xref: /freebsd/sys/netinet/raw_ip.c (revision 2b743a9e9ddc6736208dc8ca1ce06ce64ad20a19)
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  * Nominal space allocated to a raw ip socket.
110  */
111 #define	RIPSNDQ		8192
112 #define	RIPRCVQ		8192
113 
114 /*
115  * Raw interface to IP protocol.
116  */
117 
118 /*
119  * Initialize raw connection block q.
120  */
121 static void
122 rip_zone_change(void *tag)
123 {
124 
125 	uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets);
126 }
127 
128 static int
129 rip_inpcb_init(void *mem, int size, int flags)
130 {
131 	struct inpcb *inp = mem;
132 
133 	INP_LOCK_INIT(inp, "inp", "rawinp");
134 	return (0);
135 }
136 
137 void
138 rip_init()
139 {
140 	INP_INFO_LOCK_INIT(&ripcbinfo, "rip");
141 	LIST_INIT(&ripcb);
142 	ripcbinfo.listhead = &ripcb;
143 	/*
144 	 * XXX We don't use the hash list for raw IP, but it's easier
145 	 * to allocate a one entry hash list than it is to check all
146 	 * over the place for hashbase == NULL.
147 	 */
148 	ripcbinfo.hashbase = hashinit(1, M_PCB, &ripcbinfo.hashmask);
149 	ripcbinfo.porthashbase = hashinit(1, M_PCB, &ripcbinfo.porthashmask);
150 	ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb),
151 	    NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
152 	uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets);
153 	EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change,
154 		NULL, EVENTHANDLER_PRI_ANY);
155 }
156 
157 static struct	sockaddr_in ripsrc = { sizeof(ripsrc), AF_INET };
158 
159 static int
160 raw_append(struct inpcb *last, struct ip *ip, struct mbuf *n)
161 {
162 	int policyfail = 0;
163 
164 	INP_LOCK_ASSERT(last);
165 
166 #if defined(IPSEC) || defined(FAST_IPSEC)
167 	/* check AH/ESP integrity. */
168 	if (ipsec4_in_reject(n, last)) {
169 		policyfail = 1;
170 #ifdef IPSEC
171 		ipsecstat.in_polvio++;
172 #endif /*IPSEC*/
173 		/* do not inject data to pcb */
174 	}
175 #endif /*IPSEC || FAST_IPSEC*/
176 #ifdef MAC
177 	if (!policyfail && mac_check_inpcb_deliver(last, n) != 0)
178 		policyfail = 1;
179 #endif
180 	/* Check the minimum TTL for socket. */
181 	if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
182 		policyfail = 1;
183 	if (!policyfail) {
184 		struct mbuf *opts = NULL;
185 		struct socket *so;
186 
187 		so = last->inp_socket;
188 		if ((last->inp_flags & INP_CONTROLOPTS) ||
189 		    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
190 			ip_savecontrol(last, &opts, ip, n);
191 		SOCKBUF_LOCK(&so->so_rcv);
192 		if (sbappendaddr_locked(&so->so_rcv,
193 		    (struct sockaddr *)&ripsrc, n, opts) == 0) {
194 			/* should notify about lost packet */
195 			m_freem(n);
196 			if (opts)
197 				m_freem(opts);
198 			SOCKBUF_UNLOCK(&so->so_rcv);
199 		} else
200 			sorwakeup_locked(so);
201 	} else
202 		m_freem(n);
203 	return policyfail;
204 }
205 
206 /*
207  * Setup generic address and protocol structures
208  * for raw_input routine, then pass them along with
209  * mbuf chain.
