xref: /freebsd/sys/netinet/raw_ip.c (revision 830940567b49bb0c08dfaed40418999e76616909)
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_inet6.h"
37 #include "opt_ipsec.h"
38 
39 #include <sys/param.h>
40 #include <sys/jail.h>
41 #include <sys/kernel.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/protosw.h>
48 #include <sys/rwlock.h>
49 #include <sys/signalvar.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sx.h>
53 #include <sys/sysctl.h>
54 #include <sys/systm.h>
55 
56 #include <vm/uma.h>
57 
58 #include <net/if.h>
59 #include <net/route.h>
60 #include <net/vnet.h>
61 
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/in_pcb.h>
65 #include <netinet/in_var.h>
66 #include <netinet/ip.h>
67 #include <netinet/ip_var.h>
68 #include <netinet/ip_mroute.h>
69 
70 #ifdef IPSEC
71 #include <netipsec/ipsec.h>
72 #endif /*IPSEC*/
73 
74 #include <security/mac/mac_framework.h>
75 
76 VNET_DEFINE(struct inpcbhead, ripcb);
77 VNET_DEFINE(struct inpcbinfo, ripcbinfo);
78 
79 #define	V_ripcb			VNET(ripcb)
80 #define	V_ripcbinfo		VNET(ripcbinfo)
81 
82 /*
83  * Control and data hooks for ipfw and dummynet.
84  * The data hooks are not used here but it is convenient
85  * to keep them all in one place.
86  */
87 int (*ip_fw_ctl_ptr)(struct sockopt *) = NULL;
88 int (*ip_dn_ctl_ptr)(struct sockopt *) = NULL;
89 int (*ip_fw_chk_ptr)(struct ip_fw_args *args) = NULL;
90 int (*ip_dn_io_ptr)(struct mbuf **m, int dir, struct ip_fw_args *fwa) = NULL;
91 
92 /*
93  * Hooks for multicast routing. They all default to NULL, so leave them not
94  * initialized and rely on BSS being set to 0.
95  */
96 
97 /*
98  * The socket used to communicate with the multicast routing daemon.
99  */
100 VNET_DEFINE(struct socket *, ip_mrouter);
101 
102 /*
103  * The various mrouter and rsvp functions.
104  */
105 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
106 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
107 int (*ip_mrouter_done)(void);
108 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
109 		   struct ip_moptions *);
110 int (*mrt_ioctl)(u_long, caddr_t, int);
111 int (*legal_vif_num)(int);
112 u_long (*ip_mcast_src)(int);
113 
114 void (*rsvp_input_p)(struct mbuf *m, int off);
115 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
116 void (*ip_rsvp_force_done)(struct socket *);
117 
118 /*
119  * Hash functions
120  */
121 
122 #define INP_PCBHASH_RAW_SIZE	256
123 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \
124         (((proto) + (laddr) + (faddr)) % (mask) + 1)
125 
126 static void
127 rip_inshash(struct inpcb *inp)
128 {
129 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
130 	struct inpcbhead *pcbhash;
131 	int hash;
132 
133 	INP_INFO_WLOCK_ASSERT(pcbinfo);
134 	INP_WLOCK_ASSERT(inp);
135 
136 	if (inp->inp_ip_p != 0 &&
137 	    inp->inp_laddr.s_addr != INADDR_ANY &&
138 	    inp->inp_faddr.s_addr != INADDR_ANY) {
139 		hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr,
140 		    inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask);
141 	} else
142 		hash = 0;
143 	pcbhash = &pcbinfo->ipi_hashbase[hash];
144 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
145 }
146 
147 static void
148 rip_delhash(struct inpcb *inp)
149 {
150 
151 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
152 	INP_WLOCK_ASSERT(inp);
153 
154 	LIST_REMOVE(inp, inp_hash);
155 }
156 
157 /*
158  * Raw interface to IP protocol.
159  */
160 
161 /*
162  * Initialize raw connection block q.
