xref: /freebsd/sys/netinet/raw_ip.c (revision 1669d8afc64812c8d2d1d147ae1fd42ff441e1b1)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)raw_ip.c	8.7 (Berkeley) 5/15/95
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet6.h"
36 #include "opt_ipsec.h"
37 #include "opt_mac.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/signalvar.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/sx.h>
52 #include <sys/sysctl.h>
53 #include <sys/systm.h>
54 
55 #include <vm/uma.h>
56 
57 #include <net/if.h>
58 #include <net/route.h>
59 
60 #include <netinet/in.h>
61 #include <netinet/in_systm.h>
62 #include <netinet/in_pcb.h>
63 #include <netinet/in_var.h>
64 #include <netinet/ip.h>
65 #include <netinet/ip_var.h>
66 #include <netinet/ip_mroute.h>
67 
68 #include <netinet/ip_fw.h>
69 #include <netinet/ip_dummynet.h>
70 
71 #ifdef IPSEC
72 #include <netipsec/ipsec.h>
73 #endif /*IPSEC*/
74 
75 #include <security/mac/mac_framework.h>
76 
77 struct	inpcbhead ripcb;
78 struct	inpcbinfo ripcbinfo;
79 
80 /* control hooks for ipfw and dummynet */
81 ip_fw_ctl_t *ip_fw_ctl_ptr = NULL;
82 ip_dn_ctl_t *ip_dn_ctl_ptr = NULL;
83 
84 /*
85  * hooks for multicast routing. They all default to NULL,
86  * so leave them not initialized and rely on BSS being set to 0.
87  */
88 
89 /* The socket used to communicate with the multicast routing daemon.  */
90 struct socket  *ip_mrouter;
91 
92 /* The various mrouter and rsvp functions */
93 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
94 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
95 int (*ip_mrouter_done)(void);
96 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
97 		   struct ip_moptions *);
98 int (*mrt_ioctl)(int, caddr_t);
99 int (*legal_vif_num)(int);
100 u_long (*ip_mcast_src)(int);
101 
102 void (*rsvp_input_p)(struct mbuf *m, int off);
103 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
104 void (*ip_rsvp_force_done)(struct socket *);
105 
106 /*
107  * Raw interface to IP protocol.
108  */
109 
110 /*
111  * Initialize raw connection block q.
112  */
113 static void
114 rip_zone_change(void *tag)
115 {
116 
117 	uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets);
118 }
119 
120 static int
121 rip_inpcb_init(void *mem, int size, int flags)
122 {
123 	struct inpcb *inp = mem;
124 
125 	INP_LOCK_INIT(inp, "inp", "rawinp");
126 	return (0);
127 }
128 
129 void
130 rip_init(void)
131 {
132 
133 	INP_INFO_LOCK_INIT(&ripcbinfo, "rip");
134 	LIST_INIT(&ripcb);
135 	ripcbinfo.ipi_listhead = &ripcb;
136 	/*
137 	 * XXX We don't use the hash list for raw IP, but it's easier
138 	 * to allocate a one entry hash list than it is to check all
139 	 * over the place for hashbase == NULL.
140 	 */
141 	ripcbinfo.ipi_hashbase = hashinit(1, M_PCB, &ripcbinfo.ipi_hashmask);
142 	ripcbinfo.ipi_porthashbase = hashinit(1, M_PCB,
143 	    &ripcbinfo.ipi_porthashmask);
144 	ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb),
145 	    NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
146 	uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets);
147 	EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change,
148 		NULL, EVENTHANDLER_PRI_ANY);
149 }
150 
151 static struct	sockaddr_in ripsrc = { sizeof(ripsrc), AF_INET };
152 
153 static int
154 raw_append(struct inpcb *last, struct ip *ip, struct mbuf *n)
155 {
156 	int policyfail = 0;
157 
158 	INP_LOCK_ASSERT(last);
159 
160 #ifdef IPSEC
161 	/* check AH/ESP integrity. */
162 	if (ipsec4_in_reject(n, last)) {
163 		policyfail = 1;
164 	}
165 #endif /* IPSEC */
166 #ifdef MAC
167 	if (!policyfail && mac_inpcb_check_deliver(last, n) != 0)
168 		policyfail = 1;
169 #endif
170 	/* Check the minimum TTL for socket. */
171 	if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
172 		policyfail = 1;
173 	if (!policyfail) {
174 		struct mbuf *opts = NULL;
175 		struct socket *so;
176 
177 		so = last->inp_socket;
178 		if ((last->inp_flags & INP_CONTROLOPTS) ||
179 		    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
180 			ip_savecontrol(last, &opts, ip, n);
181 		SOCKBUF_LOCK(&so->so_rcv);
182 		if (sbappendaddr_locked(&so->so_rcv,
183 		    (struct sockaddr *)&ripsrc, n, opts) == 0) {
184 			/* should notify about lost packet */
185 			m_freem(n);
186 			if (opts)
187 				m_freem(opts);
188 			SOCKBUF_UNLOCK(&so->so_rcv);
189 		} else
190 			sorwakeup_locked(so);
191 	} else
192 		m_freem(n);
193 	return policyfail;
194 }
195 
196 /*
197  * Setup generic address and protocol structures
198  * for raw_input routine, then pass them along with
199  * mbuf chain.
