xref: /freebsd/sys/netinet/raw_ip.c (revision f0a75d274af375d15b97b830966b99a02b7db911)
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()
133 {
134 	INP_INFO_LOCK_INIT(&ripcbinfo, "rip");
135 	LIST_INIT(&ripcb);
136 	ripcbinfo.listhead = &ripcb;
137 	/*
138 	 * XXX We don't use the hash list for raw IP, but it's easier
139 	 * to allocate a one entry hash list than it is to check all
140 	 * over the place for hashbase == NULL.
141 	 */
142 	ripcbinfo.hashbase = hashinit(1, M_PCB, &ripcbinfo.hashmask);
143 	ripcbinfo.porthashbase = hashinit(1, M_PCB, &ripcbinfo.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 #if defined(IPSEC) || defined(FAST_IPSEC)
161 	/* check AH/ESP integrity. */
162 	if (ipsec4_in_reject(n, last)) {
163 		policyfail = 1;
164 #ifdef IPSEC
165 		ipsecstat.in_polvio++;
166 #endif /*IPSEC*/
167 		/* do not inject data to pcb */
168 	}
169 #endif /*IPSEC || FAST_IPSEC*/
170 #ifdef MAC
171 	if (!policyfail && mac_check_inpcb_deliver(last, n) != 0)
172 		policyfail = 1;
173 #endif
174 	/* Check the minimum TTL for socket. */
175 	if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
176 		policyfail = 1;
177 	if (!policyfail) {
178 		struct mbuf *opts = NULL;
179 		struct socket *so;
180 
181 		so = last->inp_socket;
182 		if ((last->inp_flags & INP_CONTROLOPTS) ||
183 		    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
184 			ip_savecontrol(last, &opts, ip, n);
185 		SOCKBUF_LOCK(&so->so_rcv);
186 		if (sbappendaddr_locked(&so->so_rcv,
187 		    (struct sockaddr *)&ripsrc, n, opts) == 0) {
188 			/* should notify about lost packet */
189 			m_freem(n);
190 			if (opts)
191 				m_freem(opts);
192 			SOCKBUF_UNLOCK(&so->so_rcv);
193 		} else
194 			sorwakeup_locked(so);
195 	} else
196 		m_freem(n);
197 	return policyfail;
198 }
199 
200 /*
201  * Setup generic address and protocol structures
202  * for raw_input routine, then pass them along with
203  * mbuf chain.
204  */
205 void
206 rip_input(struct mbuf *m, int off)
207 {
208 	struct ip *ip = mtod(m, struct ip *);
209 	int proto = ip->ip_p;
210 	struct inpcb *inp, *last;
211 
212 	INP_INFO_RLOCK(&ripcbinfo);
213 	ripsrc.sin_addr = ip->ip_src;
214 	last = NULL;
215 	LIST_FOREACH(inp, &ripcb, inp_list) {
216 		INP_LOCK(inp);
217 		if (inp->inp_ip_p && inp->inp_ip_p != proto) {
218 	docontinue:
219 			INP_UNLOCK(inp);
220 			continue;
221 		}
222 #ifdef INET6
223 		if ((inp->inp_vflag & INP_IPV4) == 0)
224 			goto docontinue;
225 #endif
226 		if (inp->inp_laddr.s_addr &&
227 		    inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
228 			goto docontinue;
229 		if (inp->inp_faddr.s_addr &&
230 		    inp->inp_faddr.s_addr != ip->ip_src.s_addr)
231 			goto docontinue;
232 		if (jailed(inp->inp_socket->so_cred))
233 			if (htonl(prison_getip(inp->inp_socket->so_cred)) !=
234 			    ip->ip_dst.s_addr)
235 				goto docontinue;
236 		if (last) {
237 			struct mbuf *n;
238 
239 			n = m_copy(m, 0, (int)M_COPYALL);
240 			if (n != NULL)
241 				(void) raw_append(last, ip, n);
242 			/* XXX count dropped packet */
243 			INP_UNLOCK(last);
244 		}
245 		last = inp;
246 	}
247 	if (last != NULL) {
248 		if (raw_append(last, ip, m) != 0)
249 			ipstat.ips_delivered--;
250 		INP_UNLOCK(last);
251 	} else {
252 		m_freem(m);
253 		ipstat.ips_noproto++;
254 		ipstat.ips_delivered--;
255 	}
256 	INP_INFO_RUNLOCK(&ripcbinfo);
257 }
258 
259 /*
260  * Generate IP header and pass packet to ip_output.
