xref: /freebsd/sys/netinet/raw_ip.c (revision 3d11b6c8f01e1fca5936a11d6996448467851a94)
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/mac.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.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 struct	inpcbhead ripcb;
78 struct	inpcbinfo ripcbinfo;
79 
80 /* control hooks for ipfw and dummynet */
81 ip_fw_ctl_t *ip_fw_ctl_ptr = NULL;
82 ip_dn_ctl_t *ip_dn_ctl_ptr = NULL;
83 
84 /*
85  * hooks for multicast routing. They all default to NULL,
86  * so leave them not initialized and rely on BSS being set to 0.
87  */
88 
89 /* The socket used to communicate with the multicast routing daemon.  */
90 struct socket  *ip_mrouter;
91 
92 /* The various mrouter and rsvp functions */
93 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
94 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
95 int (*ip_mrouter_done)(void);
96 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
97 		   struct ip_moptions *);
98 int (*mrt_ioctl)(int, caddr_t);
99 int (*legal_vif_num)(int);
100 u_long (*ip_mcast_src)(int);
101 
102 void (*rsvp_input_p)(struct mbuf *m, int off);
103 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
104 void (*ip_rsvp_force_done)(struct socket *);
105 
106 /*
107  * Nominal space allocated to a raw ip socket.
108  */
109 #define	RIPSNDQ		8192
110 #define	RIPRCVQ		8192
111 
112 /*
113  * Raw interface to IP protocol.
114  */
115 
116 /*
117  * Initialize raw connection block q.
118  */
119 void
120 rip_init()
121 {
122 	INP_INFO_LOCK_INIT(&ripcbinfo, "rip");
123 	LIST_INIT(&ripcb);
124 	ripcbinfo.listhead = &ripcb;
125 	/*
126 	 * XXX We don't use the hash list for raw IP, but it's easier
127 	 * to allocate a one entry hash list than it is to check all
128 	 * over the place for hashbase == NULL.
129 	 */
130 	ripcbinfo.hashbase = hashinit(1, M_PCB, &ripcbinfo.hashmask);
131 	ripcbinfo.porthashbase = hashinit(1, M_PCB, &ripcbinfo.porthashmask);
132 	ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb),
133 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
134 	uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets);
135 }
136 
137 static struct	sockaddr_in ripsrc = { sizeof(ripsrc), AF_INET };
138 
139 static int
140 raw_append(struct inpcb *last, struct ip *ip, struct mbuf *n)
141 {
142 	int policyfail = 0;
143 
144 	INP_LOCK_ASSERT(last);
145 
146 #if defined(IPSEC) || defined(FAST_IPSEC)
147 	/* check AH/ESP integrity. */
148 	if (ipsec4_in_reject(n, last)) {
149 		policyfail = 1;
150 #ifdef IPSEC
151 		ipsecstat.in_polvio++;
152 #endif /*IPSEC*/
153 		/* do not inject data to pcb */
154 	}
155 #endif /*IPSEC || FAST_IPSEC*/
156 #ifdef MAC
157 	if (!policyfail && mac_check_inpcb_deliver(last, n) != 0)
158 		policyfail = 1;
159 #endif
160 	/* Check the minimum TTL for socket. */
161 	if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
162 		policyfail = 1;
163 	if (!policyfail) {
164 		struct mbuf *opts = NULL;
165 		struct socket *so;
166 
167 		so = last->inp_socket;
168 		if ((last->inp_flags & INP_CONTROLOPTS) ||
169 		    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
170 			ip_savecontrol(last, &opts, ip, n);
171 		SOCKBUF_LOCK(&so->so_rcv);
172 		if (sbappendaddr_locked(&so->so_rcv,
173 		    (struct sockaddr *)&ripsrc, n, opts) == 0) {
174 			/* should notify about lost packet */
175 			m_freem(n);
176 			if (opts)
177 				m_freem(opts);
178 			SOCKBUF_UNLOCK(&so->so_rcv);
179 		} else
180 			sorwakeup_locked(so);
181 	} else
182 		m_freem(n);
183 	return policyfail;
184 }
185 
186 /*
187  * Setup generic address and protocol structures
188  * for raw_input routine, then pass them along with
189  * mbuf chain.