210  */
211 void
212 rip_input(struct mbuf *m, int off)
213 {
214 	struct ip *ip = mtod(m, struct ip *);
215 	int proto = ip->ip_p;
216 	struct inpcb *inp, *last;
217 
218 	INP_INFO_RLOCK(&ripcbinfo);
219 	ripsrc.sin_addr = ip->ip_src;
220 	last = NULL;
221 	LIST_FOREACH(inp, &ripcb, inp_list) {
222 		INP_LOCK(inp);
223 		if (inp->inp_ip_p && inp->inp_ip_p != proto) {
224 	docontinue:
225 			INP_UNLOCK(inp);
226 			continue;
227 		}
228 #ifdef INET6
229 		if ((inp->inp_vflag & INP_IPV4) == 0)
230 			goto docontinue;
231 #endif
232 		if (inp->inp_laddr.s_addr &&
233 		    inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
234 			goto docontinue;
235 		if (inp->inp_faddr.s_addr &&
236 		    inp->inp_faddr.s_addr != ip->ip_src.s_addr)
237 			goto docontinue;
238 		if (jailed(inp->inp_socket->so_cred))
239 			if (htonl(prison_getip(inp->inp_socket->so_cred)) !=
240 			    ip->ip_dst.s_addr)
241 				goto docontinue;
242 		if (last) {
243 			struct mbuf *n;
244 
245 			n = m_copy(m, 0, (int)M_COPYALL);
246 			if (n != NULL)
247 				(void) raw_append(last, ip, n);
248 			/* XXX count dropped packet */
249 			INP_UNLOCK(last);
250 		}
251 		last = inp;
252 	}
253 	if (last != NULL) {
254 		if (raw_append(last, ip, m) != 0)
255 			ipstat.ips_delivered--;
256 		INP_UNLOCK(last);
257 	} else {
258 		m_freem(m);
259 		ipstat.ips_noproto++;
260 		ipstat.ips_delivered--;
261 	}
262 	INP_INFO_RUNLOCK(&ripcbinfo);
263 }
264 
265 /*
266  * Generate IP header and pass packet to ip_output.
267  * Tack on options user may have setup with control call.
268  */
269 int
270 rip_output(struct mbuf *m, struct socket *so, u_long dst)
271 {
272 	struct ip *ip;
273 	int error;
274 	struct inpcb *inp = sotoinpcb(so);
275 	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
276 	    IP_ALLOWBROADCAST;
277 
278 	/*
279 	 * If the user handed us a complete IP packet, use it.
280 	 * Otherwise, allocate an mbuf for a header and fill it in.
281 	 */
282 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
283 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
284 			m_freem(m);
285 			return(EMSGSIZE);
286 		}
287 		M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
288 		if (m == NULL)
289 			return(ENOBUFS);
290 
291 		INP_LOCK(inp);
292 		ip = mtod(m, struct ip *);
293 		ip->ip_tos = inp->inp_ip_tos;
294 		if (inp->inp_flags & INP_DONTFRAG)
295 			ip->ip_off = IP_DF;
296 		else
297 			ip->ip_off = 0;
298 		ip->ip_p = inp->inp_ip_p;
299 		ip->ip_len = m->m_pkthdr.len;
300 		if (jailed(inp->inp_socket->so_cred))
301 			ip->ip_src.s_addr =
302 			    htonl(prison_getip(inp->inp_socket->so_cred));
303 		else
304 			ip->ip_src = inp->inp_laddr;
305 		ip->ip_dst.s_addr = dst;
306 		ip->ip_ttl = inp->inp_ip_ttl;
307 	} else {
308 		if (m->m_pkthdr.len > IP_MAXPACKET) {
309 			m_freem(m);
310 			return(EMSGSIZE);
311 		}
312 		INP_LOCK(inp);
313 		ip = mtod(m, struct ip *);
314 		if (jailed(inp->inp_socket->so_cred)) {
315 			if (ip->ip_src.s_addr !=
316 			    htonl(prison_getip(inp->inp_socket->so_cred))) {
317 				INP_UNLOCK(inp);
318 				m_freem(m);
319 				return (EPERM);
320 			}
321 		}
322 		/* don't allow both user specified and setsockopt options,
323 		   and don't allow packet length sizes that will crash */
324 		if (((ip->ip_hl != (sizeof (*ip) >> 2))
325 		     && inp->inp_options)
326 		    || (ip->ip_len > m->m_pkthdr.len)
327 		    || (ip->ip_len < (ip->ip_hl << 2))) {
328 			INP_UNLOCK(inp);
329 			m_freem(m);
330 			return EINVAL;
331 		}
332 		if (ip->ip_id == 0)
333 			ip->ip_id = ip_newid();
334 		/* XXX prevent ip_output from overwriting header fields */
335 		flags |= IP_RAWOUTPUT;
336 		ipstat.ips_rawout++;
337 	}
338 
339 	if (inp->inp_flags & INP_ONESBCAST)
340 		flags |= IP_SENDONES;
341 
342 #ifdef MAC
343 	mac_create_mbuf_from_inpcb(inp, m);
344 #endif
345 
346 	error = ip_output(m, inp->inp_options, NULL, flags,
347 	    inp->inp_moptions, inp);
348 	INP_UNLOCK(inp);
349 	return error;
350 }
351 
352 /*
353  * Raw IP socket option processing.