261  * Tack on options user may have setup with control call.
262  */
263 int
264 rip_output(struct mbuf *m, struct socket *so, u_long dst)
265 {
266 	struct ip *ip;
267 	int error;
268 	struct inpcb *inp = sotoinpcb(so);
269 	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
270 	    IP_ALLOWBROADCAST;
271 
272 	/*
273 	 * If the user handed us a complete IP packet, use it.
274 	 * Otherwise, allocate an mbuf for a header and fill it in.
275 	 */
276 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
277 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
278 			m_freem(m);
279 			return(EMSGSIZE);
280 		}
281 		M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
282 		if (m == NULL)
283 			return(ENOBUFS);
284 
285 		INP_LOCK(inp);
286 		ip = mtod(m, struct ip *);
287 		ip->ip_tos = inp->inp_ip_tos;
288 		if (inp->inp_flags & INP_DONTFRAG)
289 			ip->ip_off = IP_DF;
290 		else
291 			ip->ip_off = 0;
292 		ip->ip_p = inp->inp_ip_p;
293 		ip->ip_len = m->m_pkthdr.len;
294 		if (jailed(inp->inp_socket->so_cred))
295 			ip->ip_src.s_addr =
296 			    htonl(prison_getip(inp->inp_socket->so_cred));
297 		else
298 			ip->ip_src = inp->inp_laddr;
299 		ip->ip_dst.s_addr = dst;
300 		ip->ip_ttl = inp->inp_ip_ttl;
301 	} else {
302 		if (m->m_pkthdr.len > IP_MAXPACKET) {
303 			m_freem(m);
304 			return(EMSGSIZE);
305 		}
306 		INP_LOCK(inp);
307 		ip = mtod(m, struct ip *);
308 		if (jailed(inp->inp_socket->so_cred)) {
309 			if (ip->ip_src.s_addr !=
310 			    htonl(prison_getip(inp->inp_socket->so_cred))) {
311 				INP_UNLOCK(inp);
312 				m_freem(m);
313 				return (EPERM);
314 			}
315 		}
316 		/* don't allow both user specified and setsockopt options,
317 		   and don't allow packet length sizes that will crash */
318 		if (((ip->ip_hl != (sizeof (*ip) >> 2))
319 		     && inp->inp_options)
320 		    || (ip->ip_len > m->m_pkthdr.len)
321 		    || (ip->ip_len < (ip->ip_hl << 2))) {
322 			INP_UNLOCK(inp);
323 			m_freem(m);
324 			return EINVAL;
325 		}
326 		if (ip->ip_id == 0)
327 			ip->ip_id = ip_newid();
328 		/* XXX prevent ip_output from overwriting header fields */
329 		flags |= IP_RAWOUTPUT;
330 		ipstat.ips_rawout++;
331 	}
332 
333 	if (inp->inp_flags & INP_ONESBCAST)
334 		flags |= IP_SENDONES;
335 
336 #ifdef MAC
337 	mac_create_mbuf_from_inpcb(inp, m);
338 #endif
339 
340 	error = ip_output(m, inp->inp_options, NULL, flags,
341 	    inp->inp_moptions, inp);
342 	INP_UNLOCK(inp);
343 	return error;
344 }
345 
346 /*
347  * Raw IP socket option processing.
348  *
349  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
350  * only be created by a privileged process, and as such, socket option
351  * operations to manage system properties on any raw socket were allowed to
352  * take place without explicit additional access control checks.  However,
353  * raw sockets can now also be created in jail(), and therefore explicit
354  * checks are now required.  Likewise, raw sockets can be used by a process
355  * after it gives up privilege, so some caution is required.  For options
356  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
357  * performed in ip_ctloutput() and therefore no check occurs here.