163  */
164 static void
165 rip_zone_change(void *tag)
166 {
167 
168 	uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets);
169 }
170 
171 static int
172 rip_inpcb_init(void *mem, int size, int flags)
173 {
174 	struct inpcb *inp = mem;
175 
176 	INP_LOCK_INIT(inp, "inp", "rawinp");
177 	return (0);
178 }
179 
180 void
181 rip_init(void)
182 {
183 
184 	INP_INFO_LOCK_INIT(&V_ripcbinfo, "rip");
185 	LIST_INIT(&V_ripcb);
186 #ifdef VIMAGE
187 	V_ripcbinfo.ipi_vnet = curvnet;
188 #endif
189 	V_ripcbinfo.ipi_listhead = &V_ripcb;
190 	V_ripcbinfo.ipi_hashbase =
191 	    hashinit(INP_PCBHASH_RAW_SIZE, M_PCB, &V_ripcbinfo.ipi_hashmask);
192 	V_ripcbinfo.ipi_porthashbase =
193 	    hashinit(1, M_PCB, &V_ripcbinfo.ipi_porthashmask);
194 	V_ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb),
195 	    NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
196 	uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets);
197 	EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL,
198 	    EVENTHANDLER_PRI_ANY);
199 }
200 
201 #ifdef VIMAGE
202 void
203 rip_destroy(void)
204 {
205 
206 	hashdestroy(V_ripcbinfo.ipi_hashbase, M_PCB,
207 	    V_ripcbinfo.ipi_hashmask);
208 	hashdestroy(V_ripcbinfo.ipi_porthashbase, M_PCB,
209 	    V_ripcbinfo.ipi_porthashmask);
210 }
211 #endif
212 
213 static int
214 rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
215     struct sockaddr_in *ripsrc)
216 {
217 	int policyfail = 0;
218 
219 	INP_RLOCK_ASSERT(last);
220 
221 #ifdef IPSEC
222 	/* check AH/ESP integrity. */
223 	if (ipsec4_in_reject(n, last)) {
224 		policyfail = 1;
225 	}
226 #endif /* IPSEC */
227 #ifdef MAC
228 	if (!policyfail && mac_inpcb_check_deliver(last, n) != 0)
229 		policyfail = 1;
230 #endif
231 	/* Check the minimum TTL for socket. */
232 	if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
233 		policyfail = 1;
234 	if (!policyfail) {
235 		struct mbuf *opts = NULL;
236 		struct socket *so;
237 
238 		so = last->inp_socket;
239 		if ((last->inp_flags & INP_CONTROLOPTS) ||
240 		    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
241 			ip_savecontrol(last, &opts, ip, n);
242 		SOCKBUF_LOCK(&so->so_rcv);
243 		if (sbappendaddr_locked(&so->so_rcv,
244 		    (struct sockaddr *)ripsrc, n, opts) == 0) {
245 			/* should notify about lost packet */
246 			m_freem(n);
247 			if (opts)
248 				m_freem(opts);
249 			SOCKBUF_UNLOCK(&so->so_rcv);
250 		} else
251 			sorwakeup_locked(so);
252 	} else
253 		m_freem(n);
254 	return (policyfail);
255 }
256 
257 /*
258  * Setup generic address and protocol structures for raw_input routine, then
259  * pass them along with mbuf chain.
260  */
261 void
262 rip_input(struct mbuf *m, int off)
263 {
264 	struct ifnet *ifp;
265 	struct ip *ip = mtod(m, struct ip *);
266 	int proto = ip->ip_p;
267 	struct inpcb *inp, *last;
268 	struct sockaddr_in ripsrc;
269 	int hash;
270 
271 	bzero(&ripsrc, sizeof(ripsrc));
272 	ripsrc.sin_len = sizeof(ripsrc);
273 	ripsrc.sin_family = AF_INET;
274 	ripsrc.sin_addr = ip->ip_src;
275 	last = NULL;
276 
277 	ifp = m->m_pkthdr.rcvif;
278 
279 	hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr,
280 	    ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
281 	INP_INFO_RLOCK(&V_ripcbinfo);
282 	LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) {
283 		if (inp->inp_ip_p != proto)
284 			continue;
285 #ifdef INET6
286 		/* XXX inp locking */
287 		if ((inp->inp_vflag & INP_IPV4) == 0)
288 			continue;
289 #endif
290 		if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
291 			continue;
292 		if (inp->inp_faddr.s_addr != ip->ip_src.s_addr)
293 			continue;
294 		if (jailed(inp->inp_cred)) {
295 			/*
296 			 * XXX: If faddr was bound to multicast group,
297 			 * jailed raw socket will drop datagram.