354  *
355  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
356  * only be created by a privileged process, and as such, socket option
357  * operations to manage system properties on any raw socket were allowed to
358  * take place without explicit additional access control checks.  However,
359  * raw sockets can now also be created in jail(), and therefore explicit
360  * checks are now required.  Likewise, raw sockets can be used by a process
361  * after it gives up privilege, so some caution is required.  For options
362  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
363  * performed in ip_ctloutput() and therefore no check occurs here.
364  * Unilaterally checking suser() here breaks normal IP socket option
365  * operations on raw sockets.
366  *
367  * When adding new socket options here, make sure to add access control
368  * checks here as necessary.
369  */
370 int
371 rip_ctloutput(struct socket *so, struct sockopt *sopt)
372 {
373 	struct	inpcb *inp = sotoinpcb(so);
374 	int	error, optval;
375 
376 	if (sopt->sopt_level != IPPROTO_IP)
377 		return (EINVAL);
378 
379 	error = 0;
380 	switch (sopt->sopt_dir) {
381 	case SOPT_GET:
382 		switch (sopt->sopt_name) {
383 		case IP_HDRINCL:
384 			optval = inp->inp_flags & INP_HDRINCL;
385 			error = sooptcopyout(sopt, &optval, sizeof optval);
386 			break;
387 
388 		case IP_FW_ADD:	/* ADD actually returns the body... */
389 		case IP_FW_GET:
390 		case IP_FW_TABLE_GETSIZE:
391 		case IP_FW_TABLE_LIST:
392 		case IP_FW_NAT_GET_CONFIG:
393 		case IP_FW_NAT_GET_LOG:
394 			/*
395 			 * XXXRW: Isn't this checked one layer down?  Yes, it
396 			 * is.
397 			 */
398 			error = priv_check(curthread, PRIV_NETINET_IPFW);
399 			if (error != 0)
400 				return (error);
401 			if (ip_fw_ctl_ptr != NULL)
402 				error = ip_fw_ctl_ptr(sopt);
403 			else
404 				error = ENOPROTOOPT;
405 			break;
406 
407 		case IP_DUMMYNET_GET:
408 			error = priv_check(curthread, PRIV_NETINET_DUMMYNET);
409 			if (error != 0)
410 				return (error);
411 			if (ip_dn_ctl_ptr != NULL)
412 				error = ip_dn_ctl_ptr(sopt);
413 			else
414 				error = ENOPROTOOPT;
415 			break ;
416 
417 		case MRT_INIT:
418 		case MRT_DONE:
419 		case MRT_ADD_VIF:
420 		case MRT_DEL_VIF:
421 		case MRT_ADD_MFC:
422 		case MRT_DEL_MFC:
423 		case MRT_VERSION:
424 		case MRT_ASSERT:
425 		case MRT_API_SUPPORT:
426 		case MRT_API_CONFIG:
427 		case MRT_ADD_BW_UPCALL:
428 		case MRT_DEL_BW_UPCALL:
429 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
430 			if (error != 0)
431 				return (error);
432 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
433 				EOPNOTSUPP;
434 			break;
435 
436 		default:
437 			error = ip_ctloutput(so, sopt);
438 			break;
439 		}
440 		break;
441 
442 	case SOPT_SET:
443 		switch (sopt->sopt_name) {
444 		case IP_HDRINCL:
445 			error = sooptcopyin(sopt, &optval, sizeof optval,
446 					    sizeof optval);
447 			if (error)
448 				break;
449 			if (optval)
450 				inp->inp_flags |= INP_HDRINCL;
451 			else
452 				inp->inp_flags &= ~INP_HDRINCL;
453 			break;
454 
455 		case IP_FW_ADD:
456 		case IP_FW_DEL:
457 		case IP_FW_FLUSH:
458 		case IP_FW_ZERO:
459 		case IP_FW_RESETLOG:
460 		case IP_FW_TABLE_ADD:
461 		case IP_FW_TABLE_DEL:
462 		case IP_FW_TABLE_FLUSH:
463 		case IP_FW_NAT_CFG:
464 		case IP_FW_NAT_DEL:
465 			/*
466 			 * XXXRW: Isn't this checked one layer down?