358  * Unilaterally checking suser() here breaks normal IP socket option
359  * operations on raw sockets.
360  *
361  * When adding new socket options here, make sure to add access control
362  * checks here as necessary.
363  */
364 int
365 rip_ctloutput(struct socket *so, struct sockopt *sopt)
366 {
367 	struct	inpcb *inp = sotoinpcb(so);
368 	int	error, optval;
369 
370 	if (sopt->sopt_level != IPPROTO_IP)
371 		return (EINVAL);
372 
373 	error = 0;
374 	switch (sopt->sopt_dir) {
375 	case SOPT_GET:
376 		switch (sopt->sopt_name) {
377 		case IP_HDRINCL:
378 			optval = inp->inp_flags & INP_HDRINCL;
379 			error = sooptcopyout(sopt, &optval, sizeof optval);
380 			break;
381 
382 		case IP_FW_ADD:	/* ADD actually returns the body... */
383 		case IP_FW_GET:
384 		case IP_FW_TABLE_GETSIZE:
385 		case IP_FW_TABLE_LIST:
386 		case IP_FW_NAT_GET_CONFIG:
387 		case IP_FW_NAT_GET_LOG:
388 			/*
389 			 * XXXRW: Isn't this checked one layer down?  Yes, it
390 			 * is.
391 			 */
392 			error = priv_check(curthread, PRIV_NETINET_IPFW);
393 			if (error != 0)
394 				return (error);
395 			if (ip_fw_ctl_ptr != NULL)
396 				error = ip_fw_ctl_ptr(sopt);
397 			else
398 				error = ENOPROTOOPT;
399 			break;
400 
401 		case IP_DUMMYNET_GET:
402 			error = priv_check(curthread, PRIV_NETINET_DUMMYNET);
403 			if (error != 0)
404 				return (error);
405 			if (ip_dn_ctl_ptr != NULL)
406 				error = ip_dn_ctl_ptr(sopt);
407 			else
408 				error = ENOPROTOOPT;
409 			break ;
410 
411 		case MRT_INIT:
412 		case MRT_DONE:
413 		case MRT_ADD_VIF:
414 		case MRT_DEL_VIF:
415 		case MRT_ADD_MFC:
416 		case MRT_DEL_MFC:
417 		case MRT_VERSION:
418 		case MRT_ASSERT:
419 		case MRT_API_SUPPORT:
420 		case MRT_API_CONFIG:
421 		case MRT_ADD_BW_UPCALL:
422 		case MRT_DEL_BW_UPCALL:
423 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
424 			if (error != 0)
425 				return (error);
426 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
427 				EOPNOTSUPP;
428 			break;
429 
430 		default:
431 			error = ip_ctloutput(so, sopt);
432 			break;
433 		}
434 		break;
435 
436 	case SOPT_SET:
437 		switch (sopt->sopt_name) {
438 		case IP_HDRINCL:
439 			error = sooptcopyin(sopt, &optval, sizeof optval,
440 					    sizeof optval);
441 			if (error)
442 				break;
443 			if (optval)
444 				inp->inp_flags |= INP_HDRINCL;
445 			else
446 				inp->inp_flags &= ~INP_HDRINCL;
447 			break;
448 
449 		case IP_FW_ADD:
450 		case IP_FW_DEL:
451 		case IP_FW_FLUSH:
452 		case IP_FW_ZERO:
453 		case IP_FW_RESETLOG:
454 		case IP_FW_TABLE_ADD:
455 		case IP_FW_TABLE_DEL:
456 		case IP_FW_TABLE_FLUSH:
457 		case IP_FW_NAT_CFG:
458 		case IP_FW_NAT_DEL:
459 			/*
460 			 * XXXRW: Isn't this checked one layer down?
