xref: /freebsd/sys/netinet/raw_ip.c (revision d429ea332342fcb98d27a350d0c4944bf9aec3f9)
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 	if (!policyfail) {
161 		struct mbuf *opts = NULL;
162 		struct socket *so;
163 
164 		so = last->inp_socket;
165 		if ((last->inp_flags & INP_CONTROLOPTS) ||
166 		    (so->so_options & SO_TIMESTAMP))
167 			ip_savecontrol(last, &opts, ip, n);
168 		SOCKBUF_LOCK(&so->so_rcv);
169 		if (sbappendaddr_locked(&so->so_rcv,
170 		    (struct sockaddr *)&ripsrc, n, opts) == 0) {
171 			/* should notify about lost packet */
172 			m_freem(n);
173 			if (opts)
174 				m_freem(opts);
175 			SOCKBUF_UNLOCK(&so->so_rcv);
176 		} else
177 			sorwakeup_locked(so);
178 	} else
179 		m_freem(n);
180 	return policyfail;
181 }
182 
183 /*
184  * Setup generic address and protocol structures
185  * for raw_input routine, then pass them along with
186  * mbuf chain.
187  */
188 void
189 rip_input(struct mbuf *m, int off)
190 {
191 	struct ip *ip = mtod(m, struct ip *);
192 	int proto = ip->ip_p;
193 	struct inpcb *inp, *last;
194 
195 	INP_INFO_RLOCK(&ripcbinfo);
196 	ripsrc.sin_addr = ip->ip_src;
197 	last = NULL;
198 	LIST_FOREACH(inp, &ripcb, inp_list) {
199 		INP_LOCK(inp);
200 		if (inp->inp_ip_p && inp->inp_ip_p != proto) {
201 	docontinue:
202 			INP_UNLOCK(inp);
203 			continue;
204 		}
205 #ifdef INET6
206 		if ((inp->inp_vflag & INP_IPV4) == 0)
207 			goto docontinue;
208 #endif
209 		if (inp->inp_laddr.s_addr &&
210 		    inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
211 			goto docontinue;
212 		if (inp->inp_faddr.s_addr &&
213 		    inp->inp_faddr.s_addr != ip->ip_src.s_addr)
214 			goto docontinue;
215 		if (jailed(inp->inp_socket->so_cred))
216 			if (htonl(prison_getip(inp->inp_socket->so_cred)) !=
217 			    ip->ip_dst.s_addr)
218 				goto docontinue;
219 		if (last) {
220 			struct mbuf *n;
221 
222 			n = m_copy(m, 0, (int)M_COPYALL);
223 			if (n != NULL)
224 				(void) raw_append(last, ip, n);
225 			/* XXX count dropped packet */
226 			INP_UNLOCK(last);
227 		}
228 		last = inp;
229 	}
230 	if (last != NULL) {
231 		if (raw_append(last, ip, m) != 0)
232 			ipstat.ips_delivered--;
233 		INP_UNLOCK(last);
234 	} else {
235 		m_freem(m);
236 		ipstat.ips_noproto++;
237 		ipstat.ips_delivered--;
238 	}
239 	INP_INFO_RUNLOCK(&ripcbinfo);
240 }
241 
242 /*
243  * Generate IP header and pass packet to ip_output.
244  * Tack on options user may have setup with control call.
245  */
246 int
247 rip_output(struct mbuf *m, struct socket *so, u_long dst)
248 {
249 	struct ip *ip;
250 	int error;
251 	struct inpcb *inp = sotoinpcb(so);
252 	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
253 	    IP_ALLOWBROADCAST;
254 
255 	/*
256 	 * If the user handed us a complete IP packet, use it.
257 	 * Otherwise, allocate an mbuf for a header and fill it in.
