xref: /freebsd/sys/netinet/raw_ip.c (revision 94942af266ac119ede0ca836f9aa5a5ac0582938)
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  * $FreeBSD$
31  */
32 
33 #include "opt_inet6.h"
34 #include "opt_ipsec.h"
35 #include "opt_mac.h"
36 
37 #include <sys/param.h>
38 #include <sys/jail.h>
39 #include <sys/kernel.h>
40 #include <sys/lock.h>
41 #include <sys/malloc.h>
42 #include <sys/mbuf.h>
43 #include <sys/priv.h>
44 #include <sys/proc.h>
45 #include <sys/protosw.h>
46 #include <sys/signalvar.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/sx.h>
50 #include <sys/sysctl.h>
51 #include <sys/systm.h>
52 
53 #include <vm/uma.h>
54 
55 #include <net/if.h>
56 #include <net/route.h>
57 
58 #include <netinet/in.h>
59 #include <netinet/in_systm.h>
60 #include <netinet/in_pcb.h>
61 #include <netinet/in_var.h>
62 #include <netinet/ip.h>
63 #include <netinet/ip_var.h>
64 #include <netinet/ip_mroute.h>
65 
66 #include <netinet/ip_fw.h>
67 #include <netinet/ip_dummynet.h>
68 
69 #ifdef FAST_IPSEC
70 #include <netipsec/ipsec.h>
71 #endif /*FAST_IPSEC*/
72 
73 #ifdef IPSEC
74 #include <netinet6/ipsec.h>
75 #endif /*IPSEC*/
76 
77 #include <security/mac/mac_framework.h>
78 
79 struct	inpcbhead ripcb;
80 struct	inpcbinfo ripcbinfo;
81 
82 /* control hooks for ipfw and dummynet */
83 ip_fw_ctl_t *ip_fw_ctl_ptr = NULL;
84 ip_dn_ctl_t *ip_dn_ctl_ptr = NULL;
85 
86 /*
87  * hooks for multicast routing. They all default to NULL,
88  * so leave them not initialized and rely on BSS being set to 0.
89  */
90 
91 /* The socket used to communicate with the multicast routing daemon.  */
92 struct socket  *ip_mrouter;
93 
94 /* The various mrouter and rsvp functions */
95 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
96 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
97 int (*ip_mrouter_done)(void);
98 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
99 		   struct ip_moptions *);
100 int (*mrt_ioctl)(int, caddr_t);
101 int (*legal_vif_num)(int);
102 u_long (*ip_mcast_src)(int);
103 
104 void (*rsvp_input_p)(struct mbuf *m, int off);
105 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
106 void (*ip_rsvp_force_done)(struct socket *);
107 
108 /*
109  * Raw interface to IP protocol.
110  */
111 
112 /*
113  * Initialize raw connection block q.
114  */
115 static void
116 rip_zone_change(void *tag)
117 {
118 
119 	uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets);
120 }
121 
122 static int
123 rip_inpcb_init(void *mem, int size, int flags)
124 {
125 	struct inpcb *inp = mem;
126 
127 	INP_LOCK_INIT(inp, "inp", "rawinp");
128 	return (0);
129 }
130 
131 void
132 rip_init(void)
133 {
134 
135 	INP_INFO_LOCK_INIT(&ripcbinfo, "rip");
136 	LIST_INIT(&ripcb);
137 	ripcbinfo.ipi_listhead = &ripcb;
138 	/*
139 	 * XXX We don't use the hash list for raw IP, but it's easier
140 	 * to allocate a one entry hash list than it is to check all
141 	 * over the place for hashbase == NULL.