200  */
201 void
202 rip_input(struct mbuf *m, int off)
203 {
204 	struct ip *ip = mtod(m, struct ip *);
205 	int proto = ip->ip_p;
206 	struct inpcb *inp, *last;
207 
208 	INP_INFO_RLOCK(&ripcbinfo);
209 	ripsrc.sin_addr = ip->ip_src;
210 	last = NULL;
211 	LIST_FOREACH(inp, &ripcb, inp_list) {
212 		INP_LOCK(inp);
213 		if (inp->inp_ip_p && inp->inp_ip_p != proto) {
214 	docontinue:
215 			INP_UNLOCK(inp);
216 			continue;
217 		}
218 #ifdef INET6
219 		if ((inp->inp_vflag & INP_IPV4) == 0)
220 			goto docontinue;
221 #endif
222 		if (inp->inp_laddr.s_addr &&
223 		    inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
224 			goto docontinue;
225 		if (inp->inp_faddr.s_addr &&
226 		    inp->inp_faddr.s_addr != ip->ip_src.s_addr)
227 			goto docontinue;
228 		if (jailed(inp->inp_socket->so_cred))
229 			if (htonl(prison_getip(inp->inp_socket->so_cred)) !=
230 			    ip->ip_dst.s_addr)
231 				goto docontinue;
232 		if (last) {
233 			struct mbuf *n;
234 
235 			n = m_copy(m, 0, (int)M_COPYALL);
236 			if (n != NULL)
237 				(void) raw_append(last, ip, n);
238 			/* XXX count dropped packet */
239 			INP_UNLOCK(last);
240 		}
241 		last = inp;
242 	}
243 	if (last != NULL) {
244 		if (raw_append(last, ip, m) != 0)
245 			ipstat.ips_delivered--;
246 		INP_UNLOCK(last);
247 	} else {
248 		m_freem(m);
249 		ipstat.ips_noproto++;
250 		ipstat.ips_delivered--;
251 	}
252 	INP_INFO_RUNLOCK(&ripcbinfo);
253 }
254 
255 /*
256  * Generate IP header and pass packet to ip_output.
257  * Tack on options user may have setup with control call.
258  */
259 int
260 rip_output(struct mbuf *m, struct socket *so, u_long dst)
261 {
262 	struct ip *ip;
263 	int error;
264 	struct inpcb *inp = sotoinpcb(so);
265 	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
266 	    IP_ALLOWBROADCAST;
267 
268 	/*
269 	 * If the user handed us a complete IP packet, use it.
270 	 * Otherwise, allocate an mbuf for a header and fill it in.