461 			 */
462 			error = priv_check(curthread, PRIV_NETINET_IPFW);
463 			if (error != 0)
464 				return (error);
465 			if (ip_fw_ctl_ptr != NULL)
466 				error = ip_fw_ctl_ptr(sopt);
467 			else
468 				error = ENOPROTOOPT;
469 			break;
470 
471 		case IP_DUMMYNET_CONFIGURE:
472 		case IP_DUMMYNET_DEL:
473 		case IP_DUMMYNET_FLUSH:
474 			error = priv_check(curthread, PRIV_NETINET_DUMMYNET);
475 			if (error != 0)
476 				return (error);
477 			if (ip_dn_ctl_ptr != NULL)
478 				error = ip_dn_ctl_ptr(sopt);
479 			else
480 				error = ENOPROTOOPT ;
481 			break ;
482 
483 		case IP_RSVP_ON:
484 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
485 			if (error != 0)
486 				return (error);
487 			error = ip_rsvp_init(so);
488 			break;
489 
490 		case IP_RSVP_OFF:
491 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
492 			if (error != 0)
493 				return (error);
494 			error = ip_rsvp_done();
495 			break;
496 
497 		case IP_RSVP_VIF_ON:
498 		case IP_RSVP_VIF_OFF:
499 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
500 			if (error != 0)
501 				return (error);
502 			error = ip_rsvp_vif ?
503 				ip_rsvp_vif(so, sopt) : EINVAL;
504 			break;
505 
506 		case MRT_INIT:
507 		case MRT_DONE:
508 		case MRT_ADD_VIF:
509 		case MRT_DEL_VIF:
510 		case MRT_ADD_MFC:
511 		case MRT_DEL_MFC:
512 		case MRT_VERSION:
513 		case MRT_ASSERT:
514 		case MRT_API_SUPPORT:
515 		case MRT_API_CONFIG:
516 		case MRT_ADD_BW_UPCALL:
517 		case MRT_DEL_BW_UPCALL:
518 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
519 			if (error != 0)
520 				return (error);
521 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
522 					EOPNOTSUPP;
523 			break;
524 
525 		default:
526 			error = ip_ctloutput(so, sopt);
527 			break;
528 		}
529 		break;
530 	}
531 
532 	return (error);
533 }
534 
535 /*
536  * This function exists solely to receive the PRC_IFDOWN messages which
537  * are sent by if_down().  It looks for an ifaddr whose ifa_addr is sa,
538  * and calls in_ifadown() to remove all routes corresponding to that address.
539  * It also receives the PRC_IFUP messages from if_up() and reinstalls the
540  * interface routes.
541  */
542 void
543 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
544 {
545 	struct in_ifaddr *ia;
546 	struct ifnet *ifp;
547 	int err;
548 	int flags;
549 
550 	switch (cmd) {
551 	case PRC_IFDOWN:
552 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
553 			if (ia->ia_ifa.ifa_addr == sa
554 			    && (ia->ia_flags & IFA_ROUTE)) {
555 				/*
556 				 * in_ifscrub kills the interface route.
557 				 */
558 				in_ifscrub(ia->ia_ifp, ia);
559 				/*
560 				 * in_ifadown gets rid of all the rest of
561 				 * the routes.  This is not quite the right
562 				 * thing to do, but at least if we are running
563 				 * a routing process they will come back.
564 				 */
565 				in_ifadown(&ia->ia_ifa, 0);
566 				break;
567 			}
568 		}
569 		break;
570 
571 	case PRC_IFUP:
572 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
573 			if (ia->ia_ifa.ifa_addr == sa)
574 				break;
575 		}
576 		if (ia == 0 || (ia->ia_flags & IFA_ROUTE))
577 			return;
578 		flags = RTF_UP;
579 		ifp = ia->ia_ifa.ifa_ifp;
580 
581 		if ((ifp->if_flags & IFF_LOOPBACK)
582 		    || (ifp->if_flags & IFF_POINTOPOINT))
583 			flags |= RTF_HOST;
584 
585 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
586 		if (err == 0)
587 			ia->ia_flags |= IFA_ROUTE;
588 		break;
589 	}
590 }
591 
592 u_long	rip_sendspace = 9216;
593 u_long	rip_recvspace = 9216;
594 
595 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
596     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
597 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
598     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
599 
600 static int
601 rip_attach(struct socket *so, int proto, struct thread *td)
602 {
603 	struct inpcb *inp;
604 	int error;
605 
606 	inp = sotoinpcb(so);
607 	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
608 	/*
609 	 * XXXRW: Centralize privilege decision in kern_jail.c.
