INP_WLOCK_ASSERT(inp);
180 
181 	CK_LIST_REMOVE(inp, inp_hash);
182 }
183 #endif /* INET */
184 
185 INPCBSTORAGE_DEFINE(ripcbstor, "rawinp", "ripcb", "rip", "riphash");
186 
187 static void
188 rip_init(void *arg __unused)
189 {
190 
191 	in_pcbinfo_init(&V_ripcbinfo, &ripcbstor, INP_PCBHASH_RAW_SIZE, 1);
192 }
193 VNET_SYSINIT(rip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rip_init, NULL);
194 
195 #ifdef VIMAGE
196 static void
197 rip_destroy(void *unused __unused)
198 {
199 
200 	in_pcbinfo_destroy(&V_ripcbinfo);
201 }
202 VNET_SYSUNINIT(raw_ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, rip_destroy, NULL);
203 #endif
204 
205 #ifdef INET
206 static int
207 rip_append(struct inpcb *inp, struct ip *ip, struct mbuf *m,
208     struct sockaddr_in *ripsrc)
209 {
210 	struct socket *so = inp->inp_socket;
211 	struct mbuf *n, *opts = NULL;
212 
213 	INP_LOCK_ASSERT(inp);
214 
215 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
216 	/* check AH/ESP integrity. */
217 	if (IPSEC_ENABLED(ipv4) && IPSEC_CHECK_POLICY(ipv4, m, inp) != 0)
218 		return (0);
219 #endif /* IPSEC */
220 #ifdef MAC
221 	if (mac_inpcb_check_deliver(inp, m) != 0)
222 		return (0);
223 #endif
224 	/* Check the minimum TTL for socket. */
225 	if (inp->inp_ip_minttl && inp->inp_ip_minttl > ip->ip_ttl)
226 		return (0);
227 
228 	if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL)
229 		return (0);
230 
231 	if ((inp->inp_flags & INP_CONTROLOPTS) ||
232 	    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
233 		ip_savecontrol(inp, &opts, ip, n);
234 	SOCKBUF_LOCK(&so->so_rcv);
235 	if (sbappendaddr_locked(&so->so_rcv,
236 	    (struct sockaddr *)ripsrc, n, opts) == 0) {
237 		soroverflow_locked(so);
238 		m_freem(n);
239 		if (opts)
240 			m_freem(opts);
241 		return (0);
242 	}
243 	sorwakeup_locked(so);
244 
245 	return (1);
246 }
247 
248 struct rip_inp_match_ctx {
249 	struct ip *ip;
250 	int proto;
251 };
252 
253 static bool
254 rip_inp_match1(const struct inpcb *inp, void *v)
255 {
256 	struct rip_inp_match_ctx *ctx = v;
257 
258 	if (inp->inp_ip_p != ctx->proto)
259 		return (false);
260 #ifdef INET6
261 	/* XXX inp locking */
262 	if ((inp->inp_vflag & INP_IPV4) == 0)
263 		return (false);
264 #endif
265 	if (inp->inp_laddr.s_addr != ctx->ip->ip_dst.s_addr)
266 		return (false);
267 	if (inp->inp_faddr.s_addr != ctx->ip->ip_src.s_addr)
268 		return (false);
269 	return (true);
270 }
271 
272 static bool
273 rip_inp_match2(const struct inpcb *inp, void *v)
274 {
275 	struct rip_inp_match_ctx *ctx = v;
276 
277 	if (inp->inp_ip_p && inp->inp_ip_p != ctx->proto)
278 		return (false);
279 #ifdef INET6
280 	/* XXX inp locking */
281 	if ((inp->inp_vflag & INP_IPV4) == 0)
282 		return (false);
283 #endif
284 	if (!in_nullhost(inp->inp_laddr) &&
285 	    !in_hosteq(inp->inp_laddr, ctx->ip->ip_dst))
286 		return (false);
287 	if (!in_nullhost(inp->inp_faddr) &&
288 	    !in_hosteq(inp->inp_faddr, ctx->ip->ip_src))
289 		return (false);
290 	return (true);
291 }
292 
293 /*
294  * Setup generic address and protocol structures for raw_input routine, then
295  * pass them along with mbuf chain.