190  */
191 void
192 rip_input(struct mbuf *m, int off)
193 {
194 	struct ip *ip = mtod(m, struct ip *);
195 	int proto = ip->ip_p;
196 	struct inpcb *inp, *last;
197 
198 	INP_INFO_RLOCK(&ripcbinfo);
199 	ripsrc.sin_addr = ip->ip_src;
200 	last = NULL;
201 	LIST_FOREACH(inp, &ripcb, inp_list) {
202 		INP_LOCK(inp);
203 		if (inp->inp_ip_p && inp->inp_ip_p != proto) {
204 	docontinue:
205 			INP_UNLOCK(inp);
206 			continue;
207 		}
208 #ifdef INET6
209 		if ((inp->inp_vflag & INP_IPV4) == 0)
210 			goto docontinue;
211 #endif
212 		if (inp->inp_laddr.s_addr &&
213 		    inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
214 			goto docontinue;
215 		if (inp->inp_faddr.s_addr &&
216 		    inp->inp_faddr.s_addr != ip->ip_src.s_addr)
217 			goto docontinue;
218 		if (jailed(inp->inp_socket->so_cred))
219 			if (htonl(prison_getip(inp->inp_socket->so_cred)) !=
220 			    ip->ip_dst.s_addr)
221 				goto docontinue;
222 		if (last) {
223 			struct mbuf *n;
224 
225 			n = m_copy(m, 0, (int)M_COPYALL);
226 			if (n != NULL)
227 				(void) raw_append(last, ip, n);
228 			/* XXX count dropped packet */
229 			INP_UNLOCK(last);
230 		}
231 		last = inp;
232 	}
233 	if (last != NULL) {
234 		if (raw_append(last, ip, m) != 0)
235 			ipstat.ips_delivered--;
236 		INP_UNLOCK(last);
237 	} else {
238 		m_freem(m);
239 		ipstat.ips_noproto++;
240 		ipstat.ips_delivered--;
241 	}
242 	INP_INFO_RUNLOCK(&ripcbinfo);
243 }
244 
245 /*
246  * Generate IP header and pass packet to ip_output.
247  * Tack on options user may have setup with control call.
248  */
249 int
250 rip_output(struct mbuf *m, struct socket *so, u_long dst)
251 {
252 	struct ip *ip;
253 	int error;
254 	struct inpcb *inp = sotoinpcb(so);
255 	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
256 	    IP_ALLOWBROADCAST;
257 
258 	/*
259 	 * If the user handed us a complete IP packet, use it.
260 	 * Otherwise, allocate an mbuf for a header and fill it in.
261 	 */
262 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
263 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
264 			m_freem(m);
265 			return(EMSGSIZE);
266 		}
267 		M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
268 		if (m == NULL)
269 			return(ENOBUFS);
270 
271 		INP_LOCK(inp);
272 		ip = mtod(m, struct ip *);
273 		ip->ip_tos = inp->inp_ip_tos;
274 		if (inp->inp_flags & INP_DONTFRAG)
275 			ip->ip_off = IP_DF;
276 		else
277 			ip->ip_off = 0;
278 		ip->ip_p = inp->inp_ip_p;
279 		ip->ip_len = m->m_pkthdr.len;
280 		if (jailed(inp->inp_socket->so_cred))
281 			ip->ip_src.s_addr =
282 			    htonl(prison_getip(inp->inp_socket->so_cred));
283 		else
284 			ip->ip_src = inp->inp_laddr;
285 		ip->ip_dst.s_addr = dst;
286 		ip->ip_ttl = inp->inp_ip_ttl;
287 	} else {
288 		if (m->m_pkthdr.len > IP_MAXPACKET) {
289 			m_freem(m);
290 			return(EMSGSIZE);
291 		}
292 		INP_LOCK(inp);
293 		ip = mtod(m, struct ip *);
294 		if (jailed(inp->inp_socket->so_cred)) {
295 			if (ip->ip_src.s_addr !=
296 			    htonl(prison_getip(inp->inp_socket->so_cred))) {
297 				INP_UNLOCK(inp);
298 				m_freem(m);
299 				return (EPERM);
300 			}
301 		}
302 		/* don't allow both user specified and setsockopt options,