467 			 */
468 			error = priv_check(curthread, PRIV_NETINET_IPFW);
469 			if (error != 0)
470 				return (error);
471 			if (ip_fw_ctl_ptr != NULL)
472 				error = ip_fw_ctl_ptr(sopt);
473 			else
474 				error = ENOPROTOOPT;
475 			break;
476 
477 		case IP_DUMMYNET_CONFIGURE:
478 		case IP_DUMMYNET_DEL:
479 		case IP_DUMMYNET_FLUSH:
480 			error = priv_check(curthread, PRIV_NETINET_DUMMYNET);
481 			if (error != 0)
482 				return (error);
483 			if (ip_dn_ctl_ptr != NULL)
484 				error = ip_dn_ctl_ptr(sopt);
485 			else
486 				error = ENOPROTOOPT ;
487 			break ;
488 
489 		case IP_RSVP_ON:
490 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
491 			if (error != 0)
492 				return (error);
493 			error = ip_rsvp_init(so);
494 			break;
495 
496 		case IP_RSVP_OFF:
497 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
498 			if (error != 0)
499 				return (error);
500 			error = ip_rsvp_done();
501 			break;
502 
503 		case IP_RSVP_VIF_ON:
504 		case IP_RSVP_VIF_OFF:
505 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
506 			if (error != 0)
507 				return (error);
508 			error = ip_rsvp_vif ?
509 				ip_rsvp_vif(so, sopt) : EINVAL;
510 			break;
511 
512 		case MRT_INIT:
513 		case MRT_DONE:
514 		case MRT_ADD_VIF:
515 		case MRT_DEL_VIF:
516 		case MRT_ADD_MFC:
517 		case MRT_DEL_MFC:
518 		case MRT_VERSION:
519 		case MRT_ASSERT:
520 		case MRT_API_SUPPORT:
521 		case MRT_API_CONFIG:
522 		case MRT_ADD_BW_UPCALL:
523 		case MRT_DEL_BW_UPCALL:
524 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
525 			if (error != 0)
526 				return (error);
527 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
528 					EOPNOTSUPP;
529 			break;
530 
531 		default:
532 			error = ip_ctloutput(so, sopt);
533 			break;
534 		}
535 		break;
536 	}
537 
538 	return (error);
539 }
540 
541 /*
542  * This function exists solely to receive the PRC_IFDOWN messages which
543  * are sent by if_down().  It looks for an ifaddr whose ifa_addr is sa,
544  * and calls in_ifadown() to remove all routes corresponding to that address.
545  * It also receives the PRC_IFUP messages from if_up() and reinstalls the
546  * interface routes.
547  */
548 void
549 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
550 {
551 	struct in_ifaddr *ia;
552 	struct ifnet *ifp;
553 	int err;
554 	int flags;
555 
556 	switch (cmd) {
557 	case PRC_IFDOWN:
558 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
559 			if (ia->ia_ifa.ifa_addr == sa
560 			    && (ia->ia_flags & IFA_ROUTE)) {
561 				/*
562 				 * in_ifscrub kills the interface route.
563 				 */
564 				in_ifscrub(ia->ia_ifp, ia);
565 				/*
566 				 * in_ifadown gets rid of all the rest of
567 				 * the routes.  This is not quite the right
568 				 * thing to do, but at least if we are running
569 				 * a routing process they will come back.
