xref: /freebsd/sys/netinet/raw_ip.c (revision 2357939bc239bd5334a169b62313806178dd8f30)
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 #include "opt_random_ip_id.h"
37 
38 #include <sys/param.h>
39 #include <sys/jail.h>
40 #include <sys/kernel.h>
41 #include <sys/lock.h>
42 #include <sys/mac.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/proc.h>
46 #include <sys/protosw.h>
47 #include <sys/signalvar.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/sx.h>
51 #include <sys/sysctl.h>
52 #include <sys/systm.h>
53 
54 #include <vm/uma.h>
55 
56 #include <net/if.h>
57 #include <net/route.h>
58 
59 #include <netinet/in.h>
60 #include <netinet/in_systm.h>
61 #include <netinet/in_pcb.h>
62 #include <netinet/in_var.h>
63 #include <netinet/ip.h>
64 #include <netinet/ip_var.h>
65 #include <netinet/ip_mroute.h>
66 
67 #include <netinet/ip_fw.h>
68 #include <netinet/ip_dummynet.h>
69 
70 #ifdef FAST_IPSEC
71 #include <netipsec/ipsec.h>
72 #endif /*FAST_IPSEC*/
73 
74 #ifdef IPSEC
75 #include <netinet6/ipsec.h>
76 #endif /*IPSEC*/
77 
78 struct	inpcbhead ripcb;
79 struct	inpcbinfo ripcbinfo;
80 
81 /* control hooks for ipfw and dummynet */
82 ip_fw_ctl_t *ip_fw_ctl_ptr;
83 ip_dn_ctl_t *ip_dn_ctl_ptr;
84 
85 /*
86  * hooks for multicast routing. They all default to NULL,
87  * so leave them not initialized and rely on BSS being set to 0.
88  */
89 
90 /* The socket used to communicate with the multicast routing daemon.  */
91 struct socket  *ip_mrouter;
92 
93 /* The various mrouter and rsvp functions */
94 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
95 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
96 int (*ip_mrouter_done)(void);
97 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
98                    struct ip_moptions *);
99 int (*mrt_ioctl)(int, caddr_t);
100 int (*legal_vif_num)(int);
101 u_long (*ip_mcast_src)(int);
102 
103 void (*rsvp_input_p)(struct mbuf *m, int off);
104 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
105 void (*ip_rsvp_force_done)(struct socket *);
106 
107 /*
108  * Nominal space allocated to a raw ip socket.
109  */
110 #define	RIPSNDQ		8192
111 #define	RIPRCVQ		8192
112 
113 /*
114  * Raw interface to IP protocol.
115  */
116 
117 /*
118  * Initialize raw connection block q.
119  */
120 void
121 rip_init()
122 {
123 	INP_INFO_LOCK_INIT(&ripcbinfo, "rip");
124 	LIST_INIT(&ripcb);
125 	ripcbinfo.listhead = &ripcb;
126 	/*
127 	 * XXX We don't use the hash list for raw IP, but it's easier
128 	 * to allocate a one entry hash list than it is to check all
129 	 * over the place for hashbase == NULL.
130 	 */
131 	ripcbinfo.hashbase = hashinit(1, M_PCB, &ripcbinfo.hashmask);
132 	ripcbinfo.porthashbase = hashinit(1, M_PCB, &ripcbinfo.porthashmask);
133 	ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb),
134 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
135 	uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets);
136 }
137 
138 static struct	sockaddr_in ripsrc = { sizeof(ripsrc), AF_INET };
139 
140 static int
141 raw_append(struct inpcb *last, struct ip *ip, struct mbuf *n)
142 {
143 	int policyfail = 0;
144 
145 	INP_LOCK_ASSERT(last);
146 
147 #if defined(IPSEC) || defined(FAST_IPSEC)
148 	/* check AH/ESP integrity. */
149 	if (ipsec4_in_reject(n, last)) {
150 		policyfail = 1;
151 #ifdef IPSEC
152 		ipsecstat.in_polvio++;
153 #endif /*IPSEC*/
154 		/* do not inject data to pcb */
155 	}
156 #endif /*IPSEC || FAST_IPSEC*/
157 #ifdef MAC
158 	if (!policyfail && mac_check_inpcb_deliver(last, n) != 0)
159 		policyfail = 1;
160 #endif
161 	if (!policyfail) {
162 		struct mbuf *opts = NULL;
163 
164 		if ((last->inp_flags & INP_CONTROLOPTS) ||
165 		    (last->inp_socket->so_options & SO_TIMESTAMP))
166 			ip_savecontrol(last, &opts, ip, n);
167 		if (sbappendaddr(&last->inp_socket->so_rcv,
168 		    (struct sockaddr *)&ripsrc, n, opts) == 0) {
169 			/* should notify about lost packet */
170 			m_freem(n);
171 			if (opts)
172 				m_freem(opts);
173 		} else
174 			sorwakeup(last->inp_socket);
175 	} else
176 		m_freem(n);
177 	return policyfail;
178 }
179 
180 /*
181  * Setup generic address and protocol structures
182  * for raw_input routine, then pass them along with
183  * mbuf chain.