610 	 */
611 	if (jailed(td->td_ucred) && !jail_allow_raw_sockets)
612 		return (EPERM);
613 	error = priv_check_cred(td->td_ucred, PRIV_NETINET_RAW,
614 	    SUSER_ALLOWJAIL);
615 	if (error)
616 		return error;
617 	if (proto >= IPPROTO_MAX || proto < 0)
618 		return EPROTONOSUPPORT;
619 	error = soreserve(so, rip_sendspace, rip_recvspace);
620 	if (error)
621 		return error;
622 	INP_INFO_WLOCK(&ripcbinfo);
623 	error = in_pcballoc(so, &ripcbinfo);
624 	if (error) {
625 		INP_INFO_WUNLOCK(&ripcbinfo);
626 		return error;
627 	}
628 	inp = (struct inpcb *)so->so_pcb;
629 	INP_INFO_WUNLOCK(&ripcbinfo);
630 	inp->inp_vflag |= INP_IPV4;
631 	inp->inp_ip_p = proto;
632 	inp->inp_ip_ttl = ip_defttl;
633 	INP_UNLOCK(inp);
634 	return 0;
635 }
636 
637 static void
638 rip_detach(struct socket *so)
639 {
640 	struct inpcb *inp;
641 
642 	inp = sotoinpcb(so);
643 	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
644 	KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
645 	    ("rip_detach: not closed"));
646 
647 	INP_INFO_WLOCK(&ripcbinfo);
648 	INP_LOCK(inp);
649 	if (so == ip_mrouter && ip_mrouter_done)
650 		ip_mrouter_done();
651 	if (ip_rsvp_force_done)
652 		ip_rsvp_force_done(so);
653 	if (so == ip_rsvpd)
654 		ip_rsvp_done();
655 	in_pcbdetach(inp);
656 	in_pcbfree(inp);
657 	INP_INFO_WUNLOCK(&ripcbinfo);
658 }
659 
660 static void
661 rip_dodisconnect(struct socket *so, struct inpcb *inp)
662 {
663 
664 	INP_LOCK_ASSERT(inp);
665 
666 	inp->inp_faddr.s_addr = INADDR_ANY;
667 	SOCK_LOCK(so);
668 	so->so_state &= ~SS_ISCONNECTED;
669 	SOCK_UNLOCK(so);
670 }
671 
672 static void
673 rip_abort(struct socket *so)
674 {
675 	struct inpcb *inp;
676 
677 	inp = sotoinpcb(so);
678 	KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
679 
680 	INP_INFO_WLOCK(&ripcbinfo);
681 	INP_LOCK(inp);
682 	rip_dodisconnect(so, inp);
683 	INP_UNLOCK(inp);
684 	INP_INFO_WUNLOCK(&ripcbinfo);
685 }
686 
687 static void
688 rip_close(struct socket *so)
689 {
690 	struct inpcb *inp;
691 
692 	inp = sotoinpcb(so);
693 	KASSERT(inp != NULL, ("rip_close: inp == NULL"));
694 
695 	INP_INFO_WLOCK(&ripcbinfo);
696 	INP_LOCK(inp);
697 	rip_dodisconnect(so, inp);
698 	INP_UNLOCK(inp);
699 	INP_INFO_WUNLOCK(&ripcbinfo);
700 }
701 
702 static int
703 rip_disconnect(struct socket *so)
704 {
705 	struct inpcb *inp;
706 
707 	if ((so->so_state & SS_ISCONNECTED) == 0)
708 		return ENOTCONN;
709 
710 	inp = sotoinpcb(so);
711 	KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
712 	INP_INFO_WLOCK(&ripcbinfo);
713 	INP_LOCK(inp);
714 	rip_dodisconnect(so, inp);
715 	INP_UNLOCK(inp);
716 	INP_INFO_WUNLOCK(&ripcbinfo);
717 	return (0);
718 }
719 
720 static int
721 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
722 {
723 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
724 	struct inpcb *inp;
725 
726 	if (nam->sa_len != sizeof(*addr))
727 		return EINVAL;
728 
729 	if (jailed(td->td_ucred)) {