258 	 */
259 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
260 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
261 			m_freem(m);
262 			return(EMSGSIZE);
263 		}
264 		M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
265 		if (m == NULL)
266 			return(ENOBUFS);
267 
268 		INP_LOCK(inp);
269 		ip = mtod(m, struct ip *);
270 		ip->ip_tos = inp->inp_ip_tos;
271 		ip->ip_off = 0;
272 		ip->ip_p = inp->inp_ip_p;
273 		ip->ip_len = m->m_pkthdr.len;
274 		if (jailed(inp->inp_socket->so_cred))
275 			ip->ip_src.s_addr =
276 			    htonl(prison_getip(inp->inp_socket->so_cred));
277 		else
278 			ip->ip_src = inp->inp_laddr;
279 		ip->ip_dst.s_addr = dst;
280 		ip->ip_ttl = inp->inp_ip_ttl;
281 	} else {
282 		if (m->m_pkthdr.len > IP_MAXPACKET) {
283 			m_freem(m);
284 			return(EMSGSIZE);
285 		}
286 		INP_LOCK(inp);
287 		ip = mtod(m, struct ip *);
288 		if (jailed(inp->inp_socket->so_cred)) {
289 			if (ip->ip_src.s_addr !=
290 			    htonl(prison_getip(inp->inp_socket->so_cred))) {
291 				INP_UNLOCK(inp);
292 				m_freem(m);
293 				return (EPERM);
294 			}
295 		}
296 		/* don't allow both user specified and setsockopt options,
297 		   and don't allow packet length sizes that will crash */
298 		if (((ip->ip_hl != (sizeof (*ip) >> 2))
299 		     && inp->inp_options)
300 		    || (ip->ip_len > m->m_pkthdr.len)
301 		    || (ip->ip_len < (ip->ip_hl << 2))) {
302 			INP_UNLOCK(inp);
303 			m_freem(m);
304 			return EINVAL;
305 		}
306 		if (ip->ip_id == 0)
307 			ip->ip_id = ip_newid();
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 #ifdef MAC
317 	mac_create_mbuf_from_inpcb(inp, m);
318 #endif
319 
320 	error = ip_output(m, inp->inp_options, NULL, flags,
321 	    inp->inp_moptions, inp);
322 	INP_UNLOCK(inp);
323 	return error;
324 }
325 
326 /*
327  * Raw IP socket option processing.
328  *
329  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
330  * only be created by a privileged process, and as such, socket option
331  * operations to manage system properties on any raw socket were allowed to
332  * take place without explicit additional access control checks.  However,
333  * raw sockets can now also be created in jail(), and therefore explicit
334  * checks are now required.  Likewise, raw sockets can be used by a process
335  * after it gives up privilege, so some caution is required.  For options
336  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
337  * performed in ip_ctloutput() and therefore no check occurs here.
338  * Unilaterally checking suser() here breaks normal IP socket option
339  * operations on raw sockets.
340  *
341  * When adding new socket options here, make sure to add access control
342  * checks here as necessary.
343  */
344 int
345 rip_ctloutput(struct socket *so, struct sockopt *sopt)
346 {
347 	struct	inpcb *inp = sotoinpcb(so);
348 	int	error, optval;
349 
350 	if (sopt->sopt_level != IPPROTO_IP)
351 		return (EINVAL);
352 
353 	error = 0;
354 	switch (sopt->sopt_dir) {
355 	case SOPT_GET:
356 		switch (sopt->sopt_name) {
357 		case IP_HDRINCL:
358 			optval = inp->inp_flags & INP_HDRINCL;
359 			error = sooptcopyout(sopt, &optval, sizeof optval);
360 			break;
361 
362 		case IP_FW_ADD:	/* ADD actually returns the body... */
363 		case IP_FW_GET:
364 		case IP_FW_TABLE_GETSIZE:
365 		case IP_FW_TABLE_LIST:
366 			error = suser(curthread);
367 			if (error != 0)
368 				return (error);
369 			if (ip_fw_ctl_ptr != NULL)
370 				error = ip_fw_ctl_ptr(sopt);
371 			else
372 				error = ENOPROTOOPT;
373 			break;
374 
375 		case IP_DUMMYNET_GET:
376 			error = suser(curthread);
377 			if (error != 0)
378 				return (error);
379 			if (ip_dn_ctl_ptr != NULL)
380 				error = ip_dn_ctl_ptr(sopt);
381 			else
382 				error = ENOPROTOOPT;
383 			break ;