298 			 */
299 			if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
300 				continue;
301 		}
302 		if (last != NULL) {
303 			struct mbuf *n;
304 
305 			n = m_copy(m, 0, (int)M_COPYALL);
306 			if (n != NULL)
307 		    	    (void) rip_append(last, ip, n, &ripsrc);
308 			/* XXX count dropped packet */
309 			INP_RUNLOCK(last);
310 		}
311 		INP_RLOCK(inp);
312 		last = inp;
313 	}
314 	LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) {
315 		if (inp->inp_ip_p && inp->inp_ip_p != proto)
316 			continue;
317 #ifdef INET6
318 		/* XXX inp locking */
319 		if ((inp->inp_vflag & INP_IPV4) == 0)
320 			continue;
321 #endif
322 		if (!in_nullhost(inp->inp_laddr) &&
323 		    !in_hosteq(inp->inp_laddr, ip->ip_dst))
324 			continue;
325 		if (!in_nullhost(inp->inp_faddr) &&
326 		    !in_hosteq(inp->inp_faddr, ip->ip_src))
327 			continue;
328 		if (jailed(inp->inp_cred)) {
329 			/*
330 			 * Allow raw socket in jail to receive multicast;
331 			 * assume process had PRIV_NETINET_RAW at attach,
332 			 * and fall through into normal filter path if so.
333 			 */
334 			if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
335 			    prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
336 				continue;
337 		}
338 		/*
339 		 * If this raw socket has multicast state, and we
340 		 * have received a multicast, check if this socket
341 		 * should receive it, as multicast filtering is now
342 		 * the responsibility of the transport layer.
343 		 */
344 		if (inp->inp_moptions != NULL &&
345 		    IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
346 			struct sockaddr_in group;
347 			int blocked;
348 
349 			bzero(&group, sizeof(struct sockaddr_in));
350 			group.sin_len = sizeof(struct sockaddr_in);
351 			group.sin_family = AF_INET;
352 			group.sin_addr = ip->ip_dst;
353 
354 			blocked = imo_multi_filter(inp->inp_moptions, ifp,
355 			    (struct sockaddr *)&group,
356 			    (struct sockaddr *)&ripsrc);
357 			if (blocked != MCAST_PASS) {
358 				IPSTAT_INC(ips_notmember);
359 				continue;
360 			}
361 		}
362 		if (last != NULL) {
363 			struct mbuf *n;
364 
365 			n = m_copy(m, 0, (int)M_COPYALL);
366 			if (n != NULL)
367 				(void) rip_append(last, ip, n, &ripsrc);
368 			/* XXX count dropped packet */
369 			INP_RUNLOCK(last);
370 		}
371 		INP_RLOCK(inp);
372 		last = inp;
373 	}
374 	INP_INFO_RUNLOCK(&V_ripcbinfo);
375 	if (last != NULL) {
376 		if (rip_append(last, ip, m, &ripsrc) != 0)
377 			IPSTAT_INC(ips_delivered);
378 		INP_RUNLOCK(last);
379 	} else {
380 		m_freem(m);
381 		IPSTAT_INC(ips_noproto);
382 		IPSTAT_DEC(ips_delivered);
383 	}
384 }
385 
386 /*
387  * Generate IP header and pass packet to ip_output.  Tack on options user may
388  * have setup with control call.
389  */
390 int
391 rip_output(struct mbuf *m, struct socket *so, u_long dst)
392 {
393 	struct ip *ip;
394 	int error;
395 	struct inpcb *inp = sotoinpcb(so);
396 	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
397 	    IP_ALLOWBROADCAST;
398 
399 	/*
400 	 * If the user handed us a complete IP packet, use it.  Otherwise,
401 	 * allocate an mbuf for a header and fill it in.
402 	 */
403 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
404 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
405 			m_freem(m);
406 			return(EMSGSIZE);
407 		}
408 		M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
409 		if (m == NULL)
410 			return(ENOBUFS);
411 
412 		INP_RLOCK(inp);
413 		ip = mtod(m, struct ip *);
414 		ip->ip_tos = inp->inp_ip_tos;
415 		if (inp->inp_flags & INP_DONTFRAG)
416 			ip->ip_off = IP_DF;
417 		else
418 			ip->ip_off = 0;
419 		ip->ip_p = inp->inp_ip_p;
420 		ip->ip_len = m->m_pkthdr.len;
421 		ip->ip_src = inp->inp_laddr;
422 		error = prison_get_ip4(inp->inp_cred, &ip->ip_src);
423 		if (error != 0) {
424 			INP_RUNLOCK(inp);
425 			m_freem(m);
426 			return (error);
427 		}
428 		ip->ip_dst.s_addr = dst;
429 		ip->ip_ttl = inp->inp_ip_ttl;
430 	} else {
431 		if (m->m_pkthdr.len > IP_MAXPACKET) {
432 			m_freem(m);
433 			return(EMSGSIZE);
434 		}
435 		INP_RLOCK(inp);
436 		ip = mtod(m, struct ip *);
437 		error = prison_check_ip4(inp->inp_cred, &ip->ip_src);
438 		if (error != 0) {
439 			INP_RUNLOCK(inp);
440 			m_freem(m);
441 			return (error);
442 		}
443 
444 		/*
445 		 * Don't allow both user specified and setsockopt options,
446 		 * and don't allow packet length sizes that will crash.
