xref: /freebsd/sys/netinet/raw_ip.c (revision b0fefb25c558179e9f9c7f0d375c6a03fb567eb9)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1993
5  *	The Regents of the University of California.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)raw_ip.c	8.7 (Berkeley) 5/15/95
33  */
34 
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37 
38 #include "opt_inet.h"
39 #include "opt_inet6.h"
40 #include "opt_ipsec.h"
41 
42 #include <sys/param.h>
43 #include <sys/jail.h>
44 #include <sys/kernel.h>
45 #include <sys/eventhandler.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mbuf.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/protosw.h>
52 #include <sys/rmlock.h>
53 #include <sys/rwlock.h>
54 #include <sys/signalvar.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sx.h>
58 #include <sys/sysctl.h>
59 #include <sys/systm.h>
60 
61 #include <vm/uma.h>
62 
63 #include <net/if.h>
64 #include <net/if_var.h>
65 #include <net/route.h>
66 #include <net/vnet.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_systm.h>
70 #include <netinet/in_pcb.h>
71 #include <netinet/in_var.h>
72 #include <netinet/if_ether.h>
73 #include <netinet/ip.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_mroute.h>
76 #include <netinet/ip_icmp.h>
77 
78 #include <netipsec/ipsec_support.h>
79 
80 #include <machine/stdarg.h>
81 #include <security/mac/mac_framework.h>
82 
83 VNET_DEFINE(int, ip_defttl) = IPDEFTTL;
84 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_VNET | CTLFLAG_RW,
85     &VNET_NAME(ip_defttl), 0,
86     "Maximum TTL on IP packets");
87 
88 VNET_DEFINE(struct inpcbhead, ripcb);
89 VNET_DEFINE(struct inpcbinfo, ripcbinfo);
90 
91 #define	V_ripcb			VNET(ripcb)
92 #define	V_ripcbinfo		VNET(ripcbinfo)
93 
94 /*
95  * Control and data hooks for ipfw, dummynet, divert and so on.
96  * The data hooks are not used here but it is convenient
97  * to keep them all in one place.
98  */
99 VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL;
100 VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL;
101 
102 int	(*ip_dn_ctl_ptr)(struct sockopt *);
103 int	(*ip_dn_io_ptr)(struct mbuf **, struct ip_fw_args *);
104 void	(*ip_divert_ptr)(struct mbuf *, bool);
105 int	(*ng_ipfw_input_p)(struct mbuf **, struct ip_fw_args *, bool);
106 
107 #ifdef INET
108 /*
109  * Hooks for multicast routing. They all default to NULL, so leave them not
110  * initialized and rely on BSS being set to 0.
111  */
112 
113 /*
114  * The socket used to communicate with the multicast routing daemon.
115  */
116 VNET_DEFINE(struct socket *, ip_mrouter);
117 
118 /*
119  * The various mrouter and rsvp functions.
120  */
121 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
122 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
123 int (*ip_mrouter_done)(void);
124 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
125 		   struct ip_moptions *);
126 int (*mrt_ioctl)(u_long, caddr_t, int);
127 int (*legal_vif_num)(int);
128 u_long (*ip_mcast_src)(int);
129 
130 int (*rsvp_input_p)(struct mbuf **, int *, int);
131 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
132 void (*ip_rsvp_force_done)(struct socket *);
133 #endif /* INET */
134 
135 extern	struct protosw inetsw[];
136 
137 u_long	rip_sendspace = 9216;
138 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
139     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
140 
141 u_long	rip_recvspace = 9216;
142 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
143     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
144 
145 /*
146  * Hash functions
147  */
148 
149 #define INP_PCBHASH_RAW_SIZE	256
150 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \
151         (((proto) + (laddr) + (faddr)) % (mask) + 1)
152 
153 #ifdef INET
154 static void
155 rip_inshash(struct inpcb *inp)
156 {
157 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
158 	struct inpcbhead *pcbhash;
159 	int hash;
160 
161 	INP_INFO_WLOCK_ASSERT(pcbinfo);
162 	INP_WLOCK_ASSERT(inp);
163 
164 	if (inp->inp_ip_p != 0 &&
165 	    inp->inp_laddr.s_addr != INADDR_ANY &&
166 	    inp->inp_faddr.s_addr != INADDR_ANY) {
167 		hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr,
168 		    inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask);
169 	} else
170 		hash = 0;
171 	pcbhash = &pcbinfo->ipi_hashbase[hash];
172 	CK_LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
173 }
174 
175 static void
176 rip_delhash(struct inpcb *inp)
177 {
178 
179 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
180 	INP_WLOCK_ASSERT(inp);
181 
182 	CK_LIST_REMOVE(inp, inp_hash);
183 }
184 #endif /* INET */
185 
186 /*
187  * Raw interface to IP protocol.