384 
385 		case MRT_INIT:
386 		case MRT_DONE:
387 		case MRT_ADD_VIF:
388 		case MRT_DEL_VIF:
389 		case MRT_ADD_MFC:
390 		case MRT_DEL_MFC:
391 		case MRT_VERSION:
392 		case MRT_ASSERT:
393 		case MRT_API_SUPPORT:
394 		case MRT_API_CONFIG:
395 		case MRT_ADD_BW_UPCALL:
396 		case MRT_DEL_BW_UPCALL:
397 			error = suser(curthread);
398 			if (error != 0)
399 				return (error);
400 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
401 				EOPNOTSUPP;
402 			break;
403 
404 		default:
405 			error = ip_ctloutput(so, sopt);
406 			break;
407 		}
408 		break;
409 
410 	case SOPT_SET:
411 		switch (sopt->sopt_name) {
412 		case IP_HDRINCL:
413 			error = sooptcopyin(sopt, &optval, sizeof optval,
414 					    sizeof optval);
415 			if (error)
416 				break;
417 			if (optval)
418 				inp->inp_flags |= INP_HDRINCL;
419 			else
420 				inp->inp_flags &= ~INP_HDRINCL;
421 			break;
422 
423 		case IP_FW_ADD:
424 		case IP_FW_DEL:
425 		case IP_FW_FLUSH:
426 		case IP_FW_ZERO:
427 		case IP_FW_RESETLOG:
428 		case IP_FW_TABLE_ADD:
429 		case IP_FW_TABLE_DEL:
430 		case IP_FW_TABLE_FLUSH:
431 			error = suser(curthread);
432 			if (error != 0)
433 				return (error);
434 			if (ip_fw_ctl_ptr != NULL)
435 				error = ip_fw_ctl_ptr(sopt);
436 			else
437 				error = ENOPROTOOPT;
438 			break;
439 
440 		case IP_DUMMYNET_CONFIGURE:
441 		case IP_DUMMYNET_DEL:
442 		case IP_DUMMYNET_FLUSH:
443 			error = suser(curthread);
444 			if (error != 0)
445 				return (error);
446 			if (ip_dn_ctl_ptr != NULL)
447 				error = ip_dn_ctl_ptr(sopt);
448 			else
449 				error = ENOPROTOOPT ;
450 			break ;
451 
452 		case IP_RSVP_ON:
453 			error = suser(curthread);
454 			if (error != 0)
455 				return (error);
456 			error = ip_rsvp_init(so);
457 			break;
458 
459 		case IP_RSVP_OFF:
460 			error = suser(curthread);
461 			if (error != 0)
462 				return (error);
463 			error = ip_rsvp_done();
464 			break;
465 
466 		case IP_RSVP_VIF_ON:
467 		case IP_RSVP_VIF_OFF:
468 			error = suser(curthread);
469 			if (error != 0)
470 				return (error);
471 			error = ip_rsvp_vif ?
472 				ip_rsvp_vif(so, sopt) : EINVAL;
473 			break;
474 
475 		case MRT_INIT:
476 		case MRT_DONE:
477 		case MRT_ADD_VIF:
478 		case MRT_DEL_VIF:
479 		case MRT_ADD_MFC:
480 		case MRT_DEL_MFC:
481 		case MRT_VERSION:
482 		case MRT_ASSERT:
483 		case MRT_API_SUPPORT:
484 		case MRT_API_CONFIG:
485 		case MRT_ADD_BW_UPCALL:
486 		case MRT_DEL_BW_UPCALL:
487 			error = suser(curthread);
488 			if (error != 0)
489 				return (error);
490 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
491 					EOPNOTSUPP;
492 			break;
493 
494 		default:
495 			error = ip_ctloutput(so, sopt);
496 			break;
497 		}
498 		break;
499 	}
500 
501 	return (error);
502 }
503 
504 /*
505  * This function exists solely to receive the PRC_IFDOWN messages which
506  * are sent by if_down().  It looks for an ifaddr whose ifa_addr is sa,
507  * and calls in_ifadown() to remove all routes corresponding to that address.
508  * It also receives the PRC_IFUP messages from if_up() and reinstalls the
509  * interface routes.
510  */
511 void
512 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
513 {
514 	struct in_ifaddr *ia;
515 	struct ifnet *ifp;
516 	int err;
517 	int flags;
518 
519 	switch (cmd) {
520 	case PRC_IFDOWN:
521 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
522 			if (ia->ia_ifa.ifa_addr == sa
523 			    && (ia->ia_flags & IFA_ROUTE)) {
524 				/*
525 				 * in_ifscrub kills the interface route.
526 				 */
527 				in_ifscrub(ia->ia_ifp, ia);
528 				/*
529 				 * in_ifadown gets rid of all the rest of
530 				 * the routes.  This is not quite the right
531 				 * thing to do, but at least if we are running
532 				 * a routing process they will come back.