730 		if (addr->sin_addr.s_addr == INADDR_ANY)
731 			addr->sin_addr.s_addr =
732 			    htonl(prison_getip(td->td_ucred));
733 		if (htonl(prison_getip(td->td_ucred)) != addr->sin_addr.s_addr)
734 			return (EADDRNOTAVAIL);
735 	}
736 
737 	if (TAILQ_EMPTY(&ifnet) ||
738 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
739 	    (addr->sin_addr.s_addr &&
740 	     ifa_ifwithaddr((struct sockaddr *)addr) == 0))
741 		return EADDRNOTAVAIL;
742 
743 	inp = sotoinpcb(so);
744 	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
745 	INP_INFO_WLOCK(&ripcbinfo);
746 	INP_LOCK(inp);
747 	inp->inp_laddr = addr->sin_addr;
748 	INP_UNLOCK(inp);
749 	INP_INFO_WUNLOCK(&ripcbinfo);
750 	return 0;
751 }
752 
753 static int
754 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
755 {
756 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
757 	struct inpcb *inp;
758 
759 	if (nam->sa_len != sizeof(*addr))
760 		return EINVAL;
761 	if (TAILQ_EMPTY(&ifnet))
762 		return EADDRNOTAVAIL;
763 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
764 		return EAFNOSUPPORT;
765 
766 	inp = sotoinpcb(so);
767 	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
768 	INP_INFO_WLOCK(&ripcbinfo);
769 	INP_LOCK(inp);
770 	inp->inp_faddr = addr->sin_addr;
771 	soisconnected(so);
772 	INP_UNLOCK(inp);
773 	INP_INFO_WUNLOCK(&ripcbinfo);
774 	return 0;
775 }
776 
777 static int
778 rip_shutdown(struct socket *so)
779 {
780 	struct inpcb *inp;
781 
782 	inp = sotoinpcb(so);
783 	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
784 	INP_LOCK(inp);
785 	socantsendmore(so);
786 	INP_UNLOCK(inp);
787 	return 0;
788 }
789 
790 static int
791 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
792 	 struct mbuf *control, struct thread *td)
793 {
794 	struct inpcb *inp;
795 	u_long dst;
796 
797 	inp = sotoinpcb(so);
798 	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
799 	/*
800 	 * Note: 'dst' reads below are unlocked.
801 	 */
802 	if (so->so_state & SS_ISCONNECTED) {
803 		if (nam) {
804 			m_freem(m);
805 			return EISCONN;
806 		}
807 		dst = inp->inp_faddr.s_addr;	/* Unlocked read. */
808 	} else {
809 		if (nam == NULL) {
810 			m_freem(m);
811 			return ENOTCONN;
812 		}
813 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
814 	}
815 	return rip_output(m, so, dst);
816 }
817 
818 static int
819 rip_pcblist(SYSCTL_HANDLER_ARGS)
820 {
821 	int error, i, n;
822 	struct inpcb *inp, **inp_list;
823 	inp_gen_t gencnt;
824 	struct xinpgen xig;
825 
826 	/*
827 	 * The process of preparing the TCB list is too time-consuming and
828 	 * resource-intensive to repeat twice on every request.
829 	 */
830 	if (req->oldptr == 0) {
831 		n = ripcbinfo.ipi_count;
832 		req->oldidx = 2 * (sizeof xig)
833 			+ (n + n/8) * sizeof(struct xinpcb);
834 		return 0;
835 	}
836 
837 	if (req->newptr != 0)
838 		return EPERM;
839 
840 	/*
841 	 * OK, now we're committed to doing something.
