xref: /freebsd/sys/netinet/raw_ip.c (revision 145992504973bd16cf3518af9ba5ce185fefa82a)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 4. Neither the name of the University nor the names of its contributors
15  *    may be used to endorse or promote products derived from this software
16  *    without specific prior written permission.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28  * SUCH DAMAGE.
29  *
30  *	@(#)raw_ip.c	8.7 (Berkeley) 5/15/95
31  */
32 
33 #include <sys/cdefs.h>
34 __FBSDID("$FreeBSD$");
35 
36 #include "opt_inet.h"
37 #include "opt_inet6.h"
38 #include "opt_ipsec.h"
39 
40 #include <sys/param.h>
41 #include <sys/jail.h>
42 #include <sys/kernel.h>
43 #include <sys/lock.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/priv.h>
47 #include <sys/proc.h>
48 #include <sys/protosw.h>
49 #include <sys/rwlock.h>
50 #include <sys/signalvar.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/sx.h>
54 #include <sys/sysctl.h>
55 #include <sys/systm.h>
56 
57 #include <vm/uma.h>
58 
59 #include <net/if.h>
60 #include <net/route.h>
61 #include <net/vnet.h>
62 
63 #include <netinet/in.h>
64 #include <netinet/in_systm.h>
65 #include <netinet/in_pcb.h>
66 #include <netinet/in_var.h>
67 #include <netinet/if_ether.h>
68 #include <netinet/ip.h>
69 #include <netinet/ip_var.h>
70 #include <netinet/ip_mroute.h>
71 
72 #ifdef IPSEC
73 #include <netipsec/ipsec.h>
74 #endif /*IPSEC*/
75 
76 #include <security/mac/mac_framework.h>
77 
78 VNET_DEFINE(int, ip_defttl) = IPDEFTTL;
79 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
80     &VNET_NAME(ip_defttl), 0,
81     "Maximum TTL on IP packets");
82 
83 VNET_DEFINE(struct inpcbhead, ripcb);
84 VNET_DEFINE(struct inpcbinfo, ripcbinfo);
85 
86 #define	V_ripcb			VNET(ripcb)
87 #define	V_ripcbinfo		VNET(ripcbinfo)
88 
89 /*
90  * Control and data hooks for ipfw, dummynet, divert and so on.
91  * The data hooks are not used here but it is convenient
92  * to keep them all in one place.
93  */
94 VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL;
95 VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL;
96 
97 int	(*ip_dn_ctl_ptr)(struct sockopt *);
98 int	(*ip_dn_io_ptr)(struct mbuf **, int, struct ip_fw_args *);
99 void	(*ip_divert_ptr)(struct mbuf *, int);
100 int	(*ng_ipfw_input_p)(struct mbuf **, int,
101 			struct ip_fw_args *, int);
102 
103 #ifdef INET
104 /*
105  * Hooks for multicast routing. They all default to NULL, so leave them not
106  * initialized and rely on BSS being set to 0.
107  */
108 
109 /*
110  * The socket used to communicate with the multicast routing daemon.
111  */
112 VNET_DEFINE(struct socket *, ip_mrouter);
113 
114 /*
115  * The various mrouter and rsvp functions.
116  */
117 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
118 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
119 int (*ip_mrouter_done)(void);
120 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
121 		   struct ip_moptions *);
122 int (*mrt_ioctl)(u_long, caddr_t, int);
123 int (*legal_vif_num)(int);
124 u_long (*ip_mcast_src)(int);
125 
126 void (*rsvp_input_p)(struct mbuf *m, int off);
127 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
128 void (*ip_rsvp_force_done)(struct socket *);
129 #endif /* INET */
130 
131 u_long	rip_sendspace = 9216;
132 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
133     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
134 
135 u_long	rip_recvspace = 9216;
136 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
137     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
138 
139 /*
140  * Hash functions
141  */
142 
143 #define INP_PCBHASH_RAW_SIZE	256
144 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \
145         (((proto) + (laddr) + (faddr)) % (mask) + 1)
146 
147 #ifdef INET
148 static void
149 rip_inshash(struct inpcb *inp)
150 {
151 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
152 	struct inpcbhead *pcbhash;
153 	int hash;
154 
155 	INP_INFO_WLOCK_ASSERT(pcbinfo);
156 	INP_WLOCK_ASSERT(inp);
157 
158 	if (inp->inp_ip_p != 0 &&
159 	    inp->inp_laddr.s_addr != INADDR_ANY &&
160 	    inp->inp_faddr.s_addr != INADDR_ANY) {
161 		hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr,
162 		    inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask);
163 	} else
164 		hash = 0;
165 	pcbhash = &pcbinfo->ipi_hashbase[hash];
166 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
167 }
168 
169 static void
170 rip_delhash(struct inpcb *inp)
171 {
172 
173 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
174 	INP_WLOCK_ASSERT(inp);
175 
176 	LIST_REMOVE(inp, inp_hash);
177 }
178 #endif /* INET */
179 
180 /*
181  * Raw interface to IP protocol.