271 	 */
272 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
273 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
274 			m_freem(m);
275 			return(EMSGSIZE);
276 		}
277 		M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
278 		if (m == NULL)
279 			return(ENOBUFS);
280 
281 		INP_LOCK(inp);
282 		ip = mtod(m, struct ip *);
283 		ip->ip_tos = inp->inp_ip_tos;
284 		if (inp->inp_flags & INP_DONTFRAG)
285 			ip->ip_off = IP_DF;
286 		else
287 			ip->ip_off = 0;
288 		ip->ip_p = inp->inp_ip_p;
289 		ip->ip_len = m->m_pkthdr.len;
290 		if (jailed(inp->inp_socket->so_cred))
291 			ip->ip_src.s_addr =
292 			    htonl(prison_getip(inp->inp_socket->so_cred));
293 		else
294 			ip->ip_src = inp->inp_laddr;
295 		ip->ip_dst.s_addr = dst;
296 		ip->ip_ttl = inp->inp_ip_ttl;
297 	} else {
298 		if (m->m_pkthdr.len > IP_MAXPACKET) {
299 			m_freem(m);
300 			return(EMSGSIZE);
301 		}
302 		INP_LOCK(inp);
303 		ip = mtod(m, struct ip *);
304 		if (jailed(inp->inp_socket->so_cred)) {
305 			if (ip->ip_src.s_addr !=
306 			    htonl(prison_getip(inp->inp_socket->so_cred))) {
307 				INP_UNLOCK(inp);
308 				m_freem(m);
309 				return (EPERM);
310 			}
311 		}
312 		/* don't allow both user specified and setsockopt options,
313 		   and don't allow packet length sizes that will crash */
314 		if (((ip->ip_hl != (sizeof (*ip) >> 2))
315 		     && inp->inp_options)
316 		    || (ip->ip_len > m->m_pkthdr.len)
317 		    || (ip->ip_len < (ip->ip_hl << 2))) {
318 			INP_UNLOCK(inp);
319 			m_freem(m);
320 			return EINVAL;
321 		}
322 		if (ip->ip_id == 0)
323 			ip->ip_id = ip_newid();
324 		/* XXX prevent ip_output from overwriting header fields */
325 		flags |= IP_RAWOUTPUT;
326 		ipstat.ips_rawout++;
327 	}
328 
329 	if (inp->inp_flags & INP_ONESBCAST)
330 		flags |= IP_SENDONES;
331 
332 #ifdef MAC
333 	mac_inpcb_create_mbuf(inp, m);
334 #endif
335 
336 	error = ip_output(m, inp->inp_options, NULL, flags,
337 	    inp->inp_moptions, inp);
338 	INP_UNLOCK(inp);
339 	return error;
340 }
341 
342 /*
343  * Raw IP socket option processing.
344  *
345  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
346  * only be created by a privileged process, and as such, socket option
347  * operations to manage system properties on any raw socket were allowed to
348  * take place without explicit additional access control checks.  However,
349  * raw sockets can now also be created in jail(), and therefore explicit
350  * checks are now required.  Likewise, raw sockets can be used by a process
351  * after it gives up privilege, so some caution is required.  For options
352  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
353  * performed in ip_ctloutput() and therefore no check occurs here.
354  * Unilaterally checking priv_check() here breaks normal IP socket option
355  * operations on raw sockets.
356  *
357  * When adding new socket options here, make sure to add access control
358  * checks here as necessary.
359  */
360 int
361 rip_ctloutput(struct socket *so, struct sockopt *sopt)
362 {
363 	struct	inpcb *inp = sotoinpcb(so);
364 	int	error, optval;
365 
366 	if (sopt->sopt_level != IPPROTO_IP)
367 		return (EINVAL);
368 
369 	error = 0;
370 	switch (sopt->sopt_dir) {
371 	case SOPT_GET:
372 		switch (sopt->sopt_name) {
373 		case IP_HDRINCL:
374 			optval = inp->inp_flags & INP_HDRINCL;
375 			error = sooptcopyout(sopt, &optval, sizeof optval);
376 			break;
377 
378 		case IP_FW_ADD:	/* ADD actually returns the body... */
379 		case IP_FW_GET:
380 		case IP_FW_TABLE_GETSIZE:
381 		case IP_FW_TABLE_LIST:
382 		case IP_FW_NAT_GET_CONFIG:
383 		case IP_FW_NAT_GET_LOG:
384 			/*
385 			 * XXXRW: Isn't this checked one layer down?  Yes, it
386 			 * is.