142 	 */
143 	ripcbinfo.ipi_hashbase = hashinit(1, M_PCB, &ripcbinfo.ipi_hashmask);
144 	ripcbinfo.ipi_porthashbase = hashinit(1, M_PCB,
145 	    &ripcbinfo.ipi_porthashmask);
146 	ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb),
147 	    NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
148 	uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets);
149 	EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change,
150 		NULL, EVENTHANDLER_PRI_ANY);
151 }
152 
153 static struct	sockaddr_in ripsrc = { sizeof(ripsrc), AF_INET };
154 
155 static int
156 raw_append(struct inpcb *last, struct ip *ip, struct mbuf *n)
157 {
158 	int policyfail = 0;
159 
160 	INP_LOCK_ASSERT(last);
161 
162 #if defined(IPSEC) || defined(FAST_IPSEC)
163 	/* check AH/ESP integrity. */
164 	if (ipsec4_in_reject(n, last)) {
165 		policyfail = 1;
166 #ifdef IPSEC
167 		ipsecstat.in_polvio++;
168 #endif /*IPSEC*/
169 		/* do not inject data to pcb */
170 	}
171 #endif /*IPSEC || FAST_IPSEC*/
172 #ifdef MAC
173 	if (!policyfail && mac_check_inpcb_deliver(last, n) != 0)
174 		policyfail = 1;
175 #endif
176 	/* Check the minimum TTL for socket. */
177 	if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
178 		policyfail = 1;
179 	if (!policyfail) {
180 		struct mbuf *opts = NULL;
181 		struct socket *so;
182 
183 		so = last->inp_socket;
184 		if ((last->inp_flags & INP_CONTROLOPTS) ||
185 		    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
186 			ip_savecontrol(last, &opts, ip, n);
187 		SOCKBUF_LOCK(&so->so_rcv);
188 		if (sbappendaddr_locked(&so->so_rcv,
189 		    (struct sockaddr *)&ripsrc, n, opts) == 0) {
190 			/* should notify about lost packet */
191 			m_freem(n);
192 			if (opts)
193 				m_freem(opts);
194 			SOCKBUF_UNLOCK(&so->so_rcv);
195 		} else
196 			sorwakeup_locked(so);
197 	} else
198 		m_freem(n);
199 	return policyfail;
200 }
201 
202 /*
203  * Setup generic address and protocol structures
204  * for raw_input routine, then pass them along with
205  * mbuf chain.
206  */
207 void
208 rip_input(struct mbuf *m, int off)
209 {
210 	struct ip *ip = mtod(m, struct ip *);
211 	int proto = ip->ip_p;
212 	struct inpcb *inp, *last;
213 
214 	INP_INFO_RLOCK(&ripcbinfo);
215 	ripsrc.sin_addr = ip->ip_src;
216 	last = NULL;
217 	LIST_FOREACH(inp, &ripcb, inp_list) {
218 		INP_LOCK(inp);
219 		if (inp->inp_ip_p && inp->inp_ip_p != proto) {
220 	docontinue:
221 			INP_UNLOCK(inp);
222 			continue;
223 		}
224 #ifdef INET6
225 		if ((inp->inp_vflag & INP_IPV4) == 0)
226 			goto docontinue;
227 #endif
228 		if (inp->inp_laddr.s_addr &&
229 		    inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
230 			goto docontinue;
231 		if (inp->inp_faddr.s_addr &&
232 		    inp->inp_faddr.s_addr != ip->ip_src.s_addr)
233 			goto docontinue;
234 		if (jailed(inp->inp_socket->so_cred))
235 			if (htonl(prison_getip(inp->inp_socket->so_cred)) !=
236 			    ip->ip_dst.s_addr)
237 				goto docontinue;
238 		if (last) {
239 			struct mbuf *n;
240 
241 			n = m_copy(m, 0, (int)M_COPYALL);
242 			if (n != NULL)
243 				(void) raw_append(last, ip, n);
244 			/* XXX count dropped packet */
245 			INP_UNLOCK(last);
246 		}
247 		last = inp;
248 	}
249 	if (last != NULL) {
250 		if (raw_append(last, ip, m) != 0)
251 			ipstat.ips_delivered--;
252 		INP_UNLOCK(last);
253 	} else {
254 		m_freem(m);
255 		ipstat.ips_noproto++;
256 		ipstat.ips_delivered--;
257 	}
258 	INP_INFO_RUNLOCK(&ripcbinfo);
259 }
260 
261 /*
262  * Generate IP header and pass packet to ip_output.