447 		 */
448 		if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options)
449 		    || (ip->ip_len > m->m_pkthdr.len)
450 		    || (ip->ip_len < (ip->ip_hl << 2))) {
451 			INP_RUNLOCK(inp);
452 			m_freem(m);
453 			return (EINVAL);
454 		}
455 		if (ip->ip_id == 0)
456 			ip->ip_id = ip_newid();
457 
458 		/*
459 		 * XXX prevent ip_output from overwriting header fields.
460 		 */
461 		flags |= IP_RAWOUTPUT;
462 		IPSTAT_INC(ips_rawout);
463 	}
464 
465 	if (inp->inp_flags & INP_ONESBCAST)
466 		flags |= IP_SENDONES;
467 
468 #ifdef MAC
469 	mac_inpcb_create_mbuf(inp, m);
470 #endif
471 
472 	error = ip_output(m, inp->inp_options, NULL, flags,
473 	    inp->inp_moptions, inp);
474 	INP_RUNLOCK(inp);
475 	return (error);
476 }
477 
478 /*
479  * Raw IP socket option processing.
480  *
481  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
482  * only be created by a privileged process, and as such, socket option
483  * operations to manage system properties on any raw socket were allowed to
484  * take place without explicit additional access control checks.  However,
485  * raw sockets can now also be created in jail(), and therefore explicit
486  * checks are now required.  Likewise, raw sockets can be used by a process
487  * after it gives up privilege, so some caution is required.  For options
488  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
489  * performed in ip_ctloutput() and therefore no check occurs here.
490  * Unilaterally checking priv_check() here breaks normal IP socket option
491  * operations on raw sockets.
492  *
493  * When adding new socket options here, make sure to add access control
494  * checks here as necessary.
495  */
496 int
497 rip_ctloutput(struct socket *so, struct sockopt *sopt)
498 {
499 	struct	inpcb *inp = sotoinpcb(so);
500 	int	error, optval;
501 
502 	if (sopt->sopt_level != IPPROTO_IP) {
503 		if ((sopt->sopt_level == SOL_SOCKET) &&
504 		    (sopt->sopt_name == SO_SETFIB)) {
505 			inp->inp_inc.inc_fibnum = so->so_fibnum;
506 			return (0);
507 		}
508 		return (EINVAL);
509 	}
510 
511 	error = 0;
512 	switch (sopt->sopt_dir) {
513 	case SOPT_GET:
514 		switch (sopt->sopt_name) {
515 		case IP_HDRINCL:
516 			optval = inp->inp_flags & INP_HDRINCL;
517 			error = sooptcopyout(sopt, &optval, sizeof optval);
518 			break;
519 
520 		case IP_FW_ADD:	/* ADD actually returns the body... */
521 		case IP_FW_GET:
522 		case IP_FW_TABLE_GETSIZE:
523 		case IP_FW_TABLE_LIST:
524 		case IP_FW_NAT_GET_CONFIG:
525 		case IP_FW_NAT_GET_LOG:
526 			if (ip_fw_ctl_ptr != NULL)
527 				error = ip_fw_ctl_ptr(sopt);
528 			else
529 				error = ENOPROTOOPT;
530 			break;
531 
532 		case IP_DUMMYNET_GET:
533 			if (ip_dn_ctl_ptr != NULL)
534 				error = ip_dn_ctl_ptr(sopt);
535 			else
536 				error = ENOPROTOOPT;
537 			break ;
538 
539 		case MRT_INIT:
540 		case MRT_DONE:
541 		case MRT_ADD_VIF:
542 		case MRT_DEL_VIF:
543 		case MRT_ADD_MFC:
544 		case MRT_DEL_MFC:
545 		case MRT_VERSION:
546 		case MRT_ASSERT:
547 		case MRT_API_SUPPORT:
548 		case MRT_API_CONFIG:
549 		case MRT_ADD_BW_UPCALL:
550 		case MRT_DEL_BW_UPCALL:
551 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
552 			if (error != 0)