533 				 */
534 				in_ifadown(&ia->ia_ifa, 0);
535 				break;
536 			}
537 		}
538 		break;
539 
540 	case PRC_IFUP:
541 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
542 			if (ia->ia_ifa.ifa_addr == sa)
543 				break;
544 		}
545 		if (ia == 0 || (ia->ia_flags & IFA_ROUTE))
546 			return;
547 		flags = RTF_UP;
548 		ifp = ia->ia_ifa.ifa_ifp;
549 
550 		if ((ifp->if_flags & IFF_LOOPBACK)
551 		    || (ifp->if_flags & IFF_POINTOPOINT))
552 			flags |= RTF_HOST;
553 
554 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
555 		if (err == 0)
556 			ia->ia_flags |= IFA_ROUTE;
557 		break;
558 	}
559 }
560 
561 u_long	rip_sendspace = RIPSNDQ;
562 u_long	rip_recvspace = RIPRCVQ;
563 
564 SYSCTL_INT(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
565     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
566 SYSCTL_INT(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
567     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
568 
569 static int
570 rip_attach(struct socket *so, int proto, struct thread *td)
571 {
572 	struct inpcb *inp;
573 	int error;
574 
575 	/* XXX why not lower? */
576 	INP_INFO_WLOCK(&ripcbinfo);
577 	inp = sotoinpcb(so);
578 	if (inp) {
579 		/* XXX counter, printf */
580 		INP_INFO_WUNLOCK(&ripcbinfo);
581 		return EINVAL;
582 	}
583 	if (jailed(td->td_ucred) && !jail_allow_raw_sockets) {
584 		INP_INFO_WUNLOCK(&ripcbinfo);
585 		return (EPERM);
586 	}
587 	if ((error = suser_cred(td->td_ucred, SUSER_ALLOWJAIL)) != 0) {
588 		INP_INFO_WUNLOCK(&ripcbinfo);
589 		return error;
590 	}
591 	if (proto >= IPPROTO_MAX || proto < 0) {
592 		INP_INFO_WUNLOCK(&ripcbinfo);
593 		return EPROTONOSUPPORT;
594 	}
595 
596 	error = soreserve(so, rip_sendspace, rip_recvspace);
597 	if (error) {
598 		INP_INFO_WUNLOCK(&ripcbinfo);
599 		return error;
600 	}
601 	error = in_pcballoc(so, &ripcbinfo, "rawinp");
602 	if (error) {
603 		INP_INFO_WUNLOCK(&ripcbinfo);
604 		return error;
605 	}
606 	inp = (struct inpcb *)so->so_pcb;
607 	INP_LOCK(inp);
608 	INP_INFO_WUNLOCK(&ripcbinfo);
609 	inp->inp_vflag |= INP_IPV4;
610 	inp->inp_ip_p = proto;
611 	inp->inp_ip_ttl = ip_defttl;
612 	INP_UNLOCK(inp);
613 	return 0;
614 }
615 
616 static void
617 rip_pcbdetach(struct socket *so, struct inpcb *inp)
618 {
619 
620 	INP_INFO_WLOCK_ASSERT(&ripcbinfo);
621 	INP_LOCK_ASSERT(inp);
622 
623 	if (so == ip_mrouter && ip_mrouter_done)
624 		ip_mrouter_done();
625 	if (ip_rsvp_force_done)
626 		ip_rsvp_force_done(so);
627 	if (so == ip_rsvpd)
628 		ip_rsvp_done();
629 	in_pcbdetach(inp);
630 }
631 
632 static int
633 rip_detach(struct socket *so)
634 {
635 	struct inpcb *inp;
636 
637 	INP_INFO_WLOCK(&ripcbinfo);
638 	inp = sotoinpcb(so);
639 	if (inp == 0) {
640 		/* XXX counter, printf */
641 		INP_INFO_WUNLOCK(&ripcbinfo);
642 		return EINVAL;
643 	}
644 	INP_LOCK(inp);
645 	rip_pcbdetach(so, inp);
646 	INP_INFO_WUNLOCK(&ripcbinfo);
647 	return 0;
648 }
649 
650 static int
651 rip_abort(struct socket *so)
652 {
653 	struct inpcb *inp;
654 
655 	INP_INFO_WLOCK(&ripcbinfo);
656 	inp = sotoinpcb(so);
657 	if (inp == 0) {
658 		INP_INFO_WUNLOCK(&ripcbinfo);
659 		return EINVAL;	/* ??? possible? panic instead? */
660 	}
661 	INP_LOCK(inp);
662 	soisdisconnected(so);
663 	if (so->so_state & SS_NOFDREF)
664 		rip_pcbdetach(so, inp);
665 	else