303 		   and don't allow packet length sizes that will crash */
304 		if (((ip->ip_hl != (sizeof (*ip) >> 2))
305 		     && inp->inp_options)
306 		    || (ip->ip_len > m->m_pkthdr.len)
307 		    || (ip->ip_len < (ip->ip_hl << 2))) {
308 			INP_UNLOCK(inp);
309 			m_freem(m);
310 			return EINVAL;
311 		}
312 		if (ip->ip_id == 0)
313 			ip->ip_id = ip_newid();
314 		/* XXX prevent ip_output from overwriting header fields */
315 		flags |= IP_RAWOUTPUT;
316 		ipstat.ips_rawout++;
317 	}
318 
319 	if (inp->inp_vflag & INP_ONESBCAST)
320 		flags |= IP_SENDONES;
321 
322 #ifdef MAC
323 	mac_create_mbuf_from_inpcb(inp, m);
324 #endif
325 
326 	error = ip_output(m, inp->inp_options, NULL, flags,
327 	    inp->inp_moptions, inp);
328 	INP_UNLOCK(inp);
329 	return error;
330 }
331 
332 /*
333  * Raw IP socket option processing.
334  *
335  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
336  * only be created by a privileged process, and as such, socket option
337  * operations to manage system properties on any raw socket were allowed to
338  * take place without explicit additional access control checks.  However,
339  * raw sockets can now also be created in jail(), and therefore explicit
340  * checks are now required.  Likewise, raw sockets can be used by a process
341  * after it gives up privilege, so some caution is required.  For options
342  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
343  * performed in ip_ctloutput() and therefore no check occurs here.
344  * Unilaterally checking suser() here breaks normal IP socket option
345  * operations on raw sockets.
346  *
347  * When adding new socket options here, make sure to add access control
348  * checks here as necessary.
349  */
350 int
351 rip_ctloutput(struct socket *so, struct sockopt *sopt)
352 {
353 	struct	inpcb *inp = sotoinpcb(so);
354 	int	error, optval;
355 
356 	if (sopt->sopt_level != IPPROTO_IP)
357 		return (EINVAL);
358 
359 	error = 0;
360 	switch (sopt->sopt_dir) {
361 	case SOPT_GET:
362 		switch (sopt->sopt_name) {
363 		case IP_HDRINCL:
364 			optval = inp->inp_flags & INP_HDRINCL;
365 			error = sooptcopyout(sopt, &optval, sizeof optval);
366 			break;
367 
368 		case IP_FW_ADD:	/* ADD actually returns the body... */
369 		case IP_FW_GET:
370 		case IP_FW_TABLE_GETSIZE:
371 		case IP_FW_TABLE_LIST:
372 			error = suser(curthread);
373 			if (error != 0)
374 				return (error);
375 			if (ip_fw_ctl_ptr != NULL)
376 				error = ip_fw_ctl_ptr(sopt);
377 			else
378 				error = ENOPROTOOPT;
379 			break;
380 
381 		case IP_DUMMYNET_GET:
382 			error = suser(curthread);
383 			if (error != 0)
384 				return (error);
385 			if (ip_dn_ctl_ptr != NULL)
386 				error = ip_dn_ctl_ptr(sopt);
387 			else
388 				error = ENOPROTOOPT;
389 			break ;
390 
391 		case MRT_INIT:
392 		case MRT_DONE:
393 		case MRT_ADD_VIF:
394 		case MRT_DEL_VIF:
395 		case MRT_ADD_MFC:
396 		case MRT_DEL_MFC:
397 		case MRT_VERSION:
398 		case MRT_ASSERT:
399 		case MRT_API_SUPPORT:
400 		case MRT_API_CONFIG:
401 		case MRT_ADD_BW_UPCALL:
402 		case MRT_DEL_BW_UPCALL:
403 			error = suser(curthread);
404 			if (error != 0)