188  */
189 
190 /*
191  * Initialize raw connection block q.
192  */
193 static void
194 rip_zone_change(void *tag)
195 {
196 
197 	uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets);
198 }
199 
200 static int
201 rip_inpcb_init(void *mem, int size, int flags)
202 {
203 	struct inpcb *inp = mem;
204 
205 	INP_LOCK_INIT(inp, "inp", "rawinp");
206 	return (0);
207 }
208 
209 void
210 rip_init(void)
211 {
212 
213 	in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE,
214 	    1, "ripcb", rip_inpcb_init, IPI_HASHFIELDS_NONE);
215 	EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL,
216 	    EVENTHANDLER_PRI_ANY);
217 }
218 
219 #ifdef VIMAGE
220 static void
221 rip_destroy(void *unused __unused)
222 {
223 
224 	in_pcbinfo_destroy(&V_ripcbinfo);
225 }
226 VNET_SYSUNINIT(raw_ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, rip_destroy, NULL);
227 #endif
228 
229 #ifdef INET
230 static int
231 rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
232     struct sockaddr_in *ripsrc)
233 {
234 	int policyfail = 0;
235 
236 	INP_LOCK_ASSERT(last);
237 
238 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
239 	/* check AH/ESP integrity. */
240 	if (IPSEC_ENABLED(ipv4)) {
241 		if (IPSEC_CHECK_POLICY(ipv4, n, last) != 0)
242 			policyfail = 1;
243 	}
244 #endif /* IPSEC */
245 #ifdef MAC
246 	if (!policyfail && mac_inpcb_check_deliver(last, n) != 0)
247 		policyfail = 1;
248 #endif
249 	/* Check the minimum TTL for socket. */
250 	if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
251 		policyfail = 1;
252 	if (!policyfail) {
253 		struct mbuf *opts = NULL;
254 		struct socket *so;
255 
256 		so = last->inp_socket;
257 		if ((last->inp_flags & INP_CONTROLOPTS) ||
258 		    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
259 			ip_savecontrol(last, &opts, ip, n);
260 		SOCKBUF_LOCK(&so->so_rcv);
261 		if (sbappendaddr_locked(&so->so_rcv,
262 		    (struct sockaddr *)ripsrc, n, opts) == 0) {
263 			/* should notify about lost packet */
264 			m_freem(n);
265 			if (opts)
266 				m_freem(opts);
267 			SOCKBUF_UNLOCK(&so->so_rcv);
268 		} else
269 			sorwakeup_locked(so);
270 	} else
271 		m_freem(n);
272 	return (policyfail);
273 }
274 
275 /*
276  * Setup generic address and protocol structures for raw_input routine, then
277  * pass them along with mbuf chain.