184  */
185 void
186 rip_input(struct mbuf *m, int off)
187 {
188 	struct ip *ip = mtod(m, struct ip *);
189 	int proto = ip->ip_p;
190 	struct inpcb *inp, *last;
191 
192 	INP_INFO_RLOCK(&ripcbinfo);
193 	ripsrc.sin_addr = ip->ip_src;
194 	last = NULL;
195 	LIST_FOREACH(inp, &ripcb, inp_list) {
196 		INP_LOCK(inp);
197 		if (inp->inp_ip_p && inp->inp_ip_p != proto) {
198 	docontinue:
199 			INP_UNLOCK(inp);
200 			continue;
201 		}
202 #ifdef INET6
203 		if ((inp->inp_vflag & INP_IPV4) == 0)
204 			goto docontinue;
205 #endif
206 		if (inp->inp_laddr.s_addr &&
207                     inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
208 			goto docontinue;
209 		if (inp->inp_faddr.s_addr &&
210                     inp->inp_faddr.s_addr != ip->ip_src.s_addr)
211 			goto docontinue;
212 		if (jailed(inp->inp_socket->so_cred))
213 			if (htonl(prison_getip(inp->inp_socket->so_cred)) !=
214 			    ip->ip_dst.s_addr)
215 				goto docontinue;
216 		if (last) {
217 			struct mbuf *n;
218 
219 			n = m_copy(m, 0, (int)M_COPYALL);
220 			if (n != NULL)
221 				(void) raw_append(last, ip, n);
222 			/* XXX count dropped packet */
223 			INP_UNLOCK(last);
224 		}
225 		last = inp;
226 	}
227 	if (last != NULL) {
228 		if (raw_append(last, ip, m) != 0)
229 			ipstat.ips_delivered--;
230 		INP_UNLOCK(last);
231 	} else {
232 		m_freem(m);
233 		ipstat.ips_noproto++;
234 		ipstat.ips_delivered--;
235 	}
236 	INP_INFO_RUNLOCK(&ripcbinfo);
237 }
238 
239 /*
240  * Generate IP header and pass packet to ip_output.
241  * Tack on options user may have setup with control call.
242  */
243 int
244 rip_output(struct mbuf *m, struct socket *so, u_long dst)
245 {
246 	struct ip *ip;
247 	struct inpcb *inp = sotoinpcb(so);
248 	int flags = (so->so_options & SO_DONTROUTE) | IP_ALLOWBROADCAST;
249 
250 #ifdef MAC
251 	INP_LOCK(inp);
252 	mac_create_mbuf_from_inpcb(inp, m);
253 	INP_UNLOCK(inp);
254 #endif
255 
256 	/*
257 	 * If the user handed us a complete IP packet, use it.
258 	 * Otherwise, allocate an mbuf for a header and fill it in.