405 				return (error);
406 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
407 				EOPNOTSUPP;
408 			break;
409 
410 		default:
411 			error = ip_ctloutput(so, sopt);
412 			break;
413 		}
414 		break;
415 
416 	case SOPT_SET:
417 		switch (sopt->sopt_name) {
418 		case IP_HDRINCL:
419 			error = sooptcopyin(sopt, &optval, sizeof optval,
420 					    sizeof optval);
421 			if (error)
422 				break;
423 			if (optval)
424 				inp->inp_flags |= INP_HDRINCL;
425 			else
426 				inp->inp_flags &= ~INP_HDRINCL;
427 			break;
428 
429 		case IP_FW_ADD:
430 		case IP_FW_DEL:
431 		case IP_FW_FLUSH:
432 		case IP_FW_ZERO:
433 		case IP_FW_RESETLOG:
434 		case IP_FW_TABLE_ADD:
435 		case IP_FW_TABLE_DEL:
436 		case IP_FW_TABLE_FLUSH:
437 			error = suser(curthread);
438 			if (error != 0)
439 				return (error);
440 			if (ip_fw_ctl_ptr != NULL)
441 				error = ip_fw_ctl_ptr(sopt);
442 			else
443 				error = ENOPROTOOPT;
444 			break;
445 
446 		case IP_DUMMYNET_CONFIGURE:
447 		case IP_DUMMYNET_DEL:
448 		case IP_DUMMYNET_FLUSH:
449 			error = suser(curthread);
450 			if (error != 0)
451 				return (error);
452 			if (ip_dn_ctl_ptr != NULL)
453 				error = ip_dn_ctl_ptr(sopt);
454 			else
455 				error = ENOPROTOOPT ;
456 			break ;
457 
458 		case IP_RSVP_ON:
459 			error = suser(curthread);
460 			if (error != 0)
461 				return (error);
462 			error = ip_rsvp_init(so);
463 			break;
464 
465 		case IP_RSVP_OFF:
466 			error = suser(curthread);
467 			if (error != 0)
468 				return (error);
469 			error = ip_rsvp_done();
470 			break;
471 
472 		case IP_RSVP_VIF_ON:
473 		case IP_RSVP_VIF_OFF:
474 			error = suser(curthread);
475 			if (error != 0)
476 				return (error);
477 			error = ip_rsvp_vif ?
478 				ip_rsvp_vif(so, sopt) : EINVAL;
479 			break;
480 
481 		case MRT_INIT:
482 		case MRT_DONE:
483 		case MRT_ADD_VIF:
484 		case MRT_DEL_VIF:
485 		case MRT_ADD_MFC:
486 		case MRT_DEL_MFC:
487 		case MRT_VERSION:
488 		case MRT_ASSERT:
489 		case MRT_API_SUPPORT:
490 		case MRT_API_CONFIG:
491 		case MRT_ADD_BW_UPCALL:
492 		case MRT_DEL_BW_UPCALL:
493 			error = suser(curthread);
494 			if (error != 0)
495 				return (error);
496 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
497 					EOPNOTSUPP;
498 			break;
499 
500 		default:
501 			error = ip_ctloutput(so, sopt);
502 			break;
503 		}
504 		break;
505 	}
506 
507 	return (error);
508 }
509 
510 /*
511  * This function exists solely to receive the PRC_IFDOWN messages which
512  * are sent by if_down().  It looks for an ifaddr whose ifa_addr is sa,
513  * and calls in_ifadown() to remove all routes corresponding to that address.
514  * It also receives the PRC_IFUP messages from if_up() and reinstalls the
515  * interface routes.
516  */
517 void
518 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
519 {
520 	struct in_ifaddr *ia;
521 	struct ifnet *ifp;
522 	int err;
523 	int flags;
524 
525 	switch (cmd) {
526 	case PRC_IFDOWN:
527 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
528 			if (ia->ia_ifa.ifa_addr == sa
529 			    && (ia->ia_flags & IFA_ROUTE)) {
530 				/*
531 				 * in_ifscrub kills the interface route.