842 	 */
843 	INP_INFO_RLOCK(&ripcbinfo);
844 	gencnt = ripcbinfo.ipi_gencnt;
845 	n = ripcbinfo.ipi_count;
846 	INP_INFO_RUNLOCK(&ripcbinfo);
847 
848 	xig.xig_len = sizeof xig;
849 	xig.xig_count = n;
850 	xig.xig_gen = gencnt;
851 	xig.xig_sogen = so_gencnt;
852 	error = SYSCTL_OUT(req, &xig, sizeof xig);
853 	if (error)
854 		return error;
855 
856 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
857 	if (inp_list == 0)
858 		return ENOMEM;
859 
860 	INP_INFO_RLOCK(&ripcbinfo);
861 	for (inp = LIST_FIRST(ripcbinfo.listhead), i = 0; inp && i < n;
862 	     inp = LIST_NEXT(inp, inp_list)) {
863 		INP_LOCK(inp);
864 		if (inp->inp_gencnt <= gencnt &&
865 		    cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) {
866 			/* XXX held references? */
867 			inp_list[i++] = inp;
868 		}
869 		INP_UNLOCK(inp);
870 	}
871 	INP_INFO_RUNLOCK(&ripcbinfo);
872 	n = i;
873 
874 	error = 0;
875 	for (i = 0; i < n; i++) {
876 		inp = inp_list[i];
877 		INP_LOCK(inp);
878 		if (inp->inp_gencnt <= gencnt) {
879 			struct xinpcb xi;
880 			bzero(&xi, sizeof(xi));
881 			xi.xi_len = sizeof xi;
882 			/* XXX should avoid extra copy */
883 			bcopy(inp, &xi.xi_inp, sizeof *inp);
884 			if (inp->inp_socket)
885 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
886 			INP_UNLOCK(inp);
887 			error = SYSCTL_OUT(req, &xi, sizeof xi);
888 		} else
889 			INP_UNLOCK(inp);
890 	}
891 	if (!error) {
892 		/*
893 		 * Give the user an updated idea of our state.
894 		 * If the generation differs from what we told
895 		 * her before, she knows that something happened
896 		 * while we were processing this request, and it
897 		 * might be necessary to retry.
898 		 */
899 		INP_INFO_RLOCK(&ripcbinfo);
900 		xig.xig_gen = ripcbinfo.ipi_gencnt;
901 		xig.xig_sogen = so_gencnt;
902 		xig.xig_count = ripcbinfo.ipi_count;
903 		INP_INFO_RUNLOCK(&ripcbinfo);
904 		error = SYSCTL_OUT(req, &xig, sizeof xig);
905 	}
906 	free(inp_list, M_TEMP);
907 	return error;
908 }
909 
910 /*
911  * This is the wrapper function for in_setsockaddr.  We just pass down
912  * the pcbinfo for in_setpeeraddr to lock.
913  */
914 static int
915 rip_sockaddr(struct socket *so, struct sockaddr **nam)
916 {
917 	return (in_setsockaddr(so, nam, &ripcbinfo));
918 }
919 
920 /*
921  * This is the wrapper function for in_setpeeraddr.  We just pass down
922  * the pcbinfo for in_setpeeraddr to lock.
923  */
924 static int
925 rip_peeraddr(struct socket *so, struct sockaddr **nam)
926 {
927 	return (in_setpeeraddr(so, nam, &ripcbinfo));
928 }
929 
930 
931 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0,
932 	    rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
933 
934 struct pr_usrreqs rip_usrreqs = {
935 	.pru_abort =		rip_abort,
936 	.pru_attach =		rip_attach,
937 	.pru_bind =		rip_bind,
938 	.pru_connect =		rip_connect,
939 	.pru_control =		in_control,
940 	.pru_detach =		rip_detach,
941 	.pru_disconnect =	rip_disconnect,
942 	.pru_peeraddr =		rip_peeraddr,
943 	.pru_send =		rip_send,
944 	.pru_shutdown =		rip_shutdown,
945 	.pru_sockaddr =		rip_sockaddr,
946 	.pru_sosetlabel =	in_pcbsosetlabel,
947 	.pru_close =		rip_close,
948 };
949