259 	 */
260 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
261 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
262 			m_freem(m);
263 			return(EMSGSIZE);
264 		}
265 		M_PREPEND(m, sizeof(struct ip), M_TRYWAIT);
266 		if (m == NULL)
267 			return(ENOBUFS);
268 		ip = mtod(m, struct ip *);
269 		ip->ip_tos = inp->inp_ip_tos;
270 		ip->ip_off = 0;
271 		ip->ip_p = inp->inp_ip_p;
272 		ip->ip_len = m->m_pkthdr.len;
273 		if (jailed(inp->inp_socket->so_cred))
274 			ip->ip_src.s_addr =
275 			    htonl(prison_getip(inp->inp_socket->so_cred));
276 		else
277 			ip->ip_src = inp->inp_laddr;
278 		ip->ip_dst.s_addr = dst;
279 		ip->ip_ttl = inp->inp_ip_ttl;
280 	} else {
281 		if (m->m_pkthdr.len > IP_MAXPACKET) {
282 			m_freem(m);
283 			return(EMSGSIZE);
284 		}
285 		ip = mtod(m, struct ip *);
286 		if (jailed(inp->inp_socket->so_cred)) {
287 			if (ip->ip_src.s_addr !=
288 			    htonl(prison_getip(inp->inp_socket->so_cred))) {
289 				m_freem(m);
290 				return (EPERM);
291 			}
292 		}
293 		/* don't allow both user specified and setsockopt options,
294 		   and don't allow packet length sizes that will crash */
295 		if (((ip->ip_hl != (sizeof (*ip) >> 2))
296 		     && inp->inp_options)
297 		    || (ip->ip_len > m->m_pkthdr.len)
298 		    || (ip->ip_len < (ip->ip_hl << 2))) {
299 			m_freem(m);
300 			return EINVAL;
301 		}
302 		if (ip->ip_id == 0)
303 #ifdef RANDOM_IP_ID
304 			ip->ip_id = ip_randomid();
305 #else
306 			ip->ip_id = htons(ip_id++);
307 #endif
308 		/* XXX prevent ip_output from overwriting header fields */
309 		flags |= IP_RAWOUTPUT;
310 		ipstat.ips_rawout++;
311 	}
312 
313 	if (inp->inp_flags & INP_ONESBCAST)
314 		flags |= IP_SENDONES;
315 
316 	return (ip_output(m, inp->inp_options, NULL, flags,
317 			  inp->inp_moptions, inp));
318 }
319 
320 /*
321  * Raw IP socket option processing.
322  *
323  * Note that access to all of the IP administrative functions here is
324  * implicitly protected by suser() as gaining access to a raw socket
325  * requires either that the thread pass a suser() check, or that it be
326  * passed a raw socket by another thread that has passed a suser() check.
327  * If FreeBSD moves to a more fine-grained access control mechanism,
328  * additional checks will need to be placed here if the raw IP attachment
329  * check is not equivilent the the check required for these
330  * administrative operations; in some cases, these checks are already
331  * present.
332  */
333 int
334 rip_ctloutput(struct socket *so, struct sockopt *sopt)
335 {
336 	struct	inpcb *inp = sotoinpcb(so);
337 	int	error, optval;
338 
339 	if (sopt->sopt_level != IPPROTO_IP)
340 		return (EINVAL);
341 
342 	error = 0;
343 
344 	switch (sopt->sopt_dir) {
345 	case SOPT_GET:
346 		switch (sopt->sopt_name) {
347 		case IP_HDRINCL:
348 			optval = inp->inp_flags & INP_HDRINCL;
349 			error = sooptcopyout(sopt, &optval, sizeof optval);
350 			break;
351 
352 		case IP_FW_ADD:	/* ADD actually returns the body... */
353 		case IP_FW_GET:
354 			if (IPFW_LOADED)
355 				error = ip_fw_ctl_ptr(sopt);
356 			else
357 				error = ENOPROTOOPT;
358 			break;
359 
360 		case IP_DUMMYNET_GET:
361 			if (DUMMYNET_LOADED)