263  * Tack on options user may have setup with control call.
264  */
265 int
266 rip_output(struct mbuf *m, struct socket *so, u_long dst)
267 {
268 	struct ip *ip;
269 	int error;
270 	struct inpcb *inp = sotoinpcb(so);
271 	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
272 	    IP_ALLOWBROADCAST;
273 
274 	/*
275 	 * If the user handed us a complete IP packet, use it.
276 	 * Otherwise, allocate an mbuf for a header and fill it in.
277 	 */
278 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
279 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
280 			m_freem(m);
281 			return(EMSGSIZE);
282 		}
283 		M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
284 		if (m == NULL)
285 			return(ENOBUFS);
286 
287 		INP_LOCK(inp);
288 		ip = mtod(m, struct ip *);
289 		ip->ip_tos = inp->inp_ip_tos;
290 		if (inp->inp_flags & INP_DONTFRAG)
291 			ip->ip_off = IP_DF;
292 		else
293 			ip->ip_off = 0;
294 		ip->ip_p = inp->inp_ip_p;
295 		ip->ip_len = m->m_pkthdr.len;
296 		if (jailed(inp->inp_socket->so_cred))
297 			ip->ip_src.s_addr =
298 			    htonl(prison_getip(inp->inp_socket->so_cred));
299 		else
300 			ip->ip_src = inp->inp_laddr;
301 		ip->ip_dst.s_addr = dst;
302 		ip->ip_ttl = inp->inp_ip_ttl;
303 	} else {
304 		if (m->m_pkthdr.len > IP_MAXPACKET) {
305 			m_freem(m);
306 			return(EMSGSIZE);
307 		}
308 		INP_LOCK(inp);
309 		ip = mtod(m, struct ip *);
310 		if (jailed(inp->inp_socket->so_cred)) {
311 			if (ip->ip_src.s_addr !=
312 			    htonl(prison_getip(inp->inp_socket->so_cred))) {
313 				INP_UNLOCK(inp);
314 				m_freem(m);
315 				return (EPERM);
316 			}
317 		}
318 		/* don't allow both user specified and setsockopt options,
319 		   and don't allow packet length sizes that will crash */
320 		if (((ip->ip_hl != (sizeof (*ip) >> 2))
321 		     && inp->inp_options)
322 		    || (ip->ip_len > m->m_pkthdr.len)
323 		    || (ip->ip_len < (ip->ip_hl << 2))) {
324 			INP_UNLOCK(inp);
325 			m_freem(m);
326 			return EINVAL;
327 		}
328 		if (ip->ip_id == 0)
329 			ip->ip_id = ip_newid();
330 		/* XXX prevent ip_output from overwriting header fields */
331 		flags |= IP_RAWOUTPUT;
332 		ipstat.ips_rawout++;
333 	}
334 
335 	if (inp->inp_flags & INP_ONESBCAST)
336 		flags |= IP_SENDONES;
337 
338 #ifdef MAC
339 	mac_create_mbuf_from_inpcb(inp, m);
340 #endif
341 
342 	error = ip_output(m, inp->inp_options, NULL, flags,
343 	    inp->inp_moptions, inp);
344 	INP_UNLOCK(inp);
345 	return error;
346 }
347 
348 /*
349  * Raw IP socket option processing.
350  *
351  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
352  * only be created by a privileged process, and as such, socket option
353  * operations to manage system properties on any raw socket were allowed to
354  * take place without explicit additional access control checks.  However,
355  * raw sockets can now also be created in jail(), and therefore explicit
356  * checks are now required.  Likewise, raw sockets can be used by a process
357  * after it gives up privilege, so some caution is required.  For options
358  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
359  * performed in ip_ctloutput() and therefore no check occurs here.