553 				return (error);
554 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
555 				EOPNOTSUPP;
556 			break;
557 
558 		default:
559 			error = ip_ctloutput(so, sopt);
560 			break;
561 		}
562 		break;
563 
564 	case SOPT_SET:
565 		switch (sopt->sopt_name) {
566 		case IP_HDRINCL:
567 			error = sooptcopyin(sopt, &optval, sizeof optval,
568 					    sizeof optval);
569 			if (error)
570 				break;
571 			if (optval)
572 				inp->inp_flags |= INP_HDRINCL;
573 			else
574 				inp->inp_flags &= ~INP_HDRINCL;
575 			break;
576 
577 		case IP_FW_ADD:
578 		case IP_FW_DEL:
579 		case IP_FW_FLUSH:
580 		case IP_FW_ZERO:
581 		case IP_FW_RESETLOG:
582 		case IP_FW_TABLE_ADD:
583 		case IP_FW_TABLE_DEL:
584 		case IP_FW_TABLE_FLUSH:
585 		case IP_FW_NAT_CFG:
586 		case IP_FW_NAT_DEL:
587 			if (ip_fw_ctl_ptr != NULL)
588 				error = ip_fw_ctl_ptr(sopt);
589 			else
590 				error = ENOPROTOOPT;
591 			break;
592 
593 		case IP_DUMMYNET_CONFIGURE:
594 		case IP_DUMMYNET_DEL:
595 		case IP_DUMMYNET_FLUSH:
596 			if (ip_dn_ctl_ptr != NULL)
597 				error = ip_dn_ctl_ptr(sopt);
598 			else
599 				error = ENOPROTOOPT ;
600 			break ;
601 
602 		case IP_RSVP_ON:
603 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
604 			if (error != 0)
605 				return (error);
606 			error = ip_rsvp_init(so);
607 			break;
608 
609 		case IP_RSVP_OFF:
610 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
611 			if (error != 0)
612 				return (error);
613 			error = ip_rsvp_done();
614 			break;
615 
616 		case IP_RSVP_VIF_ON:
617 		case IP_RSVP_VIF_OFF:
618 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
619 			if (error != 0)
620 				return (error);
621 			error = ip_rsvp_vif ?
622 				ip_rsvp_vif(so, sopt) : EINVAL;
623 			break;
624 
625 		case MRT_INIT:
626 		case MRT_DONE:
627 		case MRT_ADD_VIF:
628 		case MRT_DEL_VIF:
629 		case MRT_ADD_MFC:
630 		case MRT_DEL_MFC:
631 		case MRT_VERSION:
632 		case MRT_ASSERT:
633 		case MRT_API_SUPPORT:
634 		case MRT_API_CONFIG:
635 		case MRT_ADD_BW_UPCALL:
636 		case MRT_DEL_BW_UPCALL:
637 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
638 			if (error != 0)
639 				return (error);
640 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
641 					EOPNOTSUPP;
642 			break;
643 
644 		default:
645 			error = ip_ctloutput(so, sopt);
646 			break;
647 		}
648 		break;
649 	}
650 
651 	return (error);
652 }
653 
654 /*
655  * This function exists solely to receive the PRC_IFDOWN messages which are
656  * sent by if_down().  It looks for an ifaddr whose ifa_addr is sa, and calls
657  * in_ifadown() to remove all routes corresponding to that address.  It also
658  * receives the PRC_IFUP messages from if_up() and reinstalls the interface
659  * routes.
660  */
661 void
662 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
663 {
664 	struct in_ifaddr *ia;
665 	struct ifnet *ifp;
666 	int err;
667 	int flags;
668 
669 	switch (cmd) {
670 	case PRC_IFDOWN:
671 		IN_IFADDR_RLOCK();
672 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
673 			if (ia->ia_ifa.ifa_addr == sa
674 			    && (ia->ia_flags & IFA_ROUTE)) {
675 				ifa_ref(&ia->ia_ifa);
676 				IN_IFADDR_RUNLOCK();
677 				/*
678 				 * in_ifscrub kills the interface route.