666 		INP_UNLOCK(inp);
667 	INP_INFO_WUNLOCK(&ripcbinfo);
668 	return 0;
669 }
670 
671 static int
672 rip_disconnect(struct socket *so)
673 {
674 	if ((so->so_state & SS_ISCONNECTED) == 0)
675 		return ENOTCONN;
676 	return rip_abort(so);
677 }
678 
679 static int
680 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
681 {
682 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
683 	struct inpcb *inp;
684 
685 	if (nam->sa_len != sizeof(*addr))
686 		return EINVAL;
687 
688 	if (jailed(td->td_ucred)) {
689 		if (addr->sin_addr.s_addr == INADDR_ANY)
690 			addr->sin_addr.s_addr =
691 			    htonl(prison_getip(td->td_ucred));
692 		if (htonl(prison_getip(td->td_ucred)) != addr->sin_addr.s_addr)
693 			return (EADDRNOTAVAIL);
694 	}
695 
696 	if (TAILQ_EMPTY(&ifnet) ||
697 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
698 	    (addr->sin_addr.s_addr &&
699 	     ifa_ifwithaddr((struct sockaddr *)addr) == 0))
700 		return EADDRNOTAVAIL;
701 
702 	INP_INFO_WLOCK(&ripcbinfo);
703 	inp = sotoinpcb(so);
704 	if (inp == 0) {
705 		INP_INFO_WUNLOCK(&ripcbinfo);
706 		return EINVAL;
707 	}
708 	INP_LOCK(inp);
709 	inp->inp_laddr = addr->sin_addr;
710 	INP_UNLOCK(inp);
711 	INP_INFO_WUNLOCK(&ripcbinfo);
712 	return 0;
713 }
714 
715 static int
716 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
717 {
718 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
719 	struct inpcb *inp;
720 
721 	if (nam->sa_len != sizeof(*addr))
722 		return EINVAL;
723 	if (TAILQ_EMPTY(&ifnet))
724 		return EADDRNOTAVAIL;
725 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
726 		return EAFNOSUPPORT;
727 
728 	INP_INFO_WLOCK(&ripcbinfo);
729 	inp = sotoinpcb(so);
730 	if (inp == 0) {
731 		INP_INFO_WUNLOCK(&ripcbinfo);
732 		return EINVAL;
733 	}
734 	INP_LOCK(inp);
735 	inp->inp_faddr = addr->sin_addr;
736 	soisconnected(so);
737 	INP_UNLOCK(inp);
738 	INP_INFO_WUNLOCK(&ripcbinfo);
739 	return 0;
740 }
741 
742 static int
743 rip_shutdown(struct socket *so)
744 {
745 	struct inpcb *inp;
746 
747 	INP_INFO_RLOCK(&ripcbinfo);
748 	inp = sotoinpcb(so);
749 	if (inp == 0) {
750 		INP_INFO_RUNLOCK(&ripcbinfo);
751 		return EINVAL;
752 	}
753 	INP_LOCK(inp);
754 	INP_INFO_RUNLOCK(&ripcbinfo);
755 	socantsendmore(so);
756 	INP_UNLOCK(inp);
757 	return 0;
758 }
759 
760 static int
761 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
762 	 struct mbuf *control, struct thread *td)
763 {
764 	struct inpcb *inp;
765 	u_long dst;
766 	int ret;
767 
768 	INP_INFO_WLOCK(&ripcbinfo);
769 	inp = sotoinpcb(so);
770 	if (so->so_state & SS_ISCONNECTED) {
771 		if (nam) {
772 			INP_INFO_WUNLOCK(&ripcbinfo);
773 			m_freem(m);
774 			return EISCONN;
775 		}
776 		dst = inp->inp_faddr.s_addr;
777 	} else {
778 		if (nam == NULL) {
779 			INP_INFO_WUNLOCK(&ripcbinfo);
780 			m_freem(m);
781 			return ENOTCONN;
782 		}
783 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
784 	}
785 	ret = rip_output(m, so, dst);
786 	INP_INFO_WUNLOCK(&ripcbinfo);
787 	return ret;
788 }
789 
790 static int
791 rip_pcblist(SYSCTL_HANDLER_ARGS)
792 {
793 	int error, i, n;
794 	struct inpcb *inp, **inp_list;
795 	inp_gen_t gencnt;
796 	struct xinpgen xig;
797 
798 	/*
799 	 * The process of preparing the TCB list is too time-consuming and
800 	 * resource-intensive to repeat twice on every request.