387 			 */
388 			error = priv_check(curthread, PRIV_NETINET_IPFW);
389 			if (error != 0)
390 				return (error);
391 			if (ip_fw_ctl_ptr != NULL)
392 				error = ip_fw_ctl_ptr(sopt);
393 			else
394 				error = ENOPROTOOPT;
395 			break;
396 
397 		case IP_DUMMYNET_GET:
398 			error = priv_check(curthread, PRIV_NETINET_DUMMYNET);
399 			if (error != 0)
400 				return (error);
401 			if (ip_dn_ctl_ptr != NULL)
402 				error = ip_dn_ctl_ptr(sopt);
403 			else
404 				error = ENOPROTOOPT;
405 			break ;
406 
407 		case MRT_INIT:
408 		case MRT_DONE:
409 		case MRT_ADD_VIF:
410 		case MRT_DEL_VIF:
411 		case MRT_ADD_MFC:
412 		case MRT_DEL_MFC:
413 		case MRT_VERSION:
414 		case MRT_ASSERT:
415 		case MRT_API_SUPPORT:
416 		case MRT_API_CONFIG:
417 		case MRT_ADD_BW_UPCALL:
418 		case MRT_DEL_BW_UPCALL:
419 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
420 			if (error != 0)
421 				return (error);
422 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
423 				EOPNOTSUPP;
424 			break;
425 
426 		default:
427 			error = ip_ctloutput(so, sopt);
428 			break;
429 		}
430 		break;
431 
432 	case SOPT_SET:
433 		switch (sopt->sopt_name) {
434 		case IP_HDRINCL:
435 			error = sooptcopyin(sopt, &optval, sizeof optval,
436 					    sizeof optval);
437 			if (error)
438 				break;
439 			if (optval)
440 				inp->inp_flags |= INP_HDRINCL;
441 			else
442 				inp->inp_flags &= ~INP_HDRINCL;
443 			break;
444 
445 		case IP_FW_ADD:
446 		case IP_FW_DEL:
447 		case IP_FW_FLUSH:
448 		case IP_FW_ZERO:
449 		case IP_FW_RESETLOG:
450 		case IP_FW_TABLE_ADD:
451 		case IP_FW_TABLE_DEL:
452 		case IP_FW_TABLE_FLUSH:
453 		case IP_FW_NAT_CFG:
454 		case IP_FW_NAT_DEL:
455 			/*
456 			 * XXXRW: Isn't this checked one layer down?
457 			 */
458 			error = priv_check(curthread, PRIV_NETINET_IPFW);
459 			if (error != 0)
460 				return (error);
461 			if (ip_fw_ctl_ptr != NULL)
462 				error = ip_fw_ctl_ptr(sopt);
463 			else
464 				error = ENOPROTOOPT;
465 			break;
466 
467 		case IP_DUMMYNET_CONFIGURE:
468 		case IP_DUMMYNET_DEL:
469 		case IP_DUMMYNET_FLUSH:
470 			error = priv_check(curthread, PRIV_NETINET_DUMMYNET);
471 			if (error != 0)
472 				return (error);
473 			if (ip_dn_ctl_ptr != NULL)
474 				error = ip_dn_ctl_ptr(sopt);
475 			else
476 				error = ENOPROTOOPT ;
477 			break ;
478 
479 		case IP_RSVP_ON:
480 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
481 			if (error != 0)
482 				return (error);
483 			error = ip_rsvp_init(so);
484 			break;
485 
486 		case IP_RSVP_OFF:
487 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
488 			if (error != 0)
489 				return (error);
490 			error = ip_rsvp_done();
491 			break;
492 
493 		case IP_RSVP_VIF_ON:
494 		case IP_RSVP_VIF_OFF:
495 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
496 			if (error != 0)
497 				return (error);
498 			error = ip_rsvp_vif ?
499 				ip_rsvp_vif(so, sopt) : EINVAL;
500 			break;
501 
502 		case MRT_INIT:
503 		case MRT_DONE:
504 		case MRT_ADD_VIF:
505 		case MRT_DEL_VIF:
506 		case MRT_ADD_MFC:
507 		case MRT_DEL_MFC:
508 		case MRT_VERSION:
509 		case MRT_ASSERT:
510 		case MRT_API_SUPPORT:
511 		case MRT_API_CONFIG:
512 		case MRT_ADD_BW_UPCALL:
513 		case MRT_DEL_BW_UPCALL:
514 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
515 			if (error != 0)
516 				return (error);
517 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
518 					EOPNOTSUPP;
519 			break;
520 
521 		default:
522 			error = ip_ctloutput(so, sopt);
523 			break;
524 		}
525 		break;
526 	}
527 
528 	return (error);
529 }
530 
531 /*
532  * This function exists solely to receive the PRC_IFDOWN messages which
533  * are sent by if_down().  It looks for an ifaddr whose ifa_addr is sa,
534  * and calls in_ifadown() to remove all routes corresponding to that address.