360  * Unilaterally checking suser() here breaks normal IP socket option
361  * operations on raw sockets.
362  *
363  * When adding new socket options here, make sure to add access control
364  * checks here as necessary.
365  */
366 int
367 rip_ctloutput(struct socket *so, struct sockopt *sopt)
368 {
369 	struct	inpcb *inp = sotoinpcb(so);
370 	int	error, optval;
371 
372 	if (sopt->sopt_level != IPPROTO_IP)
373 		return (EINVAL);
374 
375 	error = 0;
376 	switch (sopt->sopt_dir) {
377 	case SOPT_GET:
378 		switch (sopt->sopt_name) {
379 		case IP_HDRINCL:
380 			optval = inp->inp_flags & INP_HDRINCL;
381 			error = sooptcopyout(sopt, &optval, sizeof optval);
382 			break;
383 
384 		case IP_FW_ADD:	/* ADD actually returns the body... */
385 		case IP_FW_GET:
386 		case IP_FW_TABLE_GETSIZE:
387 		case IP_FW_TABLE_LIST:
388 		case IP_FW_NAT_GET_CONFIG:
389 		case IP_FW_NAT_GET_LOG:
390 			/*
391 			 * XXXRW: Isn't this checked one layer down?  Yes, it
392 			 * is.
393 			 */
394 			error = priv_check(curthread, PRIV_NETINET_IPFW);
395 			if (error != 0)
396 				return (error);
397 			if (ip_fw_ctl_ptr != NULL)
398 				error = ip_fw_ctl_ptr(sopt);
399 			else
400 				error = ENOPROTOOPT;
401 			break;
402 
403 		case IP_DUMMYNET_GET:
404 			error = priv_check(curthread, PRIV_NETINET_DUMMYNET);
405 			if (error != 0)
406 				return (error);
407 			if (ip_dn_ctl_ptr != NULL)
408 				error = ip_dn_ctl_ptr(sopt);
409 			else
410 				error = ENOPROTOOPT;
411 			break ;
412 
413 		case MRT_INIT:
414 		case MRT_DONE:
415 		case MRT_ADD_VIF:
416 		case MRT_DEL_VIF:
417 		case MRT_ADD_MFC:
418 		case MRT_DEL_MFC:
419 		case MRT_VERSION:
420 		case MRT_ASSERT:
421 		case MRT_API_SUPPORT:
422 		case MRT_API_CONFIG:
423 		case MRT_ADD_BW_UPCALL:
424 		case MRT_DEL_BW_UPCALL:
425 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
426 			if (error != 0)
427 				return (error);
428 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
429 				EOPNOTSUPP;
430 			break;
431 
432 		default:
433 			error = ip_ctloutput(so, sopt);
434 			break;
435 		}
436 		break;
437 
438 	case SOPT_SET:
439 		switch (sopt->sopt_name) {
440 		case IP_HDRINCL:
441 			error = sooptcopyin(sopt, &optval, sizeof optval,
442 					    sizeof optval);
443 			if (error)
444 				break;
445 			if (optval)
446 				inp->inp_flags |= INP_HDRINCL;
447 			else
448 				inp->inp_flags &= ~INP_HDRINCL;
449 			break;
450 
451 		case IP_FW_ADD:
452 		case IP_FW_DEL:
453 		case IP_FW_FLUSH:
454 		case IP_FW_ZERO:
455 		case IP_FW_RESETLOG:
456 		case IP_FW_TABLE_ADD:
457 		case IP_FW_TABLE_DEL:
458 		case IP_FW_TABLE_FLUSH:
459 		case IP_FW_NAT_CFG:
460 		case IP_FW_NAT_DEL:
461 			/*
462 			 * XXXRW: Isn't this checked one layer down?