570 				 */
571 				in_ifadown(&ia->ia_ifa, 0);
572 				break;
573 			}
574 		}
575 		break;
576 
577 	case PRC_IFUP:
578 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
579 			if (ia->ia_ifa.ifa_addr == sa)
580 				break;
581 		}
582 		if (ia == 0 || (ia->ia_flags & IFA_ROUTE))
583 			return;
584 		flags = RTF_UP;
585 		ifp = ia->ia_ifa.ifa_ifp;
586 
587 		if ((ifp->if_flags & IFF_LOOPBACK)
588 		    || (ifp->if_flags & IFF_POINTOPOINT))
589 			flags |= RTF_HOST;
590 
591 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
592 		if (err == 0)
593 			ia->ia_flags |= IFA_ROUTE;
594 		break;
595 	}
596 }
597 
598 u_long	rip_sendspace = RIPSNDQ;
599 u_long	rip_recvspace = RIPRCVQ;
600 
601 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
602     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
603 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
604     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
605 
606 static int
607 rip_attach(struct socket *so, int proto, struct thread *td)
608 {
609 	struct inpcb *inp;
610 	int error;
611 
612 	inp = sotoinpcb(so);
613 	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
614 	/*
615 	 * XXXRW: Centralize privilege decision in kern_jail.c.
616 	 */
617 	if (jailed(td->td_ucred) && !jail_allow_raw_sockets)
618 		return (EPERM);
619 	error = priv_check_cred(td->td_ucred, PRIV_NETINET_RAW,
620 	    SUSER_ALLOWJAIL);
621 	if (error)
622 		return error;
623 	if (proto >= IPPROTO_MAX || proto < 0)
624 		return EPROTONOSUPPORT;
625 	error = soreserve(so, rip_sendspace, rip_recvspace);
626 	if (error)
627 		return error;
628 	INP_INFO_WLOCK(&ripcbinfo);
629 	error = in_pcballoc(so, &ripcbinfo);
630 	if (error) {
631 		INP_INFO_WUNLOCK(&ripcbinfo);
632 		return error;
633 	}
634 	inp = (struct inpcb *)so->so_pcb;
635 	INP_INFO_WUNLOCK(&ripcbinfo);
636 	inp->inp_vflag |= INP_IPV4;
637 	inp->inp_ip_p = proto;
638 	inp->inp_ip_ttl = ip_defttl;
639 	INP_UNLOCK(inp);
640 	return 0;
641 }
642 
643 static void
644 rip_detach(struct socket *so)
645 {
646 	struct inpcb *inp;
647 
648 	inp = sotoinpcb(so);
649 	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
650 	KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
651 	    ("rip_detach: not closed"));
652 
653 	INP_INFO_WLOCK(&ripcbinfo);
654 	INP_LOCK(inp);
655 	if (so == ip_mrouter && ip_mrouter_done)
656 		ip_mrouter_done();
657 	if (ip_rsvp_force_done)
658 		ip_rsvp_force_done(so);
659 	if (so == ip_rsvpd)
660 		ip_rsvp_done();
661 	in_pcbdetach(inp);
662 	in_pcbfree(inp);
663 	INP_INFO_WUNLOCK(&ripcbinfo);
664 }
665 
666 static void
667 rip_dodisconnect(struct socket *so, struct inpcb *inp)
668 {
669 
670 	INP_LOCK_ASSERT(inp);
671 
672 	inp->inp_faddr.s_addr = INADDR_ANY;
673 	SOCK_LOCK(so);
674 	so->so_state &= ~SS_ISCONNECTED;
675 	SOCK_UNLOCK(so);
676 }
677 
678 static void
679 rip_abort(struct socket *so)
680 {
681 	struct inpcb *inp;
682 
683 	inp = sotoinpcb(so);
684 	KASSERT(inp != NULL, ("rip_abort: 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 void
694 rip_close(struct socket *so)
695 {
696 	struct inpcb *inp;
697 
698 	inp = sotoinpcb(so);
699 	KASSERT(inp != NULL, ("rip_close: inp == NULL"));
700 
701 	INP_INFO_WLOCK(&ripcbinfo);
702 	INP_LOCK(inp);
703 	rip_dodisconnect(so, inp);
704 	INP_UNLOCK(inp);
705 	INP_INFO_WUNLOCK(&ripcbinfo);
706 }
707 
708 static int
709 rip_disconnect(struct socket *so)
710 {
711 	struct inpcb *inp;
712 
713 	if ((so->so_state & SS_ISCONNECTED) == 0)