535  * It also receives the PRC_IFUP messages from if_up() and reinstalls the
536  * interface routes.
537  */
538 void
539 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
540 {
541 	struct in_ifaddr *ia;
542 	struct ifnet *ifp;
543 	int err;
544 	int flags;
545 
546 	switch (cmd) {
547 	case PRC_IFDOWN:
548 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
549 			if (ia->ia_ifa.ifa_addr == sa
550 			    && (ia->ia_flags & IFA_ROUTE)) {
551 				/*
552 				 * in_ifscrub kills the interface route.
553 				 */
554 				in_ifscrub(ia->ia_ifp, ia);
555 				/*
556 				 * in_ifadown gets rid of all the rest of
557 				 * the routes.  This is not quite the right
558 				 * thing to do, but at least if we are running
559 				 * a routing process they will come back.
560 				 */
561 				in_ifadown(&ia->ia_ifa, 0);
562 				break;
563 			}
564 		}
565 		break;
566 
567 	case PRC_IFUP:
568 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
569 			if (ia->ia_ifa.ifa_addr == sa)
570 				break;
571 		}
572 		if (ia == 0 || (ia->ia_flags & IFA_ROUTE))
573 			return;
574 		flags = RTF_UP;
575 		ifp = ia->ia_ifa.ifa_ifp;
576 
577 		if ((ifp->if_flags & IFF_LOOPBACK)
578 		    || (ifp->if_flags & IFF_POINTOPOINT))
579 			flags |= RTF_HOST;
580 
581 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
582 		if (err == 0)
583 			ia->ia_flags |= IFA_ROUTE;
584 		break;
585 	}
586 }
587 
588 u_long	rip_sendspace = 9216;
589 u_long	rip_recvspace = 9216;
590 
591 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
592     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
593 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
594     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
595 
596 static int
597 rip_attach(struct socket *so, int proto, struct thread *td)
598 {
599 	struct inpcb *inp;
600 	int error;
601 
602 	inp = sotoinpcb(so);
603 	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
604 
605 	error = priv_check(td, PRIV_NETINET_RAW);
606 	if (error)
607 		return error;
608 	if (proto >= IPPROTO_MAX || proto < 0)
609 		return EPROTONOSUPPORT;
610 	error = soreserve(so, rip_sendspace, rip_recvspace);
611 	if (error)
612 		return error;
613 	INP_INFO_WLOCK(&ripcbinfo);
614 	error = in_pcballoc(so, &ripcbinfo);
615 	if (error) {
616 		INP_INFO_WUNLOCK(&ripcbinfo);
617 		return error;
618 	}
619 	inp = (struct inpcb *)so->so_pcb;
620 	INP_INFO_WUNLOCK(&ripcbinfo);
621 	inp->inp_vflag |= INP_IPV4;
622 	inp->inp_ip_p = proto;
623 	inp->inp_ip_ttl = ip_defttl;
624 	INP_UNLOCK(inp);
625 	return 0;
626 }
627 
628 static void
629 rip_detach(struct socket *so)
630 {
631 	struct inpcb *inp;
632 
633 	inp = sotoinpcb(so);
634 	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
635 	KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
636 	    ("rip_detach: not closed"));
637 
638 	INP_INFO_WLOCK(&ripcbinfo);
639 	INP_LOCK(inp);
640 	if (so == ip_mrouter && ip_mrouter_done)
641 		ip_mrouter_done();
642 	if (ip_rsvp_force_done)
643 		ip_rsvp_force_done(so);
644 	if (so == ip_rsvpd)
645 		ip_rsvp_done();
646 	in_pcbdetach(inp);
647 	in_pcbfree(inp);
648 	INP_INFO_WUNLOCK(&ripcbinfo);
649 }
650 
651 static void
652 rip_dodisconnect(struct socket *so, struct inpcb *inp)
653 {
654 
655 	INP_LOCK_ASSERT(inp);
656 
657 	inp->inp_faddr.s_addr = INADDR_ANY;
658 	SOCK_LOCK(so);