463 			 */
464 			error = priv_check(curthread, PRIV_NETINET_IPFW);
465 			if (error != 0)
466 				return (error);
467 			if (ip_fw_ctl_ptr != NULL)
468 				error = ip_fw_ctl_ptr(sopt);
469 			else
470 				error = ENOPROTOOPT;
471 			break;
472 
473 		case IP_DUMMYNET_CONFIGURE:
474 		case IP_DUMMYNET_DEL:
475 		case IP_DUMMYNET_FLUSH:
476 			error = priv_check(curthread, PRIV_NETINET_DUMMYNET);
477 			if (error != 0)
478 				return (error);
479 			if (ip_dn_ctl_ptr != NULL)
480 				error = ip_dn_ctl_ptr(sopt);
481 			else
482 				error = ENOPROTOOPT ;
483 			break ;
484 
485 		case IP_RSVP_ON:
486 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
487 			if (error != 0)
488 				return (error);
489 			error = ip_rsvp_init(so);
490 			break;
491 
492 		case IP_RSVP_OFF:
493 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
494 			if (error != 0)
495 				return (error);
496 			error = ip_rsvp_done();
497 			break;
498 
499 		case IP_RSVP_VIF_ON:
500 		case IP_RSVP_VIF_OFF:
501 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
502 			if (error != 0)
503 				return (error);
504 			error = ip_rsvp_vif ?
505 				ip_rsvp_vif(so, sopt) : EINVAL;
506 			break;
507 
508 		case MRT_INIT:
509 		case MRT_DONE:
510 		case MRT_ADD_VIF:
511 		case MRT_DEL_VIF:
512 		case MRT_ADD_MFC:
513 		case MRT_DEL_MFC:
514 		case MRT_VERSION:
515 		case MRT_ASSERT:
516 		case MRT_API_SUPPORT:
517 		case MRT_API_CONFIG:
518 		case MRT_ADD_BW_UPCALL:
519 		case MRT_DEL_BW_UPCALL:
520 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
521 			if (error != 0)
522 				return (error);
523 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
524 					EOPNOTSUPP;
525 			break;
526 
527 		default:
528 			error = ip_ctloutput(so, sopt);
529 			break;
530 		}
531 		break;
532 	}
533 
534 	return (error);
535 }
536 
537 /*
538  * This function exists solely to receive the PRC_IFDOWN messages which
539  * are sent by if_down().  It looks for an ifaddr whose ifa_addr is sa,
540  * and calls in_ifadown() to remove all routes corresponding to that address.
541  * It also receives the PRC_IFUP messages from if_up() and reinstalls the
542  * interface routes.
543  */
544 void
545 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
546 {
547 	struct in_ifaddr *ia;
548 	struct ifnet *ifp;
549 	int err;
550 	int flags;
551 
552 	switch (cmd) {
553 	case PRC_IFDOWN:
554 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
555 			if (ia->ia_ifa.ifa_addr == sa
556 			    && (ia->ia_flags & IFA_ROUTE)) {
557 				/*
558 				 * in_ifscrub kills the interface route.
559 				 */
560 				in_ifscrub(ia->ia_ifp, ia);
561 				/*
562 				 * in_ifadown gets rid of all the rest of
563 				 * the routes.  This is not quite the right
564 				 * thing to do, but at least if we are running
565 				 * a routing process they will come back.
566 				 */
567 				in_ifadown(&ia->ia_ifa, 0);
568 				break;
569 			}
570 		}
571 		break;
572 
573 	case PRC_IFUP:
574 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
575 			if (ia->ia_ifa.ifa_addr == sa)
576 				break;
577 		}
578 		if (ia == 0 || (ia->ia_flags & IFA_ROUTE))
579 			return;
580 		flags = RTF_UP;
581 		ifp = ia->ia_ifa.ifa_ifp;
582 
583 		if ((ifp->if_flags & IFF_LOOPBACK)
584 		    || (ifp->if_flags & IFF_POINTOPOINT))
585 			flags |= RTF_HOST;
586 
587 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
588 		if (err == 0)
589 			ia->ia_flags |= IFA_ROUTE;
590 		break;
591 	}
592 }
593 
594 u_long	rip_sendspace = 9216;
595 u_long	rip_recvspace = 9216;
596 
597 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
598     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
599 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
600     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
601 
602 static int
603 rip_attach(struct socket *so, int proto, struct thread *td)
604 {
605 	struct inpcb *inp;
606 	int error;
607 
608 	inp = sotoinpcb(so);
609 	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
610 	/*
611 	 * XXXRW: Centralize privilege decision in kern_jail.c.