296  */
297 int
298 rip_input(struct mbuf **mp, int *offp, int proto)
299 {
300 	struct rip_inp_match_ctx ctx = {
301 		.ip = mtod(*mp, struct ip *),
302 		.proto = proto,
303 	};
304 	struct inpcb_iterator inpi = INP_ITERATOR(&V_ripcbinfo,
305 	    INPLOOKUP_RLOCKPCB, rip_inp_match1, &ctx);
306 	struct ifnet *ifp;
307 	struct mbuf *m = *mp;
308 	struct inpcb *inp;
309 	struct sockaddr_in ripsrc;
310 	int appended;
311 
312 	*mp = NULL;
313 	appended = 0;
314 
315 	bzero(&ripsrc, sizeof(ripsrc));
316 	ripsrc.sin_len = sizeof(ripsrc);
317 	ripsrc.sin_family = AF_INET;
318 	ripsrc.sin_addr = ctx.ip->ip_src;
319 
320 	ifp = m->m_pkthdr.rcvif;
321 
322 	inpi.hash = INP_PCBHASH_RAW(proto, ctx.ip->ip_src.s_addr,
323 	    ctx.ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
324 	while ((inp = inp_next(&inpi)) != NULL) {
325 		INP_RLOCK_ASSERT(inp);
326 		if (jailed_without_vnet(inp->inp_cred) &&
327 		    prison_check_ip4(inp->inp_cred, &ctx.ip->ip_dst) != 0) {
328 			/*
329 			 * XXX: If faddr was bound to multicast group,
330 			 * jailed raw socket will drop datagram.
331 			 */
332 			continue;
333 		}
334 		appended += rip_append(inp, ctx.ip, m, &ripsrc);
335 	}
336 
337 	inpi.hash = 0;
338 	inpi.match = rip_inp_match2;
339 	MPASS(inpi.inp == NULL);
340 	while ((inp = inp_next(&inpi)) != NULL) {
341 		INP_RLOCK_ASSERT(inp);
342 		if (jailed_without_vnet(inp->inp_cred) &&
343 		    !IN_MULTICAST(ntohl(ctx.ip->ip_dst.s_addr)) &&
344 		    prison_check_ip4(inp->inp_cred, &ctx.ip->ip_dst) != 0)
345 			/*
346 			 * Allow raw socket in jail to receive multicast;
347 			 * assume process had PRIV_NETINET_RAW at attach,
348 			 * and fall through into normal filter path if so.
349 			 */
350 			continue;
351 		/*
352 		 * If this raw socket has multicast state, and we
353 		 * have received a multicast, check if this socket
354 		 * should receive it, as multicast filtering is now
355 		 * the responsibility of the transport layer.
356 		 */
357 		if (inp->inp_moptions != NULL &&
358 		    IN_MULTICAST(ntohl(ctx.ip->ip_dst.s_addr))) {
359 			/*
360 			 * If the incoming datagram is for IGMP, allow it
361 			 * through unconditionally to the raw socket.
362 			 *
363 			 * In the case of IGMPv2, we may not have explicitly
364 			 * joined the group, and may have set IFF_ALLMULTI
365 			 * on the interface. imo_multi_filter() may discard
366 			 * control traffic we actually need to see.
367 			 *
368 			 * Userland multicast routing daemons should continue
369 			 * filter the control traffic appropriately.
370 			 */
371 			int blocked;
372 
373 			blocked = MCAST_PASS;
374 			if (proto != IPPROTO_IGMP) {
375 				struct sockaddr_in group;
376 
377 				bzero(&group, sizeof(struct sockaddr_in));
378 				group.sin_len = sizeof(struct sockaddr_in);
379 				group.sin_family = AF_INET;
380 				group.sin_addr = ctx.ip->ip_dst;
381 
382 				blocked = imo_multi_filter(inp->inp_moptions,
383 				    ifp,
384 				    (struct sockaddr *)&group,
385 				    (struct sockaddr *)&ripsrc);
386 			}
387 
388 			if (blocked != MCAST_PASS) {
389 				IPSTAT_INC(ips_notmember);
390 				continue;
391 			}
392 		}
393 		appended += rip_append(inp, ctx.ip, m, &ripsrc);
394 	}
395 	if (appended == 0 && ip_protox[ctx.ip->ip_p] == rip_input) {
396 		IPSTAT_INC(ips_noproto);
397 		IPSTAT_DEC(ips_delivered);
398 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PROTOCOL, 0, 0);
399 	} else
400 		m_freem(m);
401 	return (IPPROTO_DONE);
402 }
403 
404 /*
405  * Generate IP header and pass packet to ip_output.  Tack on options user may
406  * have setup with control call.