182  */
183 
184 /*
185  * Initialize raw connection block q.
186  */
187 static void
188 rip_zone_change(void *tag)
189 {
190 
191 	uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets);
192 }
193 
194 static int
195 rip_inpcb_init(void *mem, int size, int flags)
196 {
197 	struct inpcb *inp = mem;
198 
199 	INP_LOCK_INIT(inp, "inp", "rawinp");
200 	return (0);
201 }
202 
203 void
204 rip_init(void)
205 {
206 
207 	in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE,
208 	    1, "ripcb", rip_inpcb_init, NULL, UMA_ZONE_NOFREE,
209 	    IPI_HASHFIELDS_NONE);
210 	EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL,
211 	    EVENTHANDLER_PRI_ANY);
212 }
213 
214 #ifdef VIMAGE
215 void
216 rip_destroy(void)
217 {
218 
219 	in_pcbinfo_destroy(&V_ripcbinfo);
220 }
221 #endif
222 
223 #ifdef INET
224 static int
225 rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
226     struct sockaddr_in *ripsrc)
227 {
228 	int policyfail = 0;
229 
230 	INP_LOCK_ASSERT(last);
231 
232 #ifdef IPSEC
233 	/* check AH/ESP integrity. */
234 	if (ipsec4_in_reject(n, last)) {
235 		policyfail = 1;
236 	}
237 #endif /* IPSEC */
238 #ifdef MAC
239 	if (!policyfail && mac_inpcb_check_deliver(last, n) != 0)
240 		policyfail = 1;
241 #endif
242 	/* Check the minimum TTL for socket. */
243 	if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
244 		policyfail = 1;
245 	if (!policyfail) {
246 		struct mbuf *opts = NULL;
247 		struct socket *so;
248 
249 		so = last->inp_socket;
250 		if ((last->inp_flags & INP_CONTROLOPTS) ||
251 		    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
252 			ip_savecontrol(last, &opts, ip, n);
253 		SOCKBUF_LOCK(&so->so_rcv);
254 		if (sbappendaddr_locked(&so->so_rcv,
255 		    (struct sockaddr *)ripsrc, n, opts) == 0) {
256 			/* should notify about lost packet */
257 			m_freem(n);
258 			if (opts)
259 				m_freem(opts);
260 			SOCKBUF_UNLOCK(&so->so_rcv);
261 		} else
262 			sorwakeup_locked(so);
263 	} else
264 		m_freem(n);
265 	return (policyfail);
266 }
267 
268 /*
269  * Setup generic address and protocol structures for raw_input routine, then
270  * pass them along with mbuf chain.
271  */
272 void
273 rip_input(struct mbuf *m, int off)
274 {
275 	struct ifnet *ifp;
276 	struct ip *ip = mtod(m, struct ip *);
277 	int proto = ip->ip_p;
278 	struct inpcb *inp, *last;
279 	struct sockaddr_in ripsrc;
280 	int hash;
281 
282 	bzero(&ripsrc, sizeof(ripsrc));
283 	ripsrc.sin_len = sizeof(ripsrc);
284 	ripsrc.sin_family = AF_INET;
285 	ripsrc.sin_addr = ip->ip_src;
286 	last = NULL;
287 
288 	ifp = m->m_pkthdr.rcvif;
289 	/*
290 	 * Add back the IP header length which was
291 	 * removed by ip_input().  Raw sockets do
292 	 * not modify the packet except for some
293 	 * byte order swaps.
294 	 */
295 	ip->ip_len += off;
296 
297 	hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr,
298 	    ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
299 	INP_INFO_RLOCK(&V_ripcbinfo);
300 	LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) {
301 		if (inp->inp_ip_p != proto)
302 			continue;
303 #ifdef INET6
304 		/* XXX inp locking */
305 		if ((inp->inp_vflag & INP_IPV4) == 0)
306 			continue;
307 #endif
308 		if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
309 			continue;
310 		if (inp->inp_faddr.s_addr != ip->ip_src.s_addr)
311 			continue;
312 		if (jailed_without_vnet(inp->inp_cred)) {
313 			/*
314 			 * XXX: If faddr was bound to multicast group,
315 			 * jailed raw socket will drop datagram.