612 	 */
613 	if (jailed(td->td_ucred) && !jail_allow_raw_sockets)
614 		return (EPERM);
615 	error = priv_check_cred(td->td_ucred, PRIV_NETINET_RAW,
616 	    SUSER_ALLOWJAIL);
617 	if (error)
618 		return error;
619 	if (proto >= IPPROTO_MAX || proto < 0)
620 		return EPROTONOSUPPORT;
621 	error = soreserve(so, rip_sendspace, rip_recvspace);
622 	if (error)
623 		return error;
624 	INP_INFO_WLOCK(&ripcbinfo);
625 	error = in_pcballoc(so, &ripcbinfo);
626 	if (error) {
627 		INP_INFO_WUNLOCK(&ripcbinfo);
628 		return error;
629 	}
630 	inp = (struct inpcb *)so->so_pcb;
631 	INP_INFO_WUNLOCK(&ripcbinfo);
632 	inp->inp_vflag |= INP_IPV4;
633 	inp->inp_ip_p = proto;
634 	inp->inp_ip_ttl = ip_defttl;
635 	INP_UNLOCK(inp);
636 	return 0;
637 }
638 
639 static void
640 rip_detach(struct socket *so)
641 {
642 	struct inpcb *inp;
643 
644 	inp = sotoinpcb(so);
645 	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
646 	KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
647 	    ("rip_detach: not closed"));
648 
649 	INP_INFO_WLOCK(&ripcbinfo);
650 	INP_LOCK(inp);
651 	if (so == ip_mrouter && ip_mrouter_done)
652 		ip_mrouter_done();
653 	if (ip_rsvp_force_done)
654 		ip_rsvp_force_done(so);
655 	if (so == ip_rsvpd)
656 		ip_rsvp_done();
657 	in_pcbdetach(inp);
658 	in_pcbfree(inp);
659 	INP_INFO_WUNLOCK(&ripcbinfo);
660 }
661 
662 static void
663 rip_dodisconnect(struct socket *so, struct inpcb *inp)
664 {
665 
666 	INP_LOCK_ASSERT(inp);
667 
668 	inp->inp_faddr.s_addr = INADDR_ANY;
669 	SOCK_LOCK(so);
670 	so->so_state &= ~SS_ISCONNECTED;
671 	SOCK_UNLOCK(so);
672 }
673 
674 static void
675 rip_abort(struct socket *so)
676 {
677 	struct inpcb *inp;
678 
679 	inp = sotoinpcb(so);
680 	KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
681 
682 	INP_INFO_WLOCK(&ripcbinfo);
683 	INP_LOCK(inp);
684 	rip_dodisconnect(so, inp);
685 	INP_UNLOCK(inp);
686 	INP_INFO_WUNLOCK(&ripcbinfo);
687 }
688 
689 static void
690 rip_close(struct socket *so)
691 {
692 	struct inpcb *inp;
693 
694 	inp = sotoinpcb(so);
695 	KASSERT(inp != NULL, ("rip_close: inp == NULL"));
696 
697 	INP_INFO_WLOCK(&ripcbinfo);
698 	INP_LOCK(inp);
699 	rip_dodisconnect(so, inp);
700 	INP_UNLOCK(inp);
701 	INP_INFO_WUNLOCK(&ripcbinfo);
702 }
703 
704 static int
705 rip_disconnect(struct socket *so)
706 {
707 	struct inpcb *inp;
708 
709 	if ((so->so_state & SS_ISCONNECTED) == 0)
710 		return ENOTCONN;
711 
712 	inp = sotoinpcb(so);
713 	KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
714 	INP_INFO_WLOCK(&ripcbinfo);
715 	INP_LOCK(inp);
716 	rip_dodisconnect(so, inp);
717 	INP_UNLOCK(inp);
718 	INP_INFO_WUNLOCK(&ripcbinfo);
719 	return (0);
720 }
721 
722 static int
723 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
724 {
725 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