407  */
408 static int
409 rip_send(struct socket *so, int pruflags, struct mbuf *m, struct sockaddr *nam,
410     struct mbuf *control, struct thread *td)
411 {
412 	struct epoch_tracker et;
413 	struct ip *ip;
414 	struct inpcb *inp;
415 	in_addr_t *dst;
416 	int error, flags, cnt, hlen;
417 	u_char opttype, optlen, *cp;
418 
419 	inp = sotoinpcb(so);
420 	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
421 
422 	if (control != NULL) {
423 		m_freem(control);
424 		control = NULL;
425 	}
426 
427 	if (so->so_state & SS_ISCONNECTED) {
428 		if (nam) {
429 			error = EISCONN;
430 			m_freem(m);
431 			return (error);
432 		}
433 		dst = &inp->inp_faddr.s_addr;
434 	} else {
435 		if (nam == NULL)
436 			error = ENOTCONN;
437 		else if (nam->sa_family != AF_INET)
438 			error = EAFNOSUPPORT;
439 		else if (nam->sa_len != sizeof(struct sockaddr_in))
440 			error = EINVAL;
441 		else
442 			error = 0;
443 		if (error != 0) {
444 			m_freem(m);
445 			return (error);
446 		}
447 		dst = &((struct sockaddr_in *)nam)->sin_addr.s_addr;
448 	}
449 
450 	flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
451 	    IP_ALLOWBROADCAST;
452 
453 	/*
454 	 * If the user handed us a complete IP packet, use it.  Otherwise,
455 	 * allocate an mbuf for a header and fill it in.
456 	 */
457 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
458 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
459 			m_freem(m);
460 			return(EMSGSIZE);
461 		}
462 		M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
463 		if (m == NULL)
464 			return(ENOBUFS);
465 
466 		INP_RLOCK(inp);
467 		ip = mtod(m, struct ip *);
468 		ip->ip_tos = inp->inp_ip_tos;
469 		if (inp->inp_flags & INP_DONTFRAG)
470 			ip->ip_off = htons(IP_DF);
471 		else
472 			ip->ip_off = htons(0);
473 		ip->ip_p = inp->inp_ip_p;
474 		ip->ip_len = htons(m->m_pkthdr.len);
475 		ip->ip_src = inp->inp_laddr;
476 		ip->ip_dst.s_addr = *dst;
477 #ifdef ROUTE_MPATH
478 		if (CALC_FLOWID_OUTBOUND) {
479 			uint32_t hash_type, hash_val;
480 
481 			hash_val = fib4_calc_software_hash(ip->ip_src,
482 			    ip->ip_dst, 0, 0, ip->ip_p, &hash_type);
483 			m->m_pkthdr.flowid = hash_val;
484 			M_HASHTYPE_SET(m, hash_type);
485 			flags |= IP_NODEFAULTFLOWID;
486 		}
487 #endif
488 		if (jailed(inp->inp_cred)) {
489 			/*
490 			 * prison_local_ip4() would be good enough but would
491 			 * let a source of INADDR_ANY pass, which we do not
492 			 * want to see from jails.