316 			 */
317 			if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
318 				continue;
319 		}
320 		if (last != NULL) {
321 			struct mbuf *n;
322 
323 			n = m_copy(m, 0, (int)M_COPYALL);
324 			if (n != NULL)
325 		    	    (void) rip_append(last, ip, n, &ripsrc);
326 			/* XXX count dropped packet */
327 			INP_RUNLOCK(last);
328 		}
329 		INP_RLOCK(inp);
330 		last = inp;
331 	}
332 	LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) {
333 		if (inp->inp_ip_p && inp->inp_ip_p != proto)
334 			continue;
335 #ifdef INET6
336 		/* XXX inp locking */
337 		if ((inp->inp_vflag & INP_IPV4) == 0)
338 			continue;
339 #endif
340 		if (!in_nullhost(inp->inp_laddr) &&
341 		    !in_hosteq(inp->inp_laddr, ip->ip_dst))
342 			continue;
343 		if (!in_nullhost(inp->inp_faddr) &&
344 		    !in_hosteq(inp->inp_faddr, ip->ip_src))
345 			continue;
346 		if (jailed_without_vnet(inp->inp_cred)) {
347 			/*
348 			 * Allow raw socket in jail to receive multicast;
349 			 * assume process had PRIV_NETINET_RAW at attach,
350 			 * and fall through into normal filter path if so.
351 			 */
352 			if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
353 			    prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
354 				continue;
355 		}
356 		/*
357 		 * If this raw socket has multicast state, and we
358 		 * have received a multicast, check if this socket
359 		 * should receive it, as multicast filtering is now
360 		 * the responsibility of the transport layer.
361 		 */
362 		if (inp->inp_moptions != NULL &&
363 		    IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
364 			/*
365 			 * If the incoming datagram is for IGMP, allow it
366 			 * through unconditionally to the raw socket.
367 			 *
368 			 * In the case of IGMPv2, we may not have explicitly
369 			 * joined the group, and may have set IFF_ALLMULTI
370 			 * on the interface. imo_multi_filter() may discard
371 			 * control traffic we actually need to see.
372 			 *
373 			 * Userland multicast routing daemons should continue
374 			 * filter the control traffic appropriately.
375 			 */
376 			int blocked;
377 
378 			blocked = MCAST_PASS;
379 			if (proto != IPPROTO_IGMP) {
380 				struct sockaddr_in group;
381 
382 				bzero(&group, sizeof(struct sockaddr_in));
383 				group.sin_len = sizeof(struct sockaddr_in);
384 				group.sin_family = AF_INET;
385 				group.sin_addr = ip->ip_dst;
386 
387 				blocked = imo_multi_filter(inp->inp_moptions,
388 				    ifp,
389 				    (struct sockaddr *)&group,
390 				    (struct sockaddr *)&ripsrc);
391 			}
392 
393 			if (blocked != MCAST_PASS) {
394 				IPSTAT_INC(ips_notmember);
395 				continue;
396 			}
397 		}
398 		if (last != NULL) {
399 			struct mbuf *n;
400 
401 			n = m_copy(m, 0, (int)M_COPYALL);
402 			if (n != NULL)
403 				(void) rip_append(last, ip, n, &ripsrc);
404 			/* XXX count dropped packet */
405 			INP_RUNLOCK(last);
406 		}
407 		INP_RLOCK(inp);
408 		last = inp;
409 	}
410 	INP_INFO_RUNLOCK(&V_ripcbinfo);
411 	if (last != NULL) {
412 		if (rip_append(last, ip, m, &ripsrc) != 0)
413 			IPSTAT_INC(ips_delivered);
414 		INP_RUNLOCK(last);
415 	} else {
416 		m_freem(m);
417 		IPSTAT_INC(ips_noproto);
418 		IPSTAT_DEC(ips_delivered);
419 	}
420 }
421 
422 /*
423  * Generate IP header and pass packet to ip_output.  Tack on options user may
424  * have setup with control call.
425  */
426 int
427 rip_output(struct mbuf *m, struct socket *so, u_long dst)
428 {
429 	struct ip *ip;
430 	int error;
431 	struct inpcb *inp = sotoinpcb(so);
432 	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
433 	    IP_ALLOWBROADCAST;
434 
435 	/*
436 	 * If the user handed us a complete IP packet, use it.  Otherwise,
437 	 * allocate an mbuf for a header and fill it in.