659 	so->so_state &= ~SS_ISCONNECTED;
660 	SOCK_UNLOCK(so);
661 }
662 
663 static void
664 rip_abort(struct socket *so)
665 {
666 	struct inpcb *inp;
667 
668 	inp = sotoinpcb(so);
669 	KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
670 
671 	INP_INFO_WLOCK(&ripcbinfo);
672 	INP_LOCK(inp);
673 	rip_dodisconnect(so, inp);
674 	INP_UNLOCK(inp);
675 	INP_INFO_WUNLOCK(&ripcbinfo);
676 }
677 
678 static void
679 rip_close(struct socket *so)
680 {
681 	struct inpcb *inp;
682 
683 	inp = sotoinpcb(so);
684 	KASSERT(inp != NULL, ("rip_close: inp == NULL"));
685 
686 	INP_INFO_WLOCK(&ripcbinfo);
687 	INP_LOCK(inp);
688 	rip_dodisconnect(so, inp);
689 	INP_UNLOCK(inp);
690 	INP_INFO_WUNLOCK(&ripcbinfo);
691 }
692 
693 static int
694 rip_disconnect(struct socket *so)
695 {
696 	struct inpcb *inp;
697 
698 	if ((so->so_state & SS_ISCONNECTED) == 0)
699 		return ENOTCONN;
700 
701 	inp = sotoinpcb(so);
702 	KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
703 	INP_INFO_WLOCK(&ripcbinfo);
704 	INP_LOCK(inp);
705 	rip_dodisconnect(so, inp);
706 	INP_UNLOCK(inp);
707 	INP_INFO_WUNLOCK(&ripcbinfo);
708 	return (0);
709 }
710 
711 static int
712 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
713 {
714 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
715 	struct inpcb *inp;
716 
717 	if (nam->sa_len != sizeof(*addr))
718 		return EINVAL;
719 
720 	if (jailed(td->td_ucred)) {
721 		if (addr->sin_addr.s_addr == INADDR_ANY)
722 			addr->sin_addr.s_addr =
723 			    htonl(prison_getip(td->td_ucred));
724 		if (htonl(prison_getip(td->td_ucred)) != addr->sin_addr.s_addr)
725 			return (EADDRNOTAVAIL);
726 	}
727 
728 	if (TAILQ_EMPTY(&ifnet) ||
729 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
730 	    (addr->sin_addr.s_addr &&
731 	     ifa_ifwithaddr((struct sockaddr *)addr) == 0))
732 		return EADDRNOTAVAIL;
733 
734 	inp = sotoinpcb(so);
735 	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
736 	INP_INFO_WLOCK(&ripcbinfo);
737 	INP_LOCK(inp);
738 	inp->inp_laddr = addr->sin_addr;
739 	INP_UNLOCK(inp);
740 	INP_INFO_WUNLOCK(&ripcbinfo);
741 	return 0;
742 }
743 
744 static int
745 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
746 {
747 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
748 	struct inpcb *inp;
749 
750 	if (nam->sa_len != sizeof(*addr))
751 		return EINVAL;
752 	if (TAILQ_EMPTY(&ifnet))
753 		return EADDRNOTAVAIL;
754 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
755 		return EAFNOSUPPORT;
756 
757 	inp = sotoinpcb(so);
758 	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
759 	INP_INFO_WLOCK(&ripcbinfo);
760 	INP_LOCK(inp);
761 	inp->inp_faddr = addr->sin_addr;
762 	soisconnected(so);
763 	INP_UNLOCK(inp);
764 	INP_INFO_WUNLOCK(&ripcbinfo);
765 	return 0;
766 }
767 
768 static int
769 rip_shutdown(struct socket *so)
770 {
771 	struct inpcb *inp;
772 
773 	inp = sotoinpcb(so);
774 	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
775 	INP_LOCK(inp);
776 	socantsendmore(so);
777 	INP_UNLOCK(inp);
778 	return 0;
779 }
780 
781 static int
782 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
783     struct mbuf *control, struct thread *td)
784 {
785 	struct inpcb *inp;
786 	u_long dst;
787 
788 	inp = sotoinpcb(so);
789 	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
790 	/*
791 	 * Note: 'dst' reads below are unlocked.