493 			 */
494 			if (ip->ip_src.s_addr == INADDR_ANY) {
495 				NET_EPOCH_ENTER(et);
496 				error = in_pcbladdr(inp, &ip->ip_dst,
497 				    &ip->ip_src, inp->inp_cred);
498 				NET_EPOCH_EXIT(et);
499 			} else {
500 				error = prison_local_ip4(inp->inp_cred,
501 				    &ip->ip_src);
502 			}
503 			if (error != 0) {
504 				INP_RUNLOCK(inp);
505 				m_freem(m);
506 				return (error);
507 			}
508 		}
509 		ip->ip_ttl = inp->inp_ip_ttl;
510 	} else {
511 		if (m->m_pkthdr.len > IP_MAXPACKET) {
512 			m_freem(m);
513 			return (EMSGSIZE);
514 		}
515 		if (m->m_pkthdr.len < sizeof(*ip)) {
516 			m_freem(m);
517 			return (EINVAL);
518 		}
519 		m = m_pullup(m, sizeof(*ip));
520 		if (m == NULL)
521 			return (ENOMEM);
522 		ip = mtod(m, struct ip *);
523 		hlen = ip->ip_hl << 2;
524 		if (m->m_len < hlen) {
525 			m = m_pullup(m, hlen);
526 			if (m == NULL)
527 				return (EINVAL);
528 			ip = mtod(m, struct ip *);
529 		}
530 #ifdef ROUTE_MPATH
531 		if (CALC_FLOWID_OUTBOUND) {
532 			uint32_t hash_type, hash_val;
533 
534 			hash_val = fib4_calc_software_hash(ip->ip_dst,
535 			    ip->ip_src, 0, 0, ip->ip_p, &hash_type);
536 			m->m_pkthdr.flowid = hash_val;
537 			M_HASHTYPE_SET(m, hash_type);
538 			flags |= IP_NODEFAULTFLOWID;
539 		}
540 #endif
541 		INP_RLOCK(inp);
542 		/*
543 		 * Don't allow both user specified and setsockopt options,
544 		 * and don't allow packet length sizes that will crash.
545 		 */
546 		if ((hlen < sizeof (*ip))
547 		    || ((hlen > sizeof (*ip)) && inp->inp_options)
548 		    || (ntohs(ip->ip_len) != m->m_pkthdr.len)) {
549 			INP_RUNLOCK(inp);
550 			m_freem(m);
551 			return (EINVAL);
552 		}
553 		error = prison_check_ip4(inp->inp_cred, &ip->ip_src);
554 		if (error != 0) {
555 			INP_RUNLOCK(inp);
556 			m_freem(m);
557 			return (error);
558 		}
559 		/*
560 		 * Don't allow IP options which do not have the required
561 		 * structure as specified in section 3.1 of RFC 791 on
562 		 * pages 15-23.
563 		 */
564 		cp = (u_char *)(ip + 1);
565 		cnt = hlen - sizeof (struct ip);
566 		for (; cnt > 0; cnt -= optlen, cp += optlen) {
567 			opttype = cp[IPOPT_OPTVAL];
568 			if (opttype == IPOPT_EOL)
569 				break;
570 			if (opttype == IPOPT_NOP) {
571 				optlen = 1;
572 				continue;
573 			}
574 			if (cnt < IPOPT_OLEN + sizeof(u_char)) {
575 				INP_RUNLOCK(inp);
576 				m_freem(m);
577 				return (EINVAL);
578 			}
579 			optlen = cp[IPOPT_OLEN];
580 			if (optlen < IPOPT_OLEN + sizeof(u_char) ||
581 			    optlen > cnt) {
582 				INP_RUNLOCK(inp);
583 				m_freem(m);
584 				return (EINVAL);
585 			}
586 		}
587 		/*
588 		 * This doesn't allow application to specify ID of zero,
589 		 * but we got this limitation from the beginning of history.
590 		 */
591 		if (ip->ip_id == 0)
592 			ip_fillid(ip);
593 
594 		/*
595 		 * XXX prevent ip_output from overwriting header fields.
596 		 */
597 		flags |= IP_RAWOUTPUT;
598 		IPSTAT_INC(ips_rawout);
599 	}
600 
601 	if (inp->inp_flags & INP_ONESBCAST)
602 		flags |= IP_SENDONES;
603 
604 #ifdef MAC
605 	mac_inpcb_create_mbuf(inp, m);
606 #endif
607 
608 	NET_EPOCH_ENTER(et);
609 	error = ip_output(m, inp->inp_options, NULL, flags,
610 	    inp->inp_moptions, inp);
611 	NET_EPOCH_EXIT(et);
612 	INP_RUNLOCK(inp);
613 	return (error);
614 }
615 
616 /*
617  * Raw IP socket option processing.