362 				error = ip_dn_ctl_ptr(sopt);
363 			else
364 				error = ENOPROTOOPT;
365 			break ;
366 
367 		case MRT_INIT:
368 		case MRT_DONE:
369 		case MRT_ADD_VIF:
370 		case MRT_DEL_VIF:
371 		case MRT_ADD_MFC:
372 		case MRT_DEL_MFC:
373 		case MRT_VERSION:
374 		case MRT_ASSERT:
375 		case MRT_API_SUPPORT:
376 		case MRT_API_CONFIG:
377 		case MRT_ADD_BW_UPCALL:
378 		case MRT_DEL_BW_UPCALL:
379 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
380 				EOPNOTSUPP;
381 			break;
382 
383 		default:
384 			error = ip_ctloutput(so, sopt);
385 			break;
386 		}
387 		break;
388 
389 	case SOPT_SET:
390 		switch (sopt->sopt_name) {
391 		case IP_HDRINCL:
392 			error = sooptcopyin(sopt, &optval, sizeof optval,
393 					    sizeof optval);
394 			if (error)
395 				break;
396 			if (optval)
397 				inp->inp_flags |= INP_HDRINCL;
398 			else
399 				inp->inp_flags &= ~INP_HDRINCL;
400 			break;
401 
402 		case IP_FW_ADD:
403 		case IP_FW_DEL:
404 		case IP_FW_FLUSH:
405 		case IP_FW_ZERO:
406 		case IP_FW_RESETLOG:
407 			if (IPFW_LOADED)
408 				error = ip_fw_ctl_ptr(sopt);
409 			else
410 				error = ENOPROTOOPT;
411 			break;
412 
413 		case IP_DUMMYNET_CONFIGURE:
414 		case IP_DUMMYNET_DEL:
415 		case IP_DUMMYNET_FLUSH:
416 			if (DUMMYNET_LOADED)
417 				error = ip_dn_ctl_ptr(sopt);
418 			else
419 				error = ENOPROTOOPT ;
420 			break ;
421 
422 		case IP_RSVP_ON:
423 			error = ip_rsvp_init(so);
424 			break;
425 
426 		case IP_RSVP_OFF:
427 			error = ip_rsvp_done();
428 			break;
429 
430 		case IP_RSVP_VIF_ON:
431 		case IP_RSVP_VIF_OFF:
432 			error = ip_rsvp_vif ?
433 				ip_rsvp_vif(so, sopt) : EINVAL;
434 			break;
435 
436 		case MRT_INIT:
437 		case MRT_DONE:
438 		case MRT_ADD_VIF:
439 		case MRT_DEL_VIF:
440 		case MRT_ADD_MFC:
441 		case MRT_DEL_MFC:
442 		case MRT_VERSION:
443 		case MRT_ASSERT:
444 		case MRT_API_SUPPORT:
445 		case MRT_API_CONFIG:
446 		case MRT_ADD_BW_UPCALL:
447 		case MRT_DEL_BW_UPCALL:
448 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
449 					EOPNOTSUPP;
450 			break;
451 
452 		default:
453 			error = ip_ctloutput(so, sopt);
454 			break;
455 		}
456 		break;
457 	}
458 
459 	return (error);
460 }
461 
462 /*
463  * This function exists solely to receive the PRC_IFDOWN messages which
464  * are sent by if_down().  It looks for an ifaddr whose ifa_addr is sa,
465  * and calls in_ifadown() to remove all routes corresponding to that address.
466  * It also receives the PRC_IFUP messages from if_up() and reinstalls the
467  * interface routes.
468  */
469 void
470 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
471 {
472 	struct in_ifaddr *ia;
473 	struct ifnet *ifp;
474 	int err;
475 	int flags;
476 
477 	switch (cmd) {
478 	case PRC_IFDOWN:
479 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
480 			if (ia->ia_ifa.ifa_addr == sa
481 			    && (ia->ia_flags & IFA_ROUTE)) {
482 				/*
483 				 * in_ifscrub kills the interface route.
484 				 */
485 				in_ifscrub(ia->ia_ifp, ia);
486 				/*
487 				 * in_ifadown gets rid of all the rest of
488 				 * the routes.  This is not quite the right
489 				 * thing to do, but at least if we are running
490 				 * a routing process they will come back.