679 				 */
680 				in_ifscrub(ia->ia_ifp, ia);
681 				/*
682 				 * in_ifadown gets rid of all the rest of the
683 				 * routes.  This is not quite the right thing
684 				 * to do, but at least if we are running a
685 				 * routing process they will come back.
686 				 */
687 				in_ifadown(&ia->ia_ifa, 0);
688 				ifa_free(&ia->ia_ifa);
689 				break;
690 			}
691 		}
692 		if (ia == NULL)		/* If ia matched, already unlocked. */
693 			IN_IFADDR_RUNLOCK();
694 		break;
695 
696 	case PRC_IFUP:
697 		IN_IFADDR_RLOCK();
698 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
699 			if (ia->ia_ifa.ifa_addr == sa)
700 				break;
701 		}
702 		if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) {
703 			IN_IFADDR_RUNLOCK();
704 			return;
705 		}
706 		ifa_ref(&ia->ia_ifa);
707 		IN_IFADDR_RUNLOCK();
708 		flags = RTF_UP;
709 		ifp = ia->ia_ifa.ifa_ifp;
710 
711 		if ((ifp->if_flags & IFF_LOOPBACK)
712 		    || (ifp->if_flags & IFF_POINTOPOINT))
713 			flags |= RTF_HOST;
714 
715 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
716 		if (err == 0)
717 			ia->ia_flags |= IFA_ROUTE;
718 		ifa_free(&ia->ia_ifa);
719 		break;
720 	}
721 }
722 
723 u_long	rip_sendspace = 9216;
724 u_long	rip_recvspace = 9216;
725 
726 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
727     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
728 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
729     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
730 
731 static int
732 rip_attach(struct socket *so, int proto, struct thread *td)
733 {
734 	struct inpcb *inp;
735 	int error;
736 
737 	inp = sotoinpcb(so);
738 	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
739 
740 	error = priv_check(td, PRIV_NETINET_RAW);
741 	if (error)
742 		return (error);
743 	if (proto >= IPPROTO_MAX || proto < 0)
744 		return EPROTONOSUPPORT;
745 	error = soreserve(so, rip_sendspace, rip_recvspace);
746 	if (error)
747 		return (error);
748 	INP_INFO_WLOCK(&V_ripcbinfo);
749 	error = in_pcballoc(so, &V_ripcbinfo);
750 	if (error) {
751 		INP_INFO_WUNLOCK(&V_ripcbinfo);
752 		return (error);
753 	}
754 	inp = (struct inpcb *)so->so_pcb;
755 	inp->inp_vflag |= INP_IPV4;
756 	inp->inp_ip_p = proto;
757 	inp->inp_ip_ttl = V_ip_defttl;
758 	rip_inshash(inp);
759 	INP_INFO_WUNLOCK(&V_ripcbinfo);
760 	INP_WUNLOCK(inp);
761 	return (0);
762 }
763 
764 static void
765 rip_detach(struct socket *so)
766 {
767 	struct inpcb *inp;
768 
769 	inp = sotoinpcb(so);
770 	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
771 	KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
772 	    ("rip_detach: not closed"));
773 
774 	INP_INFO_WLOCK(&V_ripcbinfo);
775 	INP_WLOCK(inp);
776 	rip_delhash(inp);
777 	if (so == V_ip_mrouter && ip_mrouter_done)
778 		ip_mrouter_done();
779 	if (ip_rsvp_force_done)
780 		ip_rsvp_force_done(so);
781 	if (so == V_ip_rsvpd)
782 		ip_rsvp_done();
783 	in_pcbdetach(inp);
784 	in_pcbfree(inp);
785 	INP_INFO_WUNLOCK(&V_ripcbinfo);
786 }
787 
788 static void
789 rip_dodisconnect(struct socket *so, struct inpcb *inp)
790 {
791 
792 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
793 	INP_WLOCK_ASSERT(inp);
794 
795 	rip_delhash(inp);
796 	inp->inp_faddr.s_addr = INADDR_ANY;
797 	rip_inshash(inp);
798 	SOCK_LOCK(so);
799 	so->so_state &= ~SS_ISCONNECTED;
800 	SOCK_UNLOCK(so);
801 }
802 