618  *
619  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
620  * only be created by a privileged process, and as such, socket option
621  * operations to manage system properties on any raw socket were allowed to
622  * take place without explicit additional access control checks.  However,
623  * raw sockets can now also be created in jail(), and therefore explicit
624  * checks are now required.  Likewise, raw sockets can be used by a process
625  * after it gives up privilege, so some caution is required.  For options
626  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
627  * performed in ip_ctloutput() and therefore no check occurs here.
628  * Unilaterally checking priv_check() here breaks normal IP socket option
629  * operations on raw sockets.
630  *
631  * When adding new socket options here, make sure to add access control
632  * checks here as necessary.
633  *
634  * XXX-BZ inp locking?
635  */
636 int
637 rip_ctloutput(struct socket *so, struct sockopt *sopt)
638 {
639 	struct	inpcb *inp = sotoinpcb(so);
640 	int	error, optval;
641 
642 	if (sopt->sopt_level != IPPROTO_IP) {
643 		if ((sopt->sopt_level == SOL_SOCKET) &&
644 		    (sopt->sopt_name == SO_SETFIB)) {
645 			inp->inp_inc.inc_fibnum = so->so_fibnum;
646 			return (0);
647 		}
648 		return (EINVAL);
649 	}
650 
651 	error = 0;
652 	switch (sopt->sopt_dir) {
653 	case SOPT_GET:
654 		switch (sopt->sopt_name) {
655 		case IP_HDRINCL:
656 			optval = inp->inp_flags & INP_HDRINCL;
657 			error = sooptcopyout(sopt, &optval, sizeof optval);
658 			break;
659 
660 		case IP_FW3:	/* generic ipfw v.3 functions */
661 		case IP_FW_ADD:	/* ADD actually returns the body... */
662 		case IP_FW_GET:
663 		case IP_FW_TABLE_GETSIZE:
664 		case IP_FW_TABLE_LIST:
665 		case IP_FW_NAT_GET_CONFIG:
666 		case IP_FW_NAT_GET_LOG:
667 			if (V_ip_fw_ctl_ptr != NULL)
668 				error = V_ip_fw_ctl_ptr(sopt);
669 			else
670 				error = ENOPROTOOPT;
671 			break;
672 
673 		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
674 		case IP_DUMMYNET_GET:
675 			if (ip_dn_ctl_ptr != NULL)
676 				error = ip_dn_ctl_ptr(sopt);
677 			else
678 				error = ENOPROTOOPT;
679 			break ;
680 
681 		case MRT_INIT:
682 		case MRT_DONE:
683 		case MRT_ADD_VIF:
684 		case MRT_DEL_VIF:
685 		case MRT_ADD_MFC:
686 		case MRT_DEL_MFC:
687 		case MRT_VERSION:
688 		case MRT_ASSERT:
689 		case MRT_API_SUPPORT:
690 		case MRT_API_CONFIG:
691 		case MRT_ADD_BW_UPCALL:
692 		case MRT_DEL_BW_UPCALL:
693 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
694 			if (error != 0)
695 				return (error);
696 			if (inp->inp_ip_p != IPPROTO_IGMP)
697 				return (EOPNOTSUPP);
698 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
699 				EOPNOTSUPP;
700 			break;
701 
702 		default:
703 			error = ip_ctloutput(so, sopt);
704 			break;
705 		}
706 		break;
707 
708 	case SOPT_SET:
709 		switch (sopt->sopt_name) {
710 		case IP_HDRINCL:
711 			error = sooptcopyin(sopt, &optval, sizeof optval,
712 					    sizeof optval);
713 			if (error)
714 				break;
715 			if (optval)
716 				inp->inp_flags |= INP_HDRINCL;
717 			else
718 				inp->inp_flags &= ~INP_HDRINCL;
719 			break;
720 
721 		case IP_FW3:	/* generic ipfw v.3 functions */
722 		case IP_FW_ADD:
723 		case IP_FW_DEL:
724 		case IP_FW_FLUSH:
725 		case IP_FW_ZERO:
726 		case IP_FW_RESETLOG:
727 		case IP_FW_TABLE_ADD:
728 		case IP_FW_TABLE_DEL:
729 		case IP_FW_TABLE_FLUSH:
730 		case IP_FW_NAT_CFG:
731 		case IP_FW_NAT_DEL:
732 			if (V_ip_fw_ctl_ptr != NULL)
733 				error = V_ip_fw_ctl_ptr(sopt);
734 			else
735 				error = ENOPROTOOPT;
736 			break;
737 
738 		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
739 		case IP_DUMMYNET_CONFIGURE:
740 		case IP_DUMMYNET_DEL:
741 		case IP_DUMMYNET_FLUSH:
742 			if (ip_dn_ctl_ptr != NULL)
743 				error = ip_dn_ctl_ptr(sopt);
744 			else
745 				error = ENOPROTOOPT ;
746 			break ;
747 
748 		case IP_RSVP_ON:
749 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
750 			if (error != 0)
751 				return (error);
752 			if (inp->inp_ip_p != IPPROTO_RSVP)
753 				return (EOPNOTSUPP);
754 			error = ip_rsvp_init(so);
755 			break;
756 
757 		case IP_RSVP_OFF:
758 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
759 			if (error != 0)
760 				return (error);
761 			error = ip_rsvp_done();
762 			break;
763 
764 		case IP_RSVP_VIF_ON:
765 		case IP_RSVP_VIF_OFF:
766 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
767 			if (error != 0)
768 				return (error);
769 			if (inp->inp_ip_p != IPPROTO_RSVP)
770 				return (EOPNOTSUPP);
771 			error = ip_rsvp_vif ?
772 				ip_rsvp_vif(so, sopt) : EINVAL;
773 			break;
774 
775 		case MRT_INIT:
776 		case MRT_DONE:
777 		case MRT_ADD_VIF:
778 		case MRT_DEL_VIF:
779 		case MRT_ADD_MFC:
780 		case MRT_DEL_MFC:
781 		case MRT_VERSION:
782 		case MRT_ASSERT:
783 		case MRT_API_SUPPORT:
784 		case MRT_API_CONFIG:
785 		case MRT_ADD_BW_UPCALL:
786 		case MRT_DEL_BW_UPCALL:
787 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
788 			if (error != 0)
789 				return (error);
790 			if (inp->inp_ip_p != IPPROTO_IGMP)
791 				return (EOPNOTSUPP);
792 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
793 					EOPNOTSUPP;
794 			break;
795 
796 		default:
797 			error = ip_ctloutput(so, sopt);
798 			break;
799 		}
800 		break;
801 	}
802 
803 	return (error);
804 }
805 
806 void
807 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
808 {
809 
810 	switch (cmd) {
811 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
812 	case PRC_MSGSIZE:
813 		if (IPSEC_ENABLED(ipv4))
814 			IPSEC_CTLINPUT(ipv4, cmd, sa, vip);
815 		break;
816 #endif
817 	}
818 }
819 
820 static int
821 rip_attach(struct socket *so, int proto, struct thread *td)
822 {
823 	struct inpcb *inp;
824 	int error;
825 
826 	inp = sotoinpcb(so);
827 	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
828 
829 	error = priv_check(td, PRIV_NETINET_RAW);
830 	if (error)
831 		return (error);
832 	if (proto >= IPPROTO_MAX || proto < 0)
833 		return EPROTONOSUPPORT;
834 	error = soreserve(so, rip_sendspace, rip_recvspace);
835 	if (error)
836 		return (error);
837 	error = in_pcballoc(so, &V_ripcbinfo);
838 	if (error)
839 		return (error);
840 	inp = (struct inpcb *)so->so_pcb;
841 	inp->inp_ip_p = proto;
842 	inp->inp_ip_ttl = V_ip_defttl;
843 	INP_HASH_WLOCK(&V_ripcbinfo);
844 	rip_inshash(inp);
845 	INP_HASH_WUNLOCK(&V_ripcbinfo);
846 	INP_WUNLOCK(inp);
847 	return (0);
848 }
849 
850 static void
851 rip_detach(struct socket *so)
852 {