278  */
279 int
280 rip_input(struct mbuf **mp, int *offp, int proto)
281 {
282 	struct ifnet *ifp;
283 	struct mbuf *m = *mp;
284 	struct ip *ip = mtod(m, struct ip *);
285 	struct inpcb *inp, *last;
286 	struct sockaddr_in ripsrc;
287 	struct epoch_tracker et;
288 	int hash;
289 
290 	*mp = NULL;
291 
292 	bzero(&ripsrc, sizeof(ripsrc));
293 	ripsrc.sin_len = sizeof(ripsrc);
294 	ripsrc.sin_family = AF_INET;
295 	ripsrc.sin_addr = ip->ip_src;
296 	last = NULL;
297 
298 	ifp = m->m_pkthdr.rcvif;
299 
300 	hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr,
301 	    ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
302 	INP_INFO_RLOCK_ET(&V_ripcbinfo, et);
303 	CK_LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) {
304 		if (inp->inp_ip_p != proto)
305 			continue;
306 #ifdef INET6
307 		/* XXX inp locking */
308 		if ((inp->inp_vflag & INP_IPV4) == 0)
309 			continue;
310 #endif
311 		if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
312 			continue;
313 		if (inp->inp_faddr.s_addr != ip->ip_src.s_addr)
314 			continue;
315 		if (last != NULL) {
316 			struct mbuf *n;
317 
318 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
319 			if (n != NULL)
320 			    (void) rip_append(last, ip, n, &ripsrc);
321 			/* XXX count dropped packet */
322 			INP_RUNLOCK(last);
323 			last = NULL;
324 		}
325 		INP_RLOCK(inp);
326 		if (__predict_false(inp->inp_flags2 & INP_FREED))
327 			goto skip_1;
328 		if (jailed_without_vnet(inp->inp_cred)) {
329 			/*
330 			 * XXX: If faddr was bound to multicast group,
331 			 * jailed raw socket will drop datagram.
332 			 */
333 			if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
334 				goto skip_1;
335 		}
336 		last = inp;
337 		continue;
338 	skip_1:
339 		INP_RUNLOCK(inp);
340 	}
341 	CK_LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) {
342 		if (inp->inp_ip_p && inp->inp_ip_p != proto)
343 			continue;
344 #ifdef INET6
345 		/* XXX inp locking */
346 		if ((inp->inp_vflag & INP_IPV4) == 0)
347 			continue;
348 #endif
349 		if (!in_nullhost(inp->inp_laddr) &&
350 		    !in_hosteq(inp->inp_laddr, ip->ip_dst))
351 			continue;
352 		if (!in_nullhost(inp->inp_faddr) &&
353 		    !in_hosteq(inp->inp_faddr, ip->ip_src))
354 			continue;
355 		if (last != NULL) {
356 			struct mbuf *n;
357 
358 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
359 			if (n != NULL)
360 				(void) rip_append(last, ip, n, &ripsrc);
361 			/* XXX count dropped packet */
362 			INP_RUNLOCK(last);
363 			last = NULL;
364 		}
365 		INP_RLOCK(inp);
366 		if (__predict_false(inp->inp_flags2 & INP_FREED))
367 			goto skip_2;
368 		if (jailed_without_vnet(inp->inp_cred)) {
369 			/*
370 			 * Allow raw socket in jail to receive multicast;
371 			 * assume process had PRIV_NETINET_RAW at attach,
372 			 * and fall through into normal filter path if so.
373 			 */
374 			if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
375 			    prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
376 				goto skip_2;
377 		}
378 		/*
379 		 * If this raw socket has multicast state, and we
380 		 * have received a multicast, check if this socket
381 		 * should receive it, as multicast filtering is now
382 		 * the responsibility of the transport layer.
383 		 */
384 		if (inp->inp_moptions != NULL &&
385 		    IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
386 			/*
387 			 * If the incoming datagram is for IGMP, allow it
388 			 * through unconditionally to the raw socket.
389 			 *
390 			 * In the case of IGMPv2, we may not have explicitly
391 			 * joined the group, and may have set IFF_ALLMULTI
392 			 * on the interface. imo_multi_filter() may discard
393 			 * control traffic we actually need to see.
394 			 *
395 			 * Userland multicast routing daemons should continue
396 			 * filter the control traffic appropriately.