438 	 */
439 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
440 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
441 			m_freem(m);
442 			return(EMSGSIZE);
443 		}
444 		M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
445 		if (m == NULL)
446 			return(ENOBUFS);
447 
448 		INP_RLOCK(inp);
449 		ip = mtod(m, struct ip *);
450 		ip->ip_tos = inp->inp_ip_tos;
451 		if (inp->inp_flags & INP_DONTFRAG)
452 			ip->ip_off = IP_DF;
453 		else
454 			ip->ip_off = 0;
455 		ip->ip_p = inp->inp_ip_p;
456 		ip->ip_len = m->m_pkthdr.len;
457 		ip->ip_src = inp->inp_laddr;
458 		if (jailed(inp->inp_cred)) {
459 			/*
460 			 * prison_local_ip4() would be good enough but would
461 			 * let a source of INADDR_ANY pass, which we do not
462 			 * want to see from jails. We do not go through the
463 			 * pain of in_pcbladdr() for raw sockets.
464 			 */
465 			if (ip->ip_src.s_addr == INADDR_ANY)
466 				error = prison_get_ip4(inp->inp_cred,
467 				    &ip->ip_src);
468 			else
469 				error = prison_local_ip4(inp->inp_cred,
470 				    &ip->ip_src);
471 			if (error != 0) {
472 				INP_RUNLOCK(inp);
473 				m_freem(m);
474 				return (error);
475 			}
476 		}
477 		ip->ip_dst.s_addr = dst;
478 		ip->ip_ttl = inp->inp_ip_ttl;
479 	} else {
480 		if (m->m_pkthdr.len > IP_MAXPACKET) {
481 			m_freem(m);
482 			return(EMSGSIZE);
483 		}
484 		INP_RLOCK(inp);
485 		ip = mtod(m, struct ip *);
486 		error = prison_check_ip4(inp->inp_cred, &ip->ip_src);
487 		if (error != 0) {
488 			INP_RUNLOCK(inp);
489 			m_freem(m);
490 			return (error);
491 		}
492 
493 		/*
494 		 * Don't allow both user specified and setsockopt options,
495 		 * and don't allow packet length sizes that will crash.
496 		 */
497 		if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options)
498 		    || (ip->ip_len > m->m_pkthdr.len)
499 		    || (ip->ip_len < (ip->ip_hl << 2))) {
500 			INP_RUNLOCK(inp);
501 			m_freem(m);
502 			return (EINVAL);
503 		}
504 		if (ip->ip_id == 0)
505 			ip->ip_id = ip_newid();
506 
507 		/*
508 		 * XXX prevent ip_output from overwriting header fields.
509 		 */
510 		flags |= IP_RAWOUTPUT;
511 		IPSTAT_INC(ips_rawout);
512 	}
513 
514 	if (inp->inp_flags & INP_ONESBCAST)
515 		flags |= IP_SENDONES;
516 
517 #ifdef MAC
518 	mac_inpcb_create_mbuf(inp, m);
519 #endif
520 
521 	error = ip_output(m, inp->inp_options, NULL, flags,
522 	    inp->inp_moptions, inp);
523 	INP_RUNLOCK(inp);
524 	return (error);
525 }
526 
527 /*
528  * Raw IP socket option processing.
529  *
530  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
531  * only be created by a privileged process, and as such, socket option
532  * operations to manage system properties on any raw socket were allowed to
533  * take place without explicit additional access control checks.  However,
534  * raw sockets can now also be created in jail(), and therefore explicit
535  * checks are now required.  Likewise, raw sockets can be used by a process
536  * after it gives up privilege, so some caution is required.  For options
537  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
538  * performed in ip_ctloutput() and therefore no check occurs here.
539  * Unilaterally checking priv_check() here breaks normal IP socket option
540  * operations on raw sockets.
541  *
542  * When adding new socket options here, make sure to add access control
543  * checks here as necessary.
544  *
545  * XXX-BZ inp locking?