853 	struct inpcb *inp;
854 
855 	inp = sotoinpcb(so);
856 	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
857 	KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
858 	    ("rip_detach: not closed"));
859 
860 	/* Disable mrouter first */
861 	if (so == V_ip_mrouter && ip_mrouter_done)
862 		ip_mrouter_done();
863 
864 	INP_WLOCK(inp);
865 	INP_HASH_WLOCK(&V_ripcbinfo);
866 	rip_delhash(inp);
867 	INP_HASH_WUNLOCK(&V_ripcbinfo);
868 
869 	if (ip_rsvp_force_done)
870 		ip_rsvp_force_done(so);
871 	if (so == V_ip_rsvpd)
872 		ip_rsvp_done();
873 	in_pcbdetach(inp);
874 	in_pcbfree(inp);
875 }
876 
877 static void
878 rip_dodisconnect(struct socket *so, struct inpcb *inp)
879 {
880 	struct inpcbinfo *pcbinfo;
881 
882 	pcbinfo = inp->inp_pcbinfo;
883 	INP_WLOCK(inp);
884 	INP_HASH_WLOCK(pcbinfo);
885 	rip_delhash(inp);
886 	inp->inp_faddr.s_addr = INADDR_ANY;
887 	rip_inshash(inp);
888 	INP_HASH_WUNLOCK(pcbinfo);
889 	SOCK_LOCK(so);
890 	so->so_state &= ~SS_ISCONNECTED;
891 	SOCK_UNLOCK(so);
892 	INP_WUNLOCK(inp);
893 }
894 
895 static void
896 rip_abort(struct socket *so)
897 {
898 	struct inpcb *inp;
899 
900 	inp = sotoinpcb(so);
901 	KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
902 
903 	rip_dodisconnect(so, inp);
904 }
905 
906 static void
907 rip_close(struct socket *so)
908 {
909 	struct inpcb *inp;
910 
911 	inp = sotoinpcb(so);
912 	KASSERT(inp != NULL, ("rip_close: inp == NULL"));
913 
914 	rip_dodisconnect(so, inp);
915 }
916 
917 static int
918 rip_disconnect(struct socket *so)
919 {
920 	struct inpcb *inp;
921 
922 	if ((so->so_state & SS_ISCONNECTED) == 0)
923 		return (ENOTCONN);
924 
925 	inp = sotoinpcb(so);
926 	KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
927 
928 	rip_dodisconnect(so, inp);
929 	return (0);
930 }
931 
932 static int
933 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
934 {
935 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
936 	struct inpcb *inp;
937 	int error;
938 
939 	if (nam->sa_family != AF_INET)
940 		return (EAFNOSUPPORT);
941 	if (nam->sa_len != sizeof(*addr))
942 		return (EINVAL);
943 
944 	error = prison_check_ip4(td->td_ucred, &addr->sin_addr);
945 	if (error != 0)
946 		return (error);
947 
948 	inp = sotoinpcb(so);
949 	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
950 
951 	if (CK_STAILQ_EMPTY(&V_ifnet) ||
952 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
953 	    (addr->sin_addr.s_addr &&
954 	     (inp->inp_flags & INP_BINDANY) == 0 &&
955 	     ifa_ifwithaddr_check((struct sockaddr *)addr) == 0))
956 		return (EADDRNOTAVAIL);
957 
958 	INP_WLOCK(inp);
959 	INP_HASH_WLOCK(&V_ripcbinfo);
960 	rip_delhash(inp);
961 	inp->inp_laddr = addr->sin_addr;
962 	rip_inshash(inp);
963 	INP_HASH_WUNLOCK(&V_ripcbinfo);
964 	INP_WUNLOCK(inp);
965 	return (0);
966 }
967 
968 static int
969 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
970 {
971 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
972 	struct inpcb *inp;
973 
974 	if (nam->sa_len != sizeof(*addr))
975 		return (EINVAL);
976 	if (CK_STAILQ_EMPTY(&V_ifnet))
977 		return (EADDRNOTAVAIL);
978 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
979 		return (EAFNOSUPPORT);
980 
981 	inp = sotoinpcb(so);
982 	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