801 	 */
802 	if (req->oldptr == 0) {
803 		n = ripcbinfo.ipi_count;
804 		req->oldidx = 2 * (sizeof xig)
805 			+ (n + n/8) * sizeof(struct xinpcb);
806 		return 0;
807 	}
808 
809 	if (req->newptr != 0)
810 		return EPERM;
811 
812 	/*
813 	 * OK, now we're committed to doing something.
814 	 */
815 	INP_INFO_RLOCK(&ripcbinfo);
816 	gencnt = ripcbinfo.ipi_gencnt;
817 	n = ripcbinfo.ipi_count;
818 	INP_INFO_RUNLOCK(&ripcbinfo);
819 
820 	xig.xig_len = sizeof xig;
821 	xig.xig_count = n;
822 	xig.xig_gen = gencnt;
823 	xig.xig_sogen = so_gencnt;
824 	error = SYSCTL_OUT(req, &xig, sizeof xig);
825 	if (error)
826 		return error;
827 
828 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
829 	if (inp_list == 0)
830 		return ENOMEM;
831 
832 	INP_INFO_RLOCK(&ripcbinfo);
833 	for (inp = LIST_FIRST(ripcbinfo.listhead), i = 0; inp && i < n;
834 	     inp = LIST_NEXT(inp, inp_list)) {
835 		INP_LOCK(inp);
836 		if (inp->inp_gencnt <= gencnt &&
837 		    cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) {
838 			/* XXX held references? */
839 			inp_list[i++] = inp;
840 		}
841 		INP_UNLOCK(inp);
842 	}
843 	INP_INFO_RUNLOCK(&ripcbinfo);
844 	n = i;
845 
846 	error = 0;
847 	for (i = 0; i < n; i++) {
848 		inp = inp_list[i];
849 		if (inp->inp_gencnt <= gencnt) {
850 			struct xinpcb xi;
851 			bzero(&xi, sizeof(xi));
852 			xi.xi_len = sizeof xi;
853 			/* XXX should avoid extra copy */
854 			bcopy(inp, &xi.xi_inp, sizeof *inp);
855 			if (inp->inp_socket)
856 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
857 			error = SYSCTL_OUT(req, &xi, sizeof xi);
858 		}
859 	}
860 	if (!error) {
861 		/*
862 		 * Give the user an updated idea of our state.
863 		 * If the generation differs from what we told
864 		 * her before, she knows that something happened
865 		 * while we were processing this request, and it
866 		 * might be necessary to retry.
867 		 */
868 		INP_INFO_RLOCK(&ripcbinfo);
869 		xig.xig_gen = ripcbinfo.ipi_gencnt;
870 		xig.xig_sogen = so_gencnt;
871 		xig.xig_count = ripcbinfo.ipi_count;
872 		INP_INFO_RUNLOCK(&ripcbinfo);
873 		error = SYSCTL_OUT(req, &xig, sizeof xig);
874 	}
875 	free(inp_list, M_TEMP);
876 	return error;
877 }
878 
879 /*
880  * This is the wrapper function for in_setsockaddr.  We just pass down
881  * the pcbinfo for in_setpeeraddr to lock.
882  */
883 static int
884 rip_sockaddr(struct socket *so, struct sockaddr **nam)
885 {
886 	return (in_setsockaddr(so, nam, &ripcbinfo));
887 }
888 
889 /*
890  * This is the wrapper function for in_setpeeraddr.  We just pass down
891  * the pcbinfo for in_setpeeraddr to lock.
892  */
893 static int
894 rip_peeraddr(struct socket *so, struct sockaddr **nam)
895 {
896 	return (in_setpeeraddr(so, nam, &ripcbinfo));
897 }
898 
899 
900 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0,
901 	    rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
902 
903 struct pr_usrreqs rip_usrreqs = {
904 	.pru_abort =		rip_abort,
905 	.pru_attach =		rip_attach,
906 	.pru_bind =		rip_bind,
907 	.pru_connect =		rip_connect,
908 	.pru_control =		in_control,
909 	.pru_detach =		rip_detach,
910 	.pru_disconnect =	rip_disconnect,
911 	.pru_peeraddr =		rip_peeraddr,
912 	.pru_send =		rip_send,
913 	.pru_shutdown =		rip_shutdown,
914 	.pru_sockaddr =		rip_sockaddr,
915 	.pru_sosetlabel =	in_pcbsosetlabel
916 };
917