726 	struct inpcb *inp;
727 
728 	if (nam->sa_len != sizeof(*addr))
729 		return EINVAL;
730 
731 	if (jailed(td->td_ucred)) {
732 		if (addr->sin_addr.s_addr == INADDR_ANY)
733 			addr->sin_addr.s_addr =
734 			    htonl(prison_getip(td->td_ucred));
735 		if (htonl(prison_getip(td->td_ucred)) != addr->sin_addr.s_addr)
736 			return (EADDRNOTAVAIL);
737 	}
738 
739 	if (TAILQ_EMPTY(&ifnet) ||
740 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
741 	    (addr->sin_addr.s_addr &&
742 	     ifa_ifwithaddr((struct sockaddr *)addr) == 0))
743 		return EADDRNOTAVAIL;
744 
745 	inp = sotoinpcb(so);
746 	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
747 	INP_INFO_WLOCK(&ripcbinfo);
748 	INP_LOCK(inp);
749 	inp->inp_laddr = addr->sin_addr;
750 	INP_UNLOCK(inp);
751 	INP_INFO_WUNLOCK(&ripcbinfo);
752 	return 0;
753 }
754 
755 static int
756 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
757 {
758 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
759 	struct inpcb *inp;
760 
761 	if (nam->sa_len != sizeof(*addr))
762 		return EINVAL;
763 	if (TAILQ_EMPTY(&ifnet))
764 		return EADDRNOTAVAIL;
765 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
766 		return EAFNOSUPPORT;
767 
768 	inp = sotoinpcb(so);
769 	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
770 	INP_INFO_WLOCK(&ripcbinfo);
771 	INP_LOCK(inp);
772 	inp->inp_faddr = addr->sin_addr;
773 	soisconnected(so);
774 	INP_UNLOCK(inp);
775 	INP_INFO_WUNLOCK(&ripcbinfo);
776 	return 0;
777 }
778 
779 static int
780 rip_shutdown(struct socket *so)
781 {
782 	struct inpcb *inp;
783 
784 	inp = sotoinpcb(so);
785 	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
786 	INP_LOCK(inp);
787 	socantsendmore(so);
788 	INP_UNLOCK(inp);
789 	return 0;
790 }
791 
792 static int
793 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
794     struct mbuf *control, struct thread *td)
795 {
796 	struct inpcb *inp;
797 	u_long dst;
798 
799 	inp = sotoinpcb(so);
800 	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
801 	/*
802 	 * Note: 'dst' reads below are unlocked.
803 	 */
804 	if (so->so_state & SS_ISCONNECTED) {
805 		if (nam) {
806 			m_freem(m);
807 			return EISCONN;
808 		}
809 		dst = inp->inp_faddr.s_addr;	/* Unlocked read. */
810 	} else {
811 		if (nam == NULL) {
812 			m_freem(m);
813 			return ENOTCONN;
814 		}
815 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
816 	}
817 	return rip_output(m, so, dst);
818 }
819 
820 static int
821 rip_pcblist(SYSCTL_HANDLER_ARGS)
822 {
823 	int error, i, n;
824 	struct inpcb *inp, **inp_list;
825 	inp_gen_t gencnt;
826 	struct xinpgen xig;
827 
828 	/*
829 	 * The process of preparing the TCB list is too time-consuming and
830 	 * resource-intensive to repeat twice on every request.