546  */
547 int
548 rip_ctloutput(struct socket *so, struct sockopt *sopt)
549 {
550 	struct	inpcb *inp = sotoinpcb(so);
551 	int	error, optval;
552 
553 	if (sopt->sopt_level != IPPROTO_IP) {
554 		if ((sopt->sopt_level == SOL_SOCKET) &&
555 		    (sopt->sopt_name == SO_SETFIB)) {
556 			inp->inp_inc.inc_fibnum = so->so_fibnum;
557 			return (0);
558 		}
559 		return (EINVAL);
560 	}
561 
562 	error = 0;
563 	switch (sopt->sopt_dir) {
564 	case SOPT_GET:
565 		switch (sopt->sopt_name) {
566 		case IP_HDRINCL:
567 			optval = inp->inp_flags & INP_HDRINCL;
568 			error = sooptcopyout(sopt, &optval, sizeof optval);
569 			break;
570 
571 		case IP_FW3:	/* generic ipfw v.3 functions */
572 		case IP_FW_ADD:	/* ADD actually returns the body... */
573 		case IP_FW_GET:
574 		case IP_FW_TABLE_GETSIZE:
575 		case IP_FW_TABLE_LIST:
576 		case IP_FW_NAT_GET_CONFIG:
577 		case IP_FW_NAT_GET_LOG:
578 			if (V_ip_fw_ctl_ptr != NULL)
579 				error = V_ip_fw_ctl_ptr(sopt);
580 			else
581 				error = ENOPROTOOPT;
582 			break;
583 
584 		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
585 		case IP_DUMMYNET_GET:
586 			if (ip_dn_ctl_ptr != NULL)
587 				error = ip_dn_ctl_ptr(sopt);
588 			else
589 				error = ENOPROTOOPT;
590 			break ;
591 
592 		case MRT_INIT:
593 		case MRT_DONE:
594 		case MRT_ADD_VIF:
595 		case MRT_DEL_VIF:
596 		case MRT_ADD_MFC:
597 		case MRT_DEL_MFC:
598 		case MRT_VERSION:
599 		case MRT_ASSERT:
600 		case MRT_API_SUPPORT:
601 		case MRT_API_CONFIG:
602 		case MRT_ADD_BW_UPCALL:
603 		case MRT_DEL_BW_UPCALL:
604 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
605 			if (error != 0)
606 				return (error);
607 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
608 				EOPNOTSUPP;
609 			break;
610 
611 		default:
612 			error = ip_ctloutput(so, sopt);
613 			break;
614 		}
615 		break;
616 
617 	case SOPT_SET:
618 		switch (sopt->sopt_name) {
619 		case IP_HDRINCL:
620 			error = sooptcopyin(sopt, &optval, sizeof optval,
621 					    sizeof optval);
622 			if (error)
623 				break;
624 			if (optval)
625 				inp->inp_flags |= INP_HDRINCL;
626 			else
627 				inp->inp_flags &= ~INP_HDRINCL;
628 			break;
629 
630 		case IP_FW3:	/* generic ipfw v.3 functions */
631 		case IP_FW_ADD:
632 		case IP_FW_DEL:
633 		case IP_FW_FLUSH:
634 		case IP_FW_ZERO:
635 		case IP_FW_RESETLOG:
636 		case IP_FW_TABLE_ADD:
637 		case IP_FW_TABLE_DEL:
638 		case IP_FW_TABLE_FLUSH:
639 		case IP_FW_NAT_CFG:
640 		case IP_FW_NAT_DEL:
641 			if (V_ip_fw_ctl_ptr != NULL)
642 				error = V_ip_fw_ctl_ptr(sopt);
643 			else
644 				error = ENOPROTOOPT;
645 			break;
646 
647 		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
648 		case IP_DUMMYNET_CONFIGURE:
649 		case IP_DUMMYNET_DEL:
650 		case IP_DUMMYNET_FLUSH:
651 			if (ip_dn_ctl_ptr != NULL)
652 				error = ip_dn_ctl_ptr(sopt);
653 			else
654 				error = ENOPROTOOPT ;
655 			break ;
656 
657 		case IP_RSVP_ON:
658 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
659 			if (error != 0)
660 				return (error);
661 			error = ip_rsvp_init(so);
662 			break;
663 
664 		case IP_RSVP_OFF:
665 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
666 			if (error != 0)
667 				return (error);
668 			error = ip_rsvp_done();
669 			break;
670 
671 		case IP_RSVP_VIF_ON:
672 		case IP_RSVP_VIF_OFF:
673 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
674 			if (error != 0)
675 				return (error);
676 			error = ip_rsvp_vif ?
677 				ip_rsvp_vif(so, sopt) : EINVAL;
678 			break;
679 
680 		case MRT_INIT:
681 		case MRT_DONE:
682 		case MRT_ADD_VIF:
683 		case MRT_DEL_VIF:
684 		case MRT_ADD_MFC:
685 		case MRT_DEL_MFC:
686 		case MRT_VERSION:
687 		case MRT_ASSERT:
688 		case MRT_API_SUPPORT:
689 		case MRT_API_CONFIG:
690 		case MRT_ADD_BW_UPCALL:
691 		case MRT_DEL_BW_UPCALL:
692 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
693 			if (error != 0)
694 				return (error);
695 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
696 					EOPNOTSUPP;
697 			break;
698 
699 		default:
700 			error = ip_ctloutput(so, sopt);
701 			break;
702 		}
703 		break;
704 	}
705 
706 	return (error);
707 }
708 
709 /*
710  * This function exists solely to receive the PRC_IFDOWN messages which are
711  * sent by if_down().  It looks for an ifaddr whose ifa_addr is sa, and calls
712  * in_ifadown() to remove all routes corresponding to that address.  It also
713  * receives the PRC_IFUP messages from if_up() and reinstalls the interface
714  * routes.