397 			 */
398 			int blocked;
399 
400 			blocked = MCAST_PASS;
401 			if (proto != IPPROTO_IGMP) {
402 				struct sockaddr_in group;
403 
404 				bzero(&group, sizeof(struct sockaddr_in));
405 				group.sin_len = sizeof(struct sockaddr_in);
406 				group.sin_family = AF_INET;
407 				group.sin_addr = ip->ip_dst;
408 
409 				blocked = imo_multi_filter(inp->inp_moptions,
410 				    ifp,
411 				    (struct sockaddr *)&group,
412 				    (struct sockaddr *)&ripsrc);
413 			}
414 
415 			if (blocked != MCAST_PASS) {
416 				IPSTAT_INC(ips_notmember);
417 				goto skip_2;
418 			}
419 		}
420 		last = inp;
421 		continue;
422 	skip_2:
423 		INP_RUNLOCK(inp);
424 	}
425 	INP_INFO_RUNLOCK_ET(&V_ripcbinfo, et);
426 	if (last != NULL) {
427 		if (rip_append(last, ip, m, &ripsrc) != 0)
428 			IPSTAT_INC(ips_delivered);
429 		INP_RUNLOCK(last);
430 	} else {
431 		if (inetsw[ip_protox[ip->ip_p]].pr_input == rip_input) {
432 			IPSTAT_INC(ips_noproto);
433 			IPSTAT_DEC(ips_delivered);
434 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PROTOCOL, 0, 0);
435 		} else {
436 			m_freem(m);
437 		}
438 	}
439 	return (IPPROTO_DONE);
440 }
441 
442 /*
443  * Generate IP header and pass packet to ip_output.  Tack on options user may
444  * have setup with control call.
445  */
446 int
447 rip_output(struct mbuf *m, struct socket *so, ...)
448 {
449 	struct ip *ip;
450 	int error;
451 	struct inpcb *inp = sotoinpcb(so);
452 	va_list ap;
453 	u_long dst;
454 	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
455 	    IP_ALLOWBROADCAST;
456 	int cnt;
457 	u_char opttype, optlen, *cp;
458 
459 	va_start(ap, so);
460 	dst = va_arg(ap, u_long);
461 	va_end(ap);
462 
463 	/*
464 	 * If the user handed us a complete IP packet, use it.  Otherwise,
465 	 * allocate an mbuf for a header and fill it in.
466 	 */
467 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
468 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
469 			m_freem(m);
470 			return(EMSGSIZE);
471 		}
472 		M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
473 		if (m == NULL)
474 			return(ENOBUFS);
475 
476 		INP_RLOCK(inp);
477 		ip = mtod(m, struct ip *);
478 		ip->ip_tos = inp->inp_ip_tos;
479 		if (inp->inp_flags & INP_DONTFRAG)
480 			ip->ip_off = htons(IP_DF);
481 		else
482 			ip->ip_off = htons(0);
483 		ip->ip_p = inp->inp_ip_p;
484 		ip->ip_len = htons(m->m_pkthdr.len);
485 		ip->ip_src = inp->inp_laddr;
486 		ip->ip_dst.s_addr = dst;
487 		if (jailed(inp->inp_cred)) {
488 			/*
489 			 * prison_local_ip4() would be good enough but would
490 			 * let a source of INADDR_ANY pass, which we do not
491 			 * want to see from jails.
492 			 */
493 			if (ip->ip_src.s_addr == INADDR_ANY) {
494 				error = in_pcbladdr(inp, &ip->ip_dst, &ip->ip_src,
495 				    inp->inp_cred);
496 			} else {
497 				error = prison_local_ip4(inp->inp_cred,
498 				    &ip->ip_src);
499 			}
500 			if (error != 0) {
501 				INP_RUNLOCK(inp);
502 				m_freem(m);
503 				return (error);
504 			}
505 		}
506 		ip->ip_ttl = inp->inp_ip_ttl;
507 	} else {
508 		if (m->m_pkthdr.len > IP_MAXPACKET) {
509 			m_freem(m);
510 			return(EMSGSIZE);
511 		}
512 		INP_RLOCK(inp);
513 		ip = mtod(m, struct ip *);
514 		error = prison_check_ip4(inp->inp_cred, &ip->ip_src);
515 		if (error != 0) {
516 			INP_RUNLOCK(inp);
517 			m_freem(m);
518 			return (error);
519 		}
520 
521 		/*
522 		 * Don't allow both user specified and setsockopt options,
523 		 * and don't allow packet length sizes that will crash.