792 	 */
793 	if (so->so_state & SS_ISCONNECTED) {
794 		if (nam) {
795 			m_freem(m);
796 			return EISCONN;
797 		}
798 		dst = inp->inp_faddr.s_addr;	/* Unlocked read. */
799 	} else {
800 		if (nam == NULL) {
801 			m_freem(m);
802 			return ENOTCONN;
803 		}
804 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
805 	}
806 	return rip_output(m, so, dst);
807 }
808 
809 static int
810 rip_pcblist(SYSCTL_HANDLER_ARGS)
811 {
812 	int error, i, n;
813 	struct inpcb *inp, **inp_list;
814 	inp_gen_t gencnt;
815 	struct xinpgen xig;
816 
817 	/*
818 	 * The process of preparing the TCB list is too time-consuming and
819 	 * resource-intensive to repeat twice on every request.
820 	 */
821 	if (req->oldptr == 0) {
822 		n = ripcbinfo.ipi_count;
823 		req->oldidx = 2 * (sizeof xig)
824 			+ (n + n/8) * sizeof(struct xinpcb);
825 		return 0;
826 	}
827 
828 	if (req->newptr != 0)
829 		return EPERM;
830 
831 	/*
832 	 * OK, now we're committed to doing something.
833 	 */
834 	INP_INFO_RLOCK(&ripcbinfo);
835 	gencnt = ripcbinfo.ipi_gencnt;
836 	n = ripcbinfo.ipi_count;
837 	INP_INFO_RUNLOCK(&ripcbinfo);
838 
839 	xig.xig_len = sizeof xig;
840 	xig.xig_count = n;
841 	xig.xig_gen = gencnt;
842 	xig.xig_sogen = so_gencnt;
843 	error = SYSCTL_OUT(req, &xig, sizeof xig);
844 	if (error)
845 		return error;
846 
847 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
848 	if (inp_list == 0)
849 		return ENOMEM;
850 
851 	INP_INFO_RLOCK(&ripcbinfo);
852 	for (inp = LIST_FIRST(ripcbinfo.ipi_listhead), i = 0; inp && i < n;
853 	     inp = LIST_NEXT(inp, inp_list)) {
854 		INP_LOCK(inp);
855 		if (inp->inp_gencnt <= gencnt &&
856 		    cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) {
857 			/* XXX held references? */
858 			inp_list[i++] = inp;
859 		}
860 		INP_UNLOCK(inp);
861 	}
862 	INP_INFO_RUNLOCK(&ripcbinfo);
863 	n = i;
864 
865 	error = 0;
866 	for (i = 0; i < n; i++) {
867 		inp = inp_list[i];
868 		INP_LOCK(inp);
869 		if (inp->inp_gencnt <= gencnt) {
870 			struct xinpcb xi;
871 			bzero(&xi, sizeof(xi));
872 			xi.xi_len = sizeof xi;
873 			/* XXX should avoid extra copy */
874 			bcopy(inp, &xi.xi_inp, sizeof *inp);
875 			if (inp->inp_socket)
876 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
877 			INP_UNLOCK(inp);
878 			error = SYSCTL_OUT(req, &xi, sizeof xi);
879 		} else
880 			INP_UNLOCK(inp);
881 	}
882 	if (!error) {
883 		/*
884 		 * Give the user an updated idea of our state.
885 		 * If the generation differs from what we told
886 		 * her before, she knows that something happened
887 		 * while we were processing this request, and it
888 		 * might be necessary to retry.
889 		 */
890 		INP_INFO_RLOCK(&ripcbinfo);
891 		xig.xig_gen = ripcbinfo.ipi_gencnt;
892 		xig.xig_sogen = so_gencnt;
893 		xig.xig_count = ripcbinfo.ipi_count;
894 		INP_INFO_RUNLOCK(&ripcbinfo);
895 		error = SYSCTL_OUT(req, &xig, sizeof xig);
896 	}
897 	free(inp_list, M_TEMP);
898 	return error;
899 }
900 
901 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0,
902 	    rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
903 
904 struct pr_usrreqs rip_usrreqs = {
905 	.pru_abort =		rip_abort,
906 	.pru_attach =		rip_attach,
907 	.pru_bind =		rip_bind,
908 	.pru_connect =		rip_connect,
909 	.pru_control =		in_control,
910 	.pru_detach =		rip_detach,
911 	.pru_disconnect =	rip_disconnect,
912 	.pru_peeraddr =		in_getpeeraddr,
913 	.pru_send =		rip_send,
914 	.pru_shutdown =		rip_shutdown,
915 	.pru_sockaddr =		in_getsockaddr,
916 	.pru_sosetlabel =	in_pcbsosetlabel,
917 	.pru_close =		rip_close,
918 };
919