714 		return ENOTCONN;
715 
716 	inp = sotoinpcb(so);
717 	KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
718 	INP_INFO_WLOCK(&ripcbinfo);
719 	INP_LOCK(inp);
720 	rip_dodisconnect(so, inp);
721 	INP_UNLOCK(inp);
722 	INP_INFO_WUNLOCK(&ripcbinfo);
723 	return (0);
724 }
725 
726 static int
727 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
728 {
729 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
730 	struct inpcb *inp;
731 
732 	if (nam->sa_len != sizeof(*addr))
733 		return EINVAL;
734 
735 	if (jailed(td->td_ucred)) {
736 		if (addr->sin_addr.s_addr == INADDR_ANY)
737 			addr->sin_addr.s_addr =
738 			    htonl(prison_getip(td->td_ucred));
739 		if (htonl(prison_getip(td->td_ucred)) != addr->sin_addr.s_addr)
740 			return (EADDRNOTAVAIL);
741 	}
742 
743 	if (TAILQ_EMPTY(&ifnet) ||
744 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
745 	    (addr->sin_addr.s_addr &&
746 	     ifa_ifwithaddr((struct sockaddr *)addr) == 0))
747 		return EADDRNOTAVAIL;
748 
749 	inp = sotoinpcb(so);
750 	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
751 	INP_INFO_WLOCK(&ripcbinfo);
752 	INP_LOCK(inp);
753 	inp->inp_laddr = addr->sin_addr;
754 	INP_UNLOCK(inp);
755 	INP_INFO_WUNLOCK(&ripcbinfo);
756 	return 0;
757 }
758 
759 static int
760 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
761 {
762 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
763 	struct inpcb *inp;
764 
765 	if (nam->sa_len != sizeof(*addr))
766 		return EINVAL;
767 	if (TAILQ_EMPTY(&ifnet))
768 		return EADDRNOTAVAIL;
769 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
770 		return EAFNOSUPPORT;
771 
772 	inp = sotoinpcb(so);
773 	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
774 	INP_INFO_WLOCK(&ripcbinfo);
775 	INP_LOCK(inp);
776 	inp->inp_faddr = addr->sin_addr;
777 	soisconnected(so);
778 	INP_UNLOCK(inp);
779 	INP_INFO_WUNLOCK(&ripcbinfo);
780 	return 0;
781 }
782 
783 static int
784 rip_shutdown(struct socket *so)
785 {
786 	struct inpcb *inp;
787 
788 	inp = sotoinpcb(so);
789 	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
790 	INP_LOCK(inp);
791 	socantsendmore(so);
792 	INP_UNLOCK(inp);
793 	return 0;
794 }
795 
796 static int
797 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
798 	 struct mbuf *control, struct thread *td)
799 {
800 	struct inpcb *inp;
801 	u_long dst;
802 
803 	inp = sotoinpcb(so);
804 	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
805 	/*
806 	 * Note: 'dst' reads below are unlocked.
807 	 */
808 	if (so->so_state & SS_ISCONNECTED) {
809 		if (nam) {
810 			m_freem(m);
811 			return EISCONN;
812 		}
813 		dst = inp->inp_faddr.s_addr;	/* Unlocked read. */
814 	} else {
815 		if (nam == NULL) {
816 			m_freem(m);
817 			return ENOTCONN;
818 		}
819 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
820 	}
821 	return rip_output(m, so, dst);
822 }
823 
824 static int
825 rip_pcblist(SYSCTL_HANDLER_ARGS)
826 {
827 	int error, i, n;
828 	struct inpcb *inp, **inp_list;
829 	inp_gen_t gencnt;
830 	struct xinpgen xig;
831 
832 	/*
833 	 * The process of preparing the TCB list is too time-consuming and
834 	 * resource-intensive to repeat twice on every request.
835 	 */
836 	if (req->oldptr == 0) {
837 		n = ripcbinfo.ipi_count;
838 		req->oldidx = 2 * (sizeof xig)
839 			+ (n + n/8) * sizeof(struct xinpcb);
840 		return 0;
841 	}
842 
843 	if (req->newptr != 0)
844 		return EPERM;
845 
846 	/*
847 	 * OK, now we're committed to doing something.