491 				 */
492 				in_ifadown(&ia->ia_ifa, 0);
493 				break;
494 			}
495 		}
496 		break;
497 
498 	case PRC_IFUP:
499 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
500 			if (ia->ia_ifa.ifa_addr == sa)
501 				break;
502 		}
503 		if (ia == 0 || (ia->ia_flags & IFA_ROUTE))
504 			return;
505 		flags = RTF_UP;
506 		ifp = ia->ia_ifa.ifa_ifp;
507 
508 		if ((ifp->if_flags & IFF_LOOPBACK)
509 		    || (ifp->if_flags & IFF_POINTOPOINT))
510 			flags |= RTF_HOST;
511 
512 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
513 		if (err == 0)
514 			ia->ia_flags |= IFA_ROUTE;
515 		break;
516 	}
517 }
518 
519 u_long	rip_sendspace = RIPSNDQ;
520 u_long	rip_recvspace = RIPRCVQ;
521 
522 SYSCTL_INT(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
523     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
524 SYSCTL_INT(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
525     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
526 
527 static int
528 rip_attach(struct socket *so, int proto, struct thread *td)
529 {
530 	struct inpcb *inp;
531 	int error;
532 
533 	/* XXX why not lower? */
534 	INP_INFO_WLOCK(&ripcbinfo);
535 	inp = sotoinpcb(so);
536 	if (inp) {
537 		/* XXX counter, printf */
538 		INP_INFO_WUNLOCK(&ripcbinfo);
539 		return EINVAL;
540 	}
541 	if (td && jailed(td->td_ucred) && !jail_allow_raw_sockets) {
542 		INP_INFO_WUNLOCK(&ripcbinfo);
543 		return (EPERM);
544 	}
545 	if (td && (error = suser_cred(td->td_ucred, PRISON_ROOT)) != 0) {
546 		INP_INFO_WUNLOCK(&ripcbinfo);
547 		return error;
548 	}
549 	if (proto >= IPPROTO_MAX || proto < 0) {
550 		INP_INFO_WUNLOCK(&ripcbinfo);
551 		return EPROTONOSUPPORT;
552 	}
553 
554 	error = soreserve(so, rip_sendspace, rip_recvspace);
555 	if (error) {
556 		INP_INFO_WUNLOCK(&ripcbinfo);
557 		return error;
558 	}
559 	error = in_pcballoc(so, &ripcbinfo, "rawinp");
560 	if (error) {
561 		INP_INFO_WUNLOCK(&ripcbinfo);
562 		return error;
563 	}
564 	inp = (struct inpcb *)so->so_pcb;
565 	INP_LOCK(inp);
566 	INP_INFO_WUNLOCK(&ripcbinfo);
567 	inp->inp_vflag |= INP_IPV4;
568 	inp->inp_ip_p = proto;
569 	inp->inp_ip_ttl = ip_defttl;
570 	INP_UNLOCK(inp);
571 	return 0;
572 }
573 
574 static void
575 rip_pcbdetach(struct socket *so, struct inpcb *inp)
576 {
577 	INP_INFO_WLOCK_ASSERT(&ripcbinfo);
578 	INP_LOCK_ASSERT(inp);
579 
580 	if (so == ip_mrouter && ip_mrouter_done)
581 		ip_mrouter_done();
582 	if (ip_rsvp_force_done)
583 		ip_rsvp_force_done(so);
584 	if (so == ip_rsvpd)
585 		ip_rsvp_done();
586 	in_pcbdetach(inp);
587 }
588 
589 static int
590 rip_detach(struct socket *so)
591 {
592 	struct inpcb *inp;
593 
594 	INP_INFO_WLOCK(&ripcbinfo);
595 	inp = sotoinpcb(so);
596 	if (inp == 0) {
597 		/* XXX counter, printf */
598 		INP_INFO_WUNLOCK(&ripcbinfo);
599 		return EINVAL;
600 	}
601 	INP_LOCK(inp);
602 	rip_pcbdetach(so, inp);
603 	INP_INFO_WUNLOCK(&ripcbinfo);
604 	return 0;
605 }
606 
607 static int
608 rip_abort(struct socket *so)
609 {
610 	struct inpcb *inp;
611 
612 	INP_INFO_WLOCK(&ripcbinfo);
613 	inp = sotoinpcb(so);
614 	if (inp == 0) {
615 		INP_INFO_WUNLOCK(&ripcbinfo);
616 		return EINVAL;	/* ??? possible? panic instead? */
617 	}
618 	INP_LOCK(inp);
619 	soisdisconnected(so);
620 	if (so->so_state & SS_NOFDREF)
621 		rip_pcbdetach(so, inp);
622 	else
623 		INP_UNLOCK(inp);