715  */
716 void
717 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
718 {
719 	struct in_ifaddr *ia;
720 	struct ifnet *ifp;
721 	int err;
722 	int flags;
723 
724 	switch (cmd) {
725 	case PRC_IFDOWN:
726 		IN_IFADDR_RLOCK();
727 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
728 			if (ia->ia_ifa.ifa_addr == sa
729 			    && (ia->ia_flags & IFA_ROUTE)) {
730 				ifa_ref(&ia->ia_ifa);
731 				IN_IFADDR_RUNLOCK();
732 				/*
733 				 * in_ifscrub kills the interface route.
734 				 */
735 				in_ifscrub(ia->ia_ifp, ia, 0);
736 				/*
737 				 * in_ifadown gets rid of all the rest of the
738 				 * routes.  This is not quite the right thing
739 				 * to do, but at least if we are running a
740 				 * routing process they will come back.
741 				 */
742 				in_ifadown(&ia->ia_ifa, 0);
743 				ifa_free(&ia->ia_ifa);
744 				break;
745 			}
746 		}
747 		if (ia == NULL)		/* If ia matched, already unlocked. */
748 			IN_IFADDR_RUNLOCK();
749 		break;
750 
751 	case PRC_IFUP:
752 		IN_IFADDR_RLOCK();
753 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
754 			if (ia->ia_ifa.ifa_addr == sa)
755 				break;
756 		}
757 		if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) {
758 			IN_IFADDR_RUNLOCK();
759 			return;
760 		}
761 		ifa_ref(&ia->ia_ifa);
762 		IN_IFADDR_RUNLOCK();
763 		flags = RTF_UP;
764 		ifp = ia->ia_ifa.ifa_ifp;
765 
766 		if ((ifp->if_flags & IFF_LOOPBACK)
767 		    || (ifp->if_flags & IFF_POINTOPOINT))
768 			flags |= RTF_HOST;
769 
770 		err = ifa_del_loopback_route((struct ifaddr *)ia, sa);
771 		if (err == 0)
772 			ia->ia_flags &= ~IFA_RTSELF;
773 
774 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
775 		if (err == 0)
776 			ia->ia_flags |= IFA_ROUTE;
777 
778 		err = ifa_add_loopback_route((struct ifaddr *)ia, sa);
779 		if (err == 0)
780 			ia->ia_flags |= IFA_RTSELF;
781 
782 		ifa_free(&ia->ia_ifa);
783 		break;
784 	}
785 }
786 
787 static int
788 rip_attach(struct socket *so, int proto, struct thread *td)
789 {
790 	struct inpcb *inp;
791 	int error;
792 
793 	inp = sotoinpcb(so);
794 	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
795 
796 	error = priv_check(td, PRIV_NETINET_RAW);
797 	if (error)
798 		return (error);
799 	if (proto >= IPPROTO_MAX || proto < 0)
800 		return EPROTONOSUPPORT;
801 	error = soreserve(so, rip_sendspace, rip_recvspace);
802 	if (error)
803 		return (error);
804 	INP_INFO_WLOCK(&V_ripcbinfo);
805 	error = in_pcballoc(so, &V_ripcbinfo);
806 	if (error) {
807 		INP_INFO_WUNLOCK(&V_ripcbinfo);
808 		return (error);
809 	}
810 	inp = (struct inpcb *)so->so_pcb;
811 	inp->inp_vflag |= INP_IPV4;
812 	inp->inp_ip_p = proto;
813 	inp->inp_ip_ttl = V_ip_defttl;
814 	rip_inshash(inp);
815 	INP_INFO_WUNLOCK(&V_ripcbinfo);
816 	INP_WUNLOCK(inp);
817 	return (0);
818 }
819 
820 static void
821 rip_detach(struct socket *so)
822 {
823 	struct inpcb *inp;
824 
825 	inp = sotoinpcb(so);
826 	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
827 	KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
828 	    ("rip_detach: not closed"));
829 
830 	INP_INFO_WLOCK(&V_ripcbinfo);
831 	INP_WLOCK(inp);
832 	rip_delhash(inp);
833 	if (so == V_ip_mrouter && ip_mrouter_done)
834 		ip_mrouter_done();
835 	if (ip_rsvp_force_done)
836 		ip_rsvp_force_done(so);
837 	if (so == V_ip_rsvpd)
838 		ip_rsvp_done();
839 	in_pcbdetach(inp);
840 	in_pcbfree(inp);
841 	INP_INFO_WUNLOCK(&V_ripcbinfo);
842 }
843 
844 static void
845 rip_dodisconnect(struct socket *so, struct inpcb *inp)
846 {
847 	struct inpcbinfo *pcbinfo;
848 
849 	pcbinfo = inp->inp_pcbinfo;
850 	INP_INFO_WLOCK(pcbinfo);
851 	INP_WLOCK(inp);
852 	rip_delhash(inp);
853 	inp->inp_faddr.s_addr = INADDR_ANY;
854 	rip_inshash(inp);
855 	SOCK_LOCK(so);
856 	so->so_state &= ~SS_ISCONNECTED;
857 	SOCK_UNLOCK(so);