803 static void
804 rip_abort(struct socket *so)
805 {
806 	struct inpcb *inp;
807 
808 	inp = sotoinpcb(so);
809 	KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
810 
811 	INP_INFO_WLOCK(&V_ripcbinfo);
812 	INP_WLOCK(inp);
813 	rip_dodisconnect(so, inp);
814 	INP_WUNLOCK(inp);
815 	INP_INFO_WUNLOCK(&V_ripcbinfo);
816 }
817 
818 static void
819 rip_close(struct socket *so)
820 {
821 	struct inpcb *inp;
822 
823 	inp = sotoinpcb(so);
824 	KASSERT(inp != NULL, ("rip_close: inp == NULL"));
825 
826 	INP_INFO_WLOCK(&V_ripcbinfo);
827 	INP_WLOCK(inp);
828 	rip_dodisconnect(so, inp);
829 	INP_WUNLOCK(inp);
830 	INP_INFO_WUNLOCK(&V_ripcbinfo);
831 }
832 
833 static int
834 rip_disconnect(struct socket *so)
835 {
836 	struct inpcb *inp;
837 
838 	if ((so->so_state & SS_ISCONNECTED) == 0)
839 		return (ENOTCONN);
840 
841 	inp = sotoinpcb(so);
842 	KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
843 
844 	INP_INFO_WLOCK(&V_ripcbinfo);
845 	INP_WLOCK(inp);
846 	rip_dodisconnect(so, inp);
847 	INP_WUNLOCK(inp);
848 	INP_INFO_WUNLOCK(&V_ripcbinfo);
849 	return (0);
850 }
851 
852 static int
853 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
854 {
855 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
856 	struct inpcb *inp;
857 	int error;
858 
859 	if (nam->sa_len != sizeof(*addr))
860 		return (EINVAL);
861 
862 	error = prison_check_ip4(td->td_ucred, &addr->sin_addr);
863 	if (error != 0)
864 		return (error);
865 
866 	inp = sotoinpcb(so);
867 	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
868 
869 	if (TAILQ_EMPTY(&V_ifnet) ||
870 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
871 	    (addr->sin_addr.s_addr &&
872 	     (inp->inp_flags & INP_BINDANY) == 0 &&
873 	     ifa_ifwithaddr_check((struct sockaddr *)addr) == 0))
874 		return (EADDRNOTAVAIL);
875 
876 	INP_INFO_WLOCK(&V_ripcbinfo);
877 	INP_WLOCK(inp);
878 	rip_delhash(inp);
879 	inp->inp_laddr = addr->sin_addr;
880 	rip_inshash(inp);
881 	INP_WUNLOCK(inp);
882 	INP_INFO_WUNLOCK(&V_ripcbinfo);
883 	return (0);
884 }
885 
886 static int
887 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
888 {
889 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
890 	struct inpcb *inp;
891 
892 	if (nam->sa_len != sizeof(*addr))
893 		return (EINVAL);
894 	if (TAILQ_EMPTY(&V_ifnet))
895 		return (EADDRNOTAVAIL);
896 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
897 		return (EAFNOSUPPORT);
898 
899 	inp = sotoinpcb(so);
900 	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
901 
902 	INP_INFO_WLOCK(&V_ripcbinfo);
903 	INP_WLOCK(inp);
904 	rip_delhash(inp);
905 	inp->inp_faddr = addr->sin_addr;
906 	rip_inshash(inp);
907 	soisconnected(so);
908 	INP_WUNLOCK(inp);
909 	INP_INFO_WUNLOCK(&V_ripcbinfo);
910 	return (0);
911 }
912 
913 static int
914 rip_shutdown(struct socket *so)
915 {
916 	struct inpcb *inp;
917 
918 	inp = sotoinpcb(so);
919 	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
920 
921 	INP_WLOCK(inp);
922 	socantsendmore(so);
923 	INP_WUNLOCK(inp);
924 	return (0);
925 }
926 
927 static int
928 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
929     struct mbuf *control, struct thread *td)
930 {
931 	struct inpcb *inp;
932 	u_long dst;
933 
934 	inp = sotoinpcb(so);
935 	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
936 
937 	/*
938 	 * Note: 'dst' reads below are unlocked.