624 	INP_INFO_WUNLOCK(&ripcbinfo);
625 	return 0;
626 }
627 
628 static int
629 rip_disconnect(struct socket *so)
630 {
631 	if ((so->so_state & SS_ISCONNECTED) == 0)
632 		return ENOTCONN;
633 	return rip_abort(so);
634 }
635 
636 static int
637 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
638 {
639 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
640 	struct inpcb *inp;
641 
642 	if (nam->sa_len != sizeof(*addr))
643 		return EINVAL;
644 
645 	if (jailed(td->td_ucred)) {
646 		if (addr->sin_addr.s_addr == INADDR_ANY)
647 			addr->sin_addr.s_addr =
648 			    htonl(prison_getip(td->td_ucred));
649 		if (htonl(prison_getip(td->td_ucred)) != addr->sin_addr.s_addr)
650 			return (EADDRNOTAVAIL);
651 	}
652 
653 	if (TAILQ_EMPTY(&ifnet) ||
654 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
655 	    (addr->sin_addr.s_addr &&
656 	     ifa_ifwithaddr((struct sockaddr *)addr) == 0))
657 		return EADDRNOTAVAIL;
658 
659 	INP_INFO_WLOCK(&ripcbinfo);
660 	inp = sotoinpcb(so);
661 	if (inp == 0) {
662 		INP_INFO_WUNLOCK(&ripcbinfo);
663 		return EINVAL;
664 	}
665 	INP_LOCK(inp);
666 	inp->inp_laddr = addr->sin_addr;
667 	INP_UNLOCK(inp);
668 	INP_INFO_WUNLOCK(&ripcbinfo);
669 	return 0;
670 }
671 
672 static int
673 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
674 {
675 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
676 	struct inpcb *inp;
677 
678 	if (nam->sa_len != sizeof(*addr))
679 		return EINVAL;
680 	if (TAILQ_EMPTY(&ifnet))
681 		return EADDRNOTAVAIL;
682 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
683 		return EAFNOSUPPORT;
684 
685 	INP_INFO_WLOCK(&ripcbinfo);
686 	inp = sotoinpcb(so);
687 	if (inp == 0) {
688 		INP_INFO_WUNLOCK(&ripcbinfo);
689 		return EINVAL;
690 	}
691 	INP_LOCK(inp);
692 	inp->inp_faddr = addr->sin_addr;
693 	soisconnected(so);
694 	INP_UNLOCK(inp);
695 	INP_INFO_WUNLOCK(&ripcbinfo);
696 	return 0;
697 }
698 
699 static int
700 rip_shutdown(struct socket *so)
701 {
702 	struct inpcb *inp;
703 
704 	INP_INFO_RLOCK(&ripcbinfo);
705 	inp = sotoinpcb(so);
706 	if (inp == 0) {
707 		INP_INFO_RUNLOCK(&ripcbinfo);
708 		return EINVAL;
709 	}
710 	INP_LOCK(inp);
711 	INP_INFO_RUNLOCK(&ripcbinfo);
712 	socantsendmore(so);
713 	INP_UNLOCK(inp);
714 	return 0;
715 }
716 
717 static int
718 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
719 	 struct mbuf *control, struct thread *td)
720 {
721 	struct inpcb *inp;
722 	u_long dst;
723 	int ret;
724 
725 	INP_INFO_WLOCK(&ripcbinfo);
726 	inp = sotoinpcb(so);
727 	if (so->so_state & SS_ISCONNECTED) {
728 		if (nam) {
729 			INP_INFO_WUNLOCK(&ripcbinfo);
730 			m_freem(m);
731 			return EISCONN;
732 		}
733 		dst = inp->inp_faddr.s_addr;
734 	} else {
735 		if (nam == NULL) {
736 			INP_INFO_WUNLOCK(&ripcbinfo);
737 			m_freem(m);
738 			return ENOTCONN;
739 		}
740 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
741 	}
742 	INP_LOCK(inp);
743 	ret = rip_output(m, so, dst);
744 	INP_UNLOCK(inp);
745 	INP_INFO_WUNLOCK(&ripcbinfo);
746 	return ret;
747 }
748 
749 static int
750 rip_pcblist(SYSCTL_HANDLER_ARGS)
751 {
752 	int error, i, n;
753 	struct inpcb *inp, **inp_list;
754 	inp_gen_t gencnt;
755 	struct xinpgen xig;
756 
757 	/*
758 	 * The process of preparing the TCB list is too time-consuming and
759 	 * resource-intensive to repeat twice on every request.