858 	INP_WUNLOCK(inp);
859 	INP_INFO_WUNLOCK(pcbinfo);
860 }
861 
862 static void
863 rip_abort(struct socket *so)
864 {
865 	struct inpcb *inp;
866 
867 	inp = sotoinpcb(so);
868 	KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
869 
870 	rip_dodisconnect(so, inp);
871 }
872 
873 static void
874 rip_close(struct socket *so)
875 {
876 	struct inpcb *inp;
877 
878 	inp = sotoinpcb(so);
879 	KASSERT(inp != NULL, ("rip_close: inp == NULL"));
880 
881 	rip_dodisconnect(so, inp);
882 }
883 
884 static int
885 rip_disconnect(struct socket *so)
886 {
887 	struct inpcb *inp;
888 
889 	if ((so->so_state & SS_ISCONNECTED) == 0)
890 		return (ENOTCONN);
891 
892 	inp = sotoinpcb(so);
893 	KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
894 
895 	rip_dodisconnect(so, inp);
896 	return (0);
897 }
898 
899 static int
900 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
901 {
902 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
903 	struct inpcb *inp;
904 	int error;
905 
906 	if (nam->sa_len != sizeof(*addr))
907 		return (EINVAL);
908 
909 	error = prison_check_ip4(td->td_ucred, &addr->sin_addr);
910 	if (error != 0)
911 		return (error);
912 
913 	inp = sotoinpcb(so);
914 	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
915 
916 	if (TAILQ_EMPTY(&V_ifnet) ||
917 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
918 	    (addr->sin_addr.s_addr &&
919 	     (inp->inp_flags & INP_BINDANY) == 0 &&
920 	     ifa_ifwithaddr_check((struct sockaddr *)addr) == 0))
921 		return (EADDRNOTAVAIL);
922 
923 	INP_INFO_WLOCK(&V_ripcbinfo);
924 	INP_WLOCK(inp);
925 	rip_delhash(inp);
926 	inp->inp_laddr = addr->sin_addr;
927 	rip_inshash(inp);
928 	INP_WUNLOCK(inp);
929 	INP_INFO_WUNLOCK(&V_ripcbinfo);
930 	return (0);
931 }
932 
933 static int
934 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
935 {
936 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
937 	struct inpcb *inp;
938 
939 	if (nam->sa_len != sizeof(*addr))
940 		return (EINVAL);
941 	if (TAILQ_EMPTY(&V_ifnet))
942 		return (EADDRNOTAVAIL);
943 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
944 		return (EAFNOSUPPORT);
945 
946 	inp = sotoinpcb(so);
947 	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
948 
949 	INP_INFO_WLOCK(&V_ripcbinfo);
950 	INP_WLOCK(inp);
951 	rip_delhash(inp);
952 	inp->inp_faddr = addr->sin_addr;
953 	rip_inshash(inp);
954 	soisconnected(so);
955 	INP_WUNLOCK(inp);
956 	INP_INFO_WUNLOCK(&V_ripcbinfo);
957 	return (0);
958 }
959 
960 static int
961 rip_shutdown(struct socket *so)
962 {
963 	struct inpcb *inp;
964 
965 	inp = sotoinpcb(so);
966 	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
967 
968 	INP_WLOCK(inp);
969 	socantsendmore(so);
970 	INP_WUNLOCK(inp);
971 	return (0);
972 }
973 
974 static int
975 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
976     struct mbuf *control, struct thread *td)
977 {
978 	struct inpcb *inp;
979 	u_long dst;
980 
981 	inp = sotoinpcb(so);
982 	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
983 
984 	/*
985 	 * Note: 'dst' reads below are unlocked.
986 	 */
987 	if (so->so_state & SS_ISCONNECTED) {
988 		if (nam) {
989 			m_freem(m);
990 			return (EISCONN);
991 		}
992 		dst = inp->inp_faddr.s_addr;	/* Unlocked read. */
993 	} else {
994 		if (nam == NULL) {
995 			m_freem(m);
996 			return (ENOTCONN);
997 		}
998 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
999 	}
1000 	return (rip_output(m, so, dst));
1001 }
1002 #endif /* INET */
1003 
1004 static int
1005 rip_pcblist(SYSCTL_HANDLER_ARGS)
1006 {
1007 	int error, i, n;
1008 	struct inpcb *inp, **inp_list;
1009 	inp_gen_t gencnt;
1010 	struct xinpgen xig;
1011 
1012 	/*
1013 	 * The process of preparing the TCB list is too time-consuming and
1014 	 * resource-intensive to repeat twice on every request.