983 
984 	INP_WLOCK(inp);
985 	INP_HASH_WLOCK(&V_ripcbinfo);
986 	rip_delhash(inp);
987 	inp->inp_faddr = addr->sin_addr;
988 	rip_inshash(inp);
989 	INP_HASH_WUNLOCK(&V_ripcbinfo);
990 	soisconnected(so);
991 	INP_WUNLOCK(inp);
992 	return (0);
993 }
994 
995 static int
996 rip_shutdown(struct socket *so)
997 {
998 	struct inpcb *inp;
999 
1000 	inp = sotoinpcb(so);
1001 	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
1002 
1003 	INP_WLOCK(inp);
1004 	socantsendmore(so);
1005 	INP_WUNLOCK(inp);
1006 	return (0);
1007 }
1008 #endif /* INET */
1009 
1010 static int
1011 rip_pcblist(SYSCTL_HANDLER_ARGS)
1012 {
1013 	struct inpcb_iterator inpi = INP_ALL_ITERATOR(&V_ripcbinfo,
1014 	    INPLOOKUP_RLOCKPCB);
1015 	struct xinpgen xig;
1016 	struct inpcb *inp;
1017 	int error;
1018 
1019 	if (req->newptr != 0)
1020 		return (EPERM);
1021 
1022 	if (req->oldptr == 0) {
1023 		int n;
1024 
1025 		n = V_ripcbinfo.ipi_count;
1026 		n += imax(n / 8, 10);
1027 		req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
1028 		return (0);
1029 	}
1030 
1031 	if ((error = sysctl_wire_old_buffer(req, 0)) != 0)
1032 		return (error);
1033 
1034 	bzero(&xig, sizeof(xig));
1035 	xig.xig_len = sizeof xig;
1036 	xig.xig_count = V_ripcbinfo.ipi_count;
1037 	xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1038 	xig.xig_sogen = so_gencnt;
1039 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1040 	if (error)
1041 		return (error);
1042 
1043 	while ((inp = inp_next(&inpi)) != NULL) {
1044 		if (inp->inp_gencnt <= xig.xig_gen &&
1045 		    cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
1046 			struct xinpcb xi;
1047 
1048 			in_pcbtoxinpcb(inp, &xi);
1049 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1050 			if (error) {
1051 				INP_RUNLOCK(inp);
1052 				break;
1053 			}
1054 		}
1055 	}
1056 
1057 	if (!error) {
1058 		/*
1059 		 * Give the user an updated idea of our state.  If the
1060 		 * generation differs from what we told her before, she knows
1061 		 * that something happened while we were processing this
1062 		 * request, and it might be necessary to retry.
1063 		 */
1064 		xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1065 		xig.xig_sogen = so_gencnt;
1066 		xig.xig_count = V_ripcbinfo.ipi_count;
1067 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1068 	}
1069 
1070 	return (error);
1071 }
1072 
1073 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist,
1074     CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
1075     rip_pcblist, "S,xinpcb",
1076     "List of active raw IP sockets");
1077 
1078 #ifdef INET
1079 struct protosw rip_protosw = {
1080 	.pr_type =		SOCK_RAW,
1081 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1082 	.pr_ctloutput =		rip_ctloutput,
1083 	.pr_abort =		rip_abort,
1084 	.pr_attach =		rip_attach,
1085 	.pr_bind =		rip_bind,
1086 	.pr_connect =		rip_connect,
1087 	.pr_control =		in_control,
1088 	.pr_detach =		rip_detach,
1089 	.pr_disconnect =	rip_disconnect,
1090 	.pr_peeraddr =		in_getpeeraddr,
1091 	.pr_send =		rip_send,
1092 	.pr_shutdown =		rip_shutdown,
1093 	.pr_sockaddr =		in_getsockaddr,
1094 	.pr_sosetlabel =	in_pcbsosetlabel,
1095 	.pr_close =		rip_close
1096 };
1097 #endif /* INET */
1098