848 	 */
849 	INP_INFO_RLOCK(&ripcbinfo);
850 	gencnt = ripcbinfo.ipi_gencnt;
851 	n = ripcbinfo.ipi_count;
852 	INP_INFO_RUNLOCK(&ripcbinfo);
853 
854 	xig.xig_len = sizeof xig;
855 	xig.xig_count = n;
856 	xig.xig_gen = gencnt;
857 	xig.xig_sogen = so_gencnt;
858 	error = SYSCTL_OUT(req, &xig, sizeof xig);
859 	if (error)
860 		return error;
861 
862 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
863 	if (inp_list == 0)
864 		return ENOMEM;
865 
866 	INP_INFO_RLOCK(&ripcbinfo);
867 	for (inp = LIST_FIRST(ripcbinfo.listhead), i = 0; inp && i < n;
868 	     inp = LIST_NEXT(inp, inp_list)) {
869 		INP_LOCK(inp);
870 		if (inp->inp_gencnt <= gencnt &&
871 		    cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) {
872 			/* XXX held references? */
873 			inp_list[i++] = inp;
874 		}
875 		INP_UNLOCK(inp);
876 	}
877 	INP_INFO_RUNLOCK(&ripcbinfo);
878 	n = i;
879 
880 	error = 0;
881 	for (i = 0; i < n; i++) {
882 		inp = inp_list[i];
883 		INP_LOCK(inp);
884 		if (inp->inp_gencnt <= gencnt) {
885 			struct xinpcb xi;
886 			bzero(&xi, sizeof(xi));
887 			xi.xi_len = sizeof xi;
888 			/* XXX should avoid extra copy */
889 			bcopy(inp, &xi.xi_inp, sizeof *inp);
890 			if (inp->inp_socket)
891 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
892 			INP_UNLOCK(inp);
893 			error = SYSCTL_OUT(req, &xi, sizeof xi);
894 		} else
895 			INP_UNLOCK(inp);
896 	}
897 	if (!error) {
898 		/*
899 		 * Give the user an updated idea of our state.
900 		 * If the generation differs from what we told
901 		 * her before, she knows that something happened
902 		 * while we were processing this request, and it
903 		 * might be necessary to retry.
904 		 */
905 		INP_INFO_RLOCK(&ripcbinfo);
906 		xig.xig_gen = ripcbinfo.ipi_gencnt;
907 		xig.xig_sogen = so_gencnt;
908 		xig.xig_count = ripcbinfo.ipi_count;
909 		INP_INFO_RUNLOCK(&ripcbinfo);
910 		error = SYSCTL_OUT(req, &xig, sizeof xig);
911 	}
912 	free(inp_list, M_TEMP);
913 	return error;
914 }
915 
916 /*
917  * This is the wrapper function for in_setsockaddr.  We just pass down
918  * the pcbinfo for in_setpeeraddr to lock.
919  */
920 static int
921 rip_sockaddr(struct socket *so, struct sockaddr **nam)
922 {
923 	return (in_setsockaddr(so, nam, &ripcbinfo));
924 }
925 
926 /*
927  * This is the wrapper function for in_setpeeraddr.  We just pass down
928  * the pcbinfo for in_setpeeraddr to lock.
929  */
930 static int
931 rip_peeraddr(struct socket *so, struct sockaddr **nam)
932 {
933 	return (in_setpeeraddr(so, nam, &ripcbinfo));
934 }
935 
936 
937 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0,
938 	    rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
939 
940 struct pr_usrreqs rip_usrreqs = {
941 	.pru_abort =		rip_abort,
942 	.pru_attach =		rip_attach,
943 	.pru_bind =		rip_bind,
944 	.pru_connect =		rip_connect,
945 	.pru_control =		in_control,
946 	.pru_detach =		rip_detach,
947 	.pru_disconnect =	rip_disconnect,
948 	.pru_peeraddr =		rip_peeraddr,
949 	.pru_send =		rip_send,
950 	.pru_shutdown =		rip_shutdown,
951 	.pru_sockaddr =		rip_sockaddr,
952 	.pru_sosetlabel =	in_pcbsosetlabel,
953 	.pru_close =		rip_close,
954 };
955