1015 	 */
1016 	if (req->oldptr == 0) {
1017 		n = V_ripcbinfo.ipi_count;
1018 		n += imax(n / 8, 10);
1019 		req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
1020 		return (0);
1021 	}
1022 
1023 	if (req->newptr != 0)
1024 		return (EPERM);
1025 
1026 	/*
1027 	 * OK, now we're committed to doing something.
1028 	 */
1029 	INP_INFO_RLOCK(&V_ripcbinfo);
1030 	gencnt = V_ripcbinfo.ipi_gencnt;
1031 	n = V_ripcbinfo.ipi_count;
1032 	INP_INFO_RUNLOCK(&V_ripcbinfo);
1033 
1034 	xig.xig_len = sizeof xig;
1035 	xig.xig_count = n;
1036 	xig.xig_gen = gencnt;
1037 	xig.xig_sogen = so_gencnt;
1038 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1039 	if (error)
1040 		return (error);
1041 
1042 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
1043 	if (inp_list == 0)
1044 		return (ENOMEM);
1045 
1046 	INP_INFO_RLOCK(&V_ripcbinfo);
1047 	for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n;
1048 	     inp = LIST_NEXT(inp, inp_list)) {
1049 		INP_WLOCK(inp);
1050 		if (inp->inp_gencnt <= gencnt &&
1051 		    cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
1052 			in_pcbref(inp);
1053 			inp_list[i++] = inp;
1054 		}
1055 		INP_WUNLOCK(inp);
1056 	}
1057 	INP_INFO_RUNLOCK(&V_ripcbinfo);
1058 	n = i;
1059 
1060 	error = 0;
1061 	for (i = 0; i < n; i++) {
1062 		inp = inp_list[i];
1063 		INP_RLOCK(inp);
1064 		if (inp->inp_gencnt <= gencnt) {
1065 			struct xinpcb xi;
1066 
1067 			bzero(&xi, sizeof(xi));
1068 			xi.xi_len = sizeof xi;
1069 			/* XXX should avoid extra copy */
1070 			bcopy(inp, &xi.xi_inp, sizeof *inp);
1071 			if (inp->inp_socket)
1072 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
1073 			INP_RUNLOCK(inp);
1074 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1075 		} else
1076 			INP_RUNLOCK(inp);
1077 	}
1078 	INP_INFO_WLOCK(&V_ripcbinfo);
1079 	for (i = 0; i < n; i++) {
1080 		inp = inp_list[i];
1081 		INP_RLOCK(inp);
1082 		if (!in_pcbrele_rlocked(inp))
1083 			INP_RUNLOCK(inp);
1084 	}
1085 	INP_INFO_WUNLOCK(&V_ripcbinfo);
1086 
1087 	if (!error) {
1088 		/*
1089 		 * Give the user an updated idea of our state.  If the
1090 		 * generation differs from what we told her before, she knows
1091 		 * that something happened while we were processing this
1092 		 * request, and it might be necessary to retry.
1093 		 */
1094 		INP_INFO_RLOCK(&V_ripcbinfo);
1095 		xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1096 		xig.xig_sogen = so_gencnt;
1097 		xig.xig_count = V_ripcbinfo.ipi_count;
1098 		INP_INFO_RUNLOCK(&V_ripcbinfo);
1099 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1100 	}
1101 	free(inp_list, M_TEMP);
1102 	return (error);
1103 }
1104 
1105 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist,
1106     CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0,
1107     rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
1108 
1109 #ifdef INET
1110 struct pr_usrreqs rip_usrreqs = {
1111 	.pru_abort =		rip_abort,
1112 	.pru_attach =		rip_attach,
1113 	.pru_bind =		rip_bind,
1114 	.pru_connect =		rip_connect,
1115 	.pru_control =		in_control,
1116 	.pru_detach =		rip_detach,
1117 	.pru_disconnect =	rip_disconnect,
1118 	.pru_peeraddr =		in_getpeeraddr,
1119 	.pru_send =		rip_send,
1120 	.pru_shutdown =		rip_shutdown,
1121 	.pru_sockaddr =		in_getsockaddr,
1122 	.pru_sosetlabel =	in_pcbsosetlabel,
1123 	.pru_close =		rip_close,
1124 };
1125 #endif /* INET */
1126