760 	 */
761 	if (req->oldptr == 0) {
762 		n = ripcbinfo.ipi_count;
763 		req->oldidx = 2 * (sizeof xig)
764 			+ (n + n/8) * sizeof(struct xinpcb);
765 		return 0;
766 	}
767 
768 	if (req->newptr != 0)
769 		return EPERM;
770 
771 	/*
772 	 * OK, now we're committed to doing something.
773 	 */
774 	INP_INFO_RLOCK(&ripcbinfo);
775 	gencnt = ripcbinfo.ipi_gencnt;
776 	n = ripcbinfo.ipi_count;
777 	INP_INFO_RUNLOCK(&ripcbinfo);
778 
779 	xig.xig_len = sizeof xig;
780 	xig.xig_count = n;
781 	xig.xig_gen = gencnt;
782 	xig.xig_sogen = so_gencnt;
783 	error = SYSCTL_OUT(req, &xig, sizeof xig);
784 	if (error)
785 		return error;
786 
787 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
788 	if (inp_list == 0)
789 		return ENOMEM;
790 
791 	INP_INFO_RLOCK(&ripcbinfo);
792 	for (inp = LIST_FIRST(ripcbinfo.listhead), i = 0; inp && i < n;
793 	     inp = LIST_NEXT(inp, inp_list)) {
794 		INP_LOCK(inp);
795 		if (inp->inp_gencnt <= gencnt &&
796 		    cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) {
797 			/* XXX held references? */
798 			inp_list[i++] = inp;
799 		}
800 		INP_UNLOCK(inp);
801 	}
802 	INP_INFO_RUNLOCK(&ripcbinfo);
803 	n = i;
804 
805 	error = 0;
806 	for (i = 0; i < n; i++) {
807 		inp = inp_list[i];
808 		if (inp->inp_gencnt <= gencnt) {
809 			struct xinpcb xi;
810 			xi.xi_len = sizeof xi;
811 			/* XXX should avoid extra copy */
812 			bcopy(inp, &xi.xi_inp, sizeof *inp);
813 			if (inp->inp_socket)
814 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
815 			error = SYSCTL_OUT(req, &xi, sizeof xi);
816 		}
817 	}
818 	if (!error) {
819 		/*
820 		 * Give the user an updated idea of our state.
821 		 * If the generation differs from what we told
822 		 * her before, she knows that something happened
823 		 * while we were processing this request, and it
824 		 * might be necessary to retry.
825 		 */
826 		INP_INFO_RLOCK(&ripcbinfo);
827 		xig.xig_gen = ripcbinfo.ipi_gencnt;
828 		xig.xig_sogen = so_gencnt;
829 		xig.xig_count = ripcbinfo.ipi_count;
830 		INP_INFO_RUNLOCK(&ripcbinfo);
831 		error = SYSCTL_OUT(req, &xig, sizeof xig);
832 	}
833 	free(inp_list, M_TEMP);
834 	return error;
835 }
836 
837 /*
838  * This is the wrapper function for in_setsockaddr.  We just pass down
839  * the pcbinfo for in_setpeeraddr to lock.
840  */
841 static int
842 rip_sockaddr(struct socket *so, struct sockaddr **nam)
843 {
844 	return (in_setsockaddr(so, nam, &ripcbinfo));
845 }
846 
847 /*
848  * This is the wrapper function for in_setpeeraddr.  We just pass down
849  * the pcbinfo for in_setpeeraddr to lock.
850  */
851 static int
852 rip_peeraddr(struct socket *so, struct sockaddr **nam)
853 {
854 	return (in_setpeeraddr(so, nam, &ripcbinfo));
855 }
856 
857 
858 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0,
859 	    rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
860 
861 struct pr_usrreqs rip_usrreqs = {
862 	rip_abort, pru_accept_notsupp, rip_attach, rip_bind, rip_connect,
863 	pru_connect2_notsupp, in_control, rip_detach, rip_disconnect,
864 	pru_listen_notsupp, rip_peeraddr, pru_rcvd_notsupp,
865 	pru_rcvoob_notsupp, rip_send, pru_sense_null, rip_shutdown,
866 	rip_sockaddr, sosend, soreceive, sopoll, in_pcbsosetlabel
867 };
868