939 	 */
940 	if (so->so_state & SS_ISCONNECTED) {
941 		if (nam) {
942 			m_freem(m);
943 			return (EISCONN);
944 		}
945 		dst = inp->inp_faddr.s_addr;	/* Unlocked read. */
946 	} else {
947 		if (nam == NULL) {
948 			m_freem(m);
949 			return (ENOTCONN);
950 		}
951 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
952 	}
953 	return (rip_output(m, so, dst));
954 }
955 
956 static int
957 rip_pcblist(SYSCTL_HANDLER_ARGS)
958 {
959 	int error, i, n;
960 	struct inpcb *inp, **inp_list;
961 	inp_gen_t gencnt;
962 	struct xinpgen xig;
963 
964 	/*
965 	 * The process of preparing the TCB list is too time-consuming and
966 	 * resource-intensive to repeat twice on every request.
967 	 */
968 	if (req->oldptr == 0) {
969 		n = V_ripcbinfo.ipi_count;
970 		req->oldidx = 2 * (sizeof xig)
971 		    + (n + n/8) * sizeof(struct xinpcb);
972 		return (0);
973 	}
974 
975 	if (req->newptr != 0)
976 		return (EPERM);
977 
978 	/*
979 	 * OK, now we're committed to doing something.
980 	 */
981 	INP_INFO_RLOCK(&V_ripcbinfo);
982 	gencnt = V_ripcbinfo.ipi_gencnt;
983 	n = V_ripcbinfo.ipi_count;
984 	INP_INFO_RUNLOCK(&V_ripcbinfo);
985 
986 	xig.xig_len = sizeof xig;
987 	xig.xig_count = n;
988 	xig.xig_gen = gencnt;
989 	xig.xig_sogen = so_gencnt;
990 	error = SYSCTL_OUT(req, &xig, sizeof xig);
991 	if (error)
992 		return (error);
993 
994 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
995 	if (inp_list == 0)
996 		return (ENOMEM);
997 
998 	INP_INFO_RLOCK(&V_ripcbinfo);
999 	for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n;
1000 	     inp = LIST_NEXT(inp, inp_list)) {
1001 		INP_RLOCK(inp);
1002 		if (inp->inp_gencnt <= gencnt &&
1003 		    cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
1004 			/* XXX held references? */
1005 			inp_list[i++] = inp;
1006 		}
1007 		INP_RUNLOCK(inp);
1008 	}
1009 	INP_INFO_RUNLOCK(&V_ripcbinfo);
1010 	n = i;
1011 
1012 	error = 0;
1013 	for (i = 0; i < n; i++) {
1014 		inp = inp_list[i];
1015 		INP_RLOCK(inp);
1016 		if (inp->inp_gencnt <= gencnt) {
1017 			struct xinpcb xi;
1018 
1019 			bzero(&xi, sizeof(xi));
1020 			xi.xi_len = sizeof xi;
1021 			/* XXX should avoid extra copy */
1022 			bcopy(inp, &xi.xi_inp, sizeof *inp);
1023 			if (inp->inp_socket)
1024 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
1025 			INP_RUNLOCK(inp);
1026 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1027 		} else
1028 			INP_RUNLOCK(inp);
1029 	}
1030 	if (!error) {
1031 		/*
1032 		 * Give the user an updated idea of our state.  If the
1033 		 * generation differs from what we told her before, she knows
1034 		 * that something happened while we were processing this
1035 		 * request, and it might be necessary to retry.
1036 		 */
1037 		INP_INFO_RLOCK(&V_ripcbinfo);
1038 		xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1039 		xig.xig_sogen = so_gencnt;
1040 		xig.xig_count = V_ripcbinfo.ipi_count;
1041 		INP_INFO_RUNLOCK(&V_ripcbinfo);
1042 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1043 	}
1044 	free(inp_list, M_TEMP);
1045 	return (error);
1046 }
1047 
1048 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0,
1049     rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
1050 
1051 struct pr_usrreqs rip_usrreqs = {
1052 	.pru_abort =		rip_abort,
1053 	.pru_attach =		rip_attach,
1054 	.pru_bind =		rip_bind,
1055 	.pru_connect =		rip_connect,
1056 	.pru_control =		in_control,
1057 	.pru_detach =		rip_detach,
1058 	.pru_disconnect =	rip_disconnect,
1059 	.pru_peeraddr =		in_getpeeraddr,
1060 	.pru_send =		rip_send,
1061 	.pru_shutdown =		rip_shutdown,
1062 	.pru_sockaddr =		in_getsockaddr,
1063 	.pru_sosetlabel =	in_pcbsosetlabel,
1064 	.pru_close =		rip_close,
1065 };
1066