831 	 */
832 	if (req->oldptr == 0) {
833 		n = ripcbinfo.ipi_count;
834 		req->oldidx = 2 * (sizeof xig)
835 			+ (n + n/8) * sizeof(struct xinpcb);
836 		return 0;
837 	}
838 
839 	if (req->newptr != 0)
840 		return EPERM;
841 
842 	/*
843 	 * OK, now we're committed to doing something.
844 	 */
845 	INP_INFO_RLOCK(&ripcbinfo);
846 	gencnt = ripcbinfo.ipi_gencnt;
847 	n = ripcbinfo.ipi_count;
848 	INP_INFO_RUNLOCK(&ripcbinfo);
849 
850 	xig.xig_len = sizeof xig;
851 	xig.xig_count = n;
852 	xig.xig_gen = gencnt;
853 	xig.xig_sogen = so_gencnt;
854 	error = SYSCTL_OUT(req, &xig, sizeof xig);
855 	if (error)
856 		return error;
857 
858 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
859 	if (inp_list == 0)
860 		return ENOMEM;
861 
862 	INP_INFO_RLOCK(&ripcbinfo);
863 	for (inp = LIST_FIRST(ripcbinfo.ipi_listhead), i = 0; inp && i < n;
864 	     inp = LIST_NEXT(inp, inp_list)) {
865 		INP_LOCK(inp);
866 		if (inp->inp_gencnt <= gencnt &&
867 		    cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) {
868 			/* XXX held references? */
869 			inp_list[i++] = inp;
870 		}
871 		INP_UNLOCK(inp);
872 	}
873 	INP_INFO_RUNLOCK(&ripcbinfo);
874 	n = i;
875 
876 	error = 0;
877 	for (i = 0; i < n; i++) {
878 		inp = inp_list[i];
879 		INP_LOCK(inp);
880 		if (inp->inp_gencnt <= gencnt) {
881 			struct xinpcb xi;
882 			bzero(&xi, sizeof(xi));
883 			xi.xi_len = sizeof xi;
884 			/* XXX should avoid extra copy */
885 			bcopy(inp, &xi.xi_inp, sizeof *inp);
886 			if (inp->inp_socket)
887 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
888 			INP_UNLOCK(inp);
889 			error = SYSCTL_OUT(req, &xi, sizeof xi);
890 		} else
891 			INP_UNLOCK(inp);
892 	}
893 	if (!error) {
894 		/*
895 		 * Give the user an updated idea of our state.
896 		 * If the generation differs from what we told
897 		 * her before, she knows that something happened
898 		 * while we were processing this request, and it
899 		 * might be necessary to retry.
900 		 */
901 		INP_INFO_RLOCK(&ripcbinfo);
902 		xig.xig_gen = ripcbinfo.ipi_gencnt;
903 		xig.xig_sogen = so_gencnt;
904 		xig.xig_count = ripcbinfo.ipi_count;
905 		INP_INFO_RUNLOCK(&ripcbinfo);
906 		error = SYSCTL_OUT(req, &xig, sizeof xig);
907 	}
908 	free(inp_list, M_TEMP);
909 	return error;
910 }
911 
912 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0,
913 	    rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
914 
915 struct pr_usrreqs rip_usrreqs = {
916 	.pru_abort =		rip_abort,
917 	.pru_attach =		rip_attach,
918 	.pru_bind =		rip_bind,
919 	.pru_connect =		rip_connect,
920 	.pru_control =		in_control,
921 	.pru_detach =		rip_detach,
922 	.pru_disconnect =	rip_disconnect,
923 	.pru_peeraddr =		in_getpeeraddr,
924 	.pru_send =		rip_send,
925 	.pru_shutdown =		rip_shutdown,
926 	.pru_sockaddr =		in_getsockaddr,
927 	.pru_sosetlabel =	in_pcbsosetlabel,
928 	.pru_close =		rip_close,
929 };
930