532 				 */
533 				in_ifscrub(ia->ia_ifp, ia);
534 				/*
535 				 * in_ifadown gets rid of all the rest of
536 				 * the routes.  This is not quite the right
537 				 * thing to do, but at least if we are running
538 				 * a routing process they will come back.
539 				 */
540 				in_ifadown(&ia->ia_ifa, 0);
541 				break;
542 			}
543 		}
544 		break;
545 
546 	case PRC_IFUP:
547 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
548 			if (ia->ia_ifa.ifa_addr == sa)
549 				break;
550 		}
551 		if (ia == 0 || (ia->ia_flags & IFA_ROUTE))
552 			return;
553 		flags = RTF_UP;
554 		ifp = ia->ia_ifa.ifa_ifp;
555 
556 		if ((ifp->if_flags & IFF_LOOPBACK)
557 		    || (ifp->if_flags & IFF_POINTOPOINT))
558 			flags |= RTF_HOST;
559 
560 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
561 		if (err == 0)
562 			ia->ia_flags |= IFA_ROUTE;
563 		break;
564 	}
565 }
566 
567 u_long	rip_sendspace = RIPSNDQ;
568 u_long	rip_recvspace = RIPRCVQ;
569 
570 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
571     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
572 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
573     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
574 
575 static int
576 rip_attach(struct socket *so, int proto, struct thread *td)
577 {
578 	struct inpcb *inp;
579 	int error;
580 
581 	inp = sotoinpcb(so);
582 	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
583 	if (jailed(td->td_ucred) && !jail_allow_raw_sockets)
584 		return (EPERM);
585 	if ((error = suser_cred(td->td_ucred, SUSER_ALLOWJAIL)) != 0)
586 		return error;
587 	if (proto >= IPPROTO_MAX || proto < 0)
588 		return EPROTONOSUPPORT;
589 	error = soreserve(so, rip_sendspace, rip_recvspace);
590 	if (error)
591 		return error;
592 	INP_INFO_WLOCK(&ripcbinfo);
593 	error = in_pcballoc(so, &ripcbinfo, "rawinp");
594 	if (error) {
595 		INP_INFO_WUNLOCK(&ripcbinfo);
596 		return error;
597 	}
598 	inp = (struct inpcb *)so->so_pcb;
599 	INP_LOCK(inp);
600 	INP_INFO_WUNLOCK(&ripcbinfo);
601 	inp->inp_vflag |= INP_IPV4;
602 	inp->inp_ip_p = proto;
603 	inp->inp_ip_ttl = ip_defttl;
604 	INP_UNLOCK(inp);
605 	return 0;
606 }
607 
608 static void
609 rip_pcbdetach(struct socket *so, struct inpcb *inp)
610 {
611 
612 	INP_INFO_WLOCK_ASSERT(&ripcbinfo);
613 	INP_LOCK_ASSERT(inp);
614 
615 	if (so == ip_mrouter && ip_mrouter_done)
616 		ip_mrouter_done();
617 	if (ip_rsvp_force_done)
618 		ip_rsvp_force_done(so);
619 	if (so == ip_rsvpd)
620 		ip_rsvp_done();
621 	in_pcbdetach(inp);
622 	in_pcbfree(inp);
623 }
624 
625 static void
626 rip_detach(struct socket *so)
627 {
628 	struct inpcb *inp;
629 
630 	inp = sotoinpcb(so);
631 	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
632 	INP_INFO_WLOCK(&ripcbinfo);
633 	INP_LOCK(inp);
634 	rip_pcbdetach(so, inp);
635 	INP_INFO_WUNLOCK(&ripcbinfo);
636 }
637 
638 static void
639 rip_abort(struct socket *so)
640 {
641 	struct inpcb *inp;
642 
643 	inp = sotoinpcb(so);
644 	KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
645 	INP_INFO_WLOCK(&ripcbinfo);
646 	INP_LOCK(inp);
647 	soisdisconnected(so);
648 	rip_pcbdetach(so, inp);
649 	INP_INFO_WUNLOCK(&ripcbinfo);
650 }
651 
652 static int
653 rip_disconnect(struct socket *so)
654 {
655 	if ((so->so_state & SS_ISCONNECTED) == 0)
656 		return ENOTCONN;
657 	rip_abort(so);
658 	return (0);
659 }
660 
661 static int
662 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
663 {
664 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
665 	struct inpcb *inp;
666 
667 	if (nam->sa_len != sizeof(*addr))
668 		return EINVAL;
669 
670 	if (jailed(td->td_ucred)) {
671 		if (addr->sin_addr.s_addr == INADDR_ANY)
672 			addr->sin_addr.s_addr =
673 			    htonl(prison_getip(td->td_ucred));
674 		if (htonl(prison_getip(td->td_ucred)) != addr->sin_addr.s_addr)
675 			return (EADDRNOTAVAIL);
676 	}
677 
678 	if (TAILQ_EMPTY(&ifnet) ||
679 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
680 	    (addr->sin_addr.s_addr &&
681 	     ifa_ifwithaddr((struct sockaddr *)addr) == 0))
682 		return EADDRNOTAVAIL;
683 
684 	inp = sotoinpcb(so);
685 	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
686 	INP_INFO_WLOCK(&ripcbinfo);
687 	INP_LOCK(inp);
688 	inp->inp_laddr = addr->sin_addr;
689 	INP_UNLOCK(inp);
690 	INP_INFO_WUNLOCK(&ripcbinfo);
691 	return 0;
692 }
693 
694 static int
695 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
696 {
697 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
698 	struct inpcb *inp;
699 
700 	if (nam->sa_len != sizeof(*addr))
701 		return EINVAL;
702 	if (TAILQ_EMPTY(&ifnet))
703 		return EADDRNOTAVAIL;
704 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
705 		return EAFNOSUPPORT;
706 
707 	inp = sotoinpcb(so);
708 	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
709 	INP_INFO_WLOCK(&ripcbinfo);
710 	INP_LOCK(inp);
711 	inp->inp_faddr = addr->sin_addr;
712 	soisconnected(so);
713 	INP_UNLOCK(inp);
714 	INP_INFO_WUNLOCK(&ripcbinfo);
715 	return 0;
716 }
717 
718 static int
719 rip_shutdown(struct socket *so)
720 {
721 	struct inpcb *inp;
722 
723 	inp = sotoinpcb(so);
724 	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
725 	INP_LOCK(inp);
726 	socantsendmore(so);
727 	INP_UNLOCK(inp);
728 	return 0;
729 }
730 
731 static int
732 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
733 	 struct mbuf *control, struct thread *td)
734 {
735 	struct inpcb *inp;
736 	u_long dst;
737 
738 	inp = sotoinpcb(so);
739 	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
740 	/*
741 	 * Note: 'dst' reads below are unlocked.
742 	 */
743 	if (so->so_state & SS_ISCONNECTED) {
744 		if (nam) {
745 			m_freem(m);
746 			return EISCONN;
747 		}
748 		dst = inp->inp_faddr.s_addr;	/* Unlocked read. */
749 	} else {
750 		if (nam == NULL) {
751 			m_freem(m);
752 			return ENOTCONN;
753 		}
754 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
755 	}
756 	return rip_output(m, so, dst);
757 }
758 
759 static int
760 rip_pcblist(SYSCTL_HANDLER_ARGS)
761 {
762 	int error, i, n;
763 	struct inpcb *inp, **inp_list;
764 	inp_gen_t gencnt;
765 	struct xinpgen xig;
766 
767 	/*
768 	 * The process of preparing the TCB list is too time-consuming and
769 	 * resource-intensive to repeat twice on every request.
770 	 */
771 	if (req->oldptr == 0) {
772 		n = ripcbinfo.ipi_count;
773 		req->oldidx = 2 * (sizeof xig)
774 			+ (n + n/8) * sizeof(struct xinpcb);
775 		return 0;
776 	}
777 
778 	if (req->newptr != 0)
779 		return EPERM;
780 
781 	/*
782 	 * OK, now we're committed to doing something.
783 	 */
784 	INP_INFO_RLOCK(&ripcbinfo);
785 	gencnt = ripcbinfo.ipi_gencnt;
786 	n = ripcbinfo.ipi_count;
787 	INP_INFO_RUNLOCK(&ripcbinfo);
788 
789 	xig.xig_len = sizeof xig;
790 	xig.xig_count = n;
791 	xig.xig_gen = gencnt;
792 	xig.xig_sogen = so_gencnt;
793 	error = SYSCTL_OUT(req, &xig, sizeof xig);
794 	if (error)
795 		return error;
796 
797 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
798 	if (inp_list == 0)
799 		return ENOMEM;
800 
801 	INP_INFO_RLOCK(&ripcbinfo);
802 	for (inp = LIST_FIRST(ripcbinfo.listhead), i = 0; inp && i < n;
803 	     inp = LIST_NEXT(inp, inp_list)) {
804 		INP_LOCK(inp);
805 		if (inp->inp_gencnt <= gencnt &&
806 		    cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) {
807 			/* XXX held references? */
808 			inp_list[i++] = inp;
809 		}
810 		INP_UNLOCK(inp);
811 	}
812 	INP_INFO_RUNLOCK(&ripcbinfo);
813 	n = i;
814 
815 	error = 0;
816 	for (i = 0; i < n; i++) {
817 		inp = inp_list[i];
818 		if (inp->inp_gencnt <= gencnt) {
819 			struct xinpcb xi;
820 			bzero(&xi, sizeof(xi));
821 			xi.xi_len = sizeof xi;
822 			/* XXX should avoid extra copy */
823 			bcopy(inp, &xi.xi_inp, sizeof *inp);
824 			if (inp->inp_socket)
825 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
826 			error = SYSCTL_OUT(req, &xi, sizeof xi);
827 		}
828 	}
829 	if (!error) {
830 		/*
831 		 * Give the user an updated idea of our state.
832 		 * If the generation differs from what we told
833 		 * her before, she knows that something happened
834 		 * while we were processing this request, and it
835 		 * might be necessary to retry.
836 		 */
837 		INP_INFO_RLOCK(&ripcbinfo);
838 		xig.xig_gen = ripcbinfo.ipi_gencnt;
839 		xig.xig_sogen = so_gencnt;
840 		xig.xig_count = ripcbinfo.ipi_count;
841 		INP_INFO_RUNLOCK(&ripcbinfo);
842 		error = SYSCTL_OUT(req, &xig, sizeof xig);
843 	}
844 	free(inp_list, M_TEMP);
845 	return error;
846 }
847 
848 /*
849  * This is the wrapper function for in_setsockaddr.  We just pass down
850  * the pcbinfo for in_setpeeraddr to lock.
851  */
852 static int
853 rip_sockaddr(struct socket *so, struct sockaddr **nam)
854 {
855 	return (in_setsockaddr(so, nam, &ripcbinfo));
856 }
857 
858 /*
859  * This is the wrapper function for in_setpeeraddr.  We just pass down
860  * the pcbinfo for in_setpeeraddr to lock.
861  */
862 static int
863 rip_peeraddr(struct socket *so, struct sockaddr **nam)
864 {
865 	return (in_setpeeraddr(so, nam, &ripcbinfo));
866 }
867 
868 
869 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0,
870 	    rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
871 
872 struct pr_usrreqs rip_usrreqs = {
873 	.pru_abort =		rip_abort,
874 	.pru_attach =		rip_attach,
875 	.pru_bind =		rip_bind,
876 	.pru_connect =		rip_connect,
877 	.pru_control =		in_control,
878 	.pru_detach =		rip_detach,
879 	.pru_disconnect =	rip_disconnect,
880 	.pru_peeraddr =		rip_peeraddr,
881 	.pru_send =		rip_send,
882 	.pru_shutdown =		rip_shutdown,
883 	.pru_sockaddr =		rip_sockaddr,
884 	.pru_sosetlabel =	in_pcbsosetlabel
885 };
886