524 		 */
525 		if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options)
526 		    || (ntohs(ip->ip_len) != m->m_pkthdr.len)
527 		    || (ntohs(ip->ip_len) < (ip->ip_hl << 2))) {
528 			INP_RUNLOCK(inp);
529 			m_freem(m);
530 			return (EINVAL);
531 		}
532 		/*
533 		 * Don't allow IP options which do not have the required
534 		 * structure as specified in section 3.1 of RFC 791 on
535 		 * pages 15-23.
536 		 */
537 		cp = (u_char *)(ip + 1);
538 		cnt = (ip->ip_hl << 2) - sizeof (struct ip);
539 		for (; cnt > 0; cnt -= optlen, cp += optlen) {
540 			opttype = cp[IPOPT_OPTVAL];
541 			if (opttype == IPOPT_EOL)
542 				break;
543 			if (opttype == IPOPT_NOP) {
544 				optlen = 1;
545 				continue;
546 			}
547 			if (cnt < IPOPT_OLEN + sizeof(u_char)) {
548 				INP_RUNLOCK(inp);
549 				m_freem(m);
550 				return (EINVAL);
551 			}
552 			optlen = cp[IPOPT_OLEN];
553 			if (optlen < IPOPT_OLEN + sizeof(u_char) ||
554 			    optlen > cnt) {
555 				INP_RUNLOCK(inp);
556 				m_freem(m);
557 				return (EINVAL);
558 			}
559 		}
560 		/*
561 		 * This doesn't allow application to specify ID of zero,
562 		 * but we got this limitation from the beginning of history.
563 		 */
564 		if (ip->ip_id == 0)
565 			ip_fillid(ip);
566 
567 		/*
568 		 * XXX prevent ip_output from overwriting header fields.
569 		 */
570 		flags |= IP_RAWOUTPUT;
571 		IPSTAT_INC(ips_rawout);
572 	}
573 
574 	if (inp->inp_flags & INP_ONESBCAST)
575 		flags |= IP_SENDONES;
576 
577 #ifdef MAC
578 	mac_inpcb_create_mbuf(inp, m);
579 #endif
580 
581 	error = ip_output(m, inp->inp_options, NULL, flags,
582 	    inp->inp_moptions, inp);
583 	INP_RUNLOCK(inp);
584 	return (error);
585 }
586 
587 /*
588  * Raw IP socket option processing.
589  *
590  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
591  * only be created by a privileged process, and as such, socket option
592  * operations to manage system properties on any raw socket were allowed to
593  * take place without explicit additional access control checks.  However,
594  * raw sockets can now also be created in jail(), and therefore explicit
595  * checks are now required.  Likewise, raw sockets can be used by a process
596  * after it gives up privilege, so some caution is required.  For options
597  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
598  * performed in ip_ctloutput() and therefore no check occurs here.
599  * Unilaterally checking priv_check() here breaks normal IP socket option
600  * operations on raw sockets.
601  *
602  * When adding new socket options here, make sure to add access control
603  * checks here as necessary.
604  *
605  * XXX-BZ inp locking?
606  */
607 int
608 rip_ctloutput(struct socket *so, struct sockopt *sopt)
609 {
610 	struct	inpcb *inp = sotoinpcb(so);
611 	int	error, optval;
612 
613 	if (sopt->sopt_level != IPPROTO_IP) {
614 		if ((sopt->sopt_level == SOL_SOCKET) &&
615 		    (sopt->sopt_name == SO_SETFIB)) {
616 			inp->inp_inc.inc_fibnum = so->so_fibnum;
617 			return (0);
618 		}
619 		return (EINVAL);
620 	}
621 
622 	error = 0;
623 	switch (sopt->sopt_dir) {
624 	case SOPT_GET:
625 		switch (sopt->sopt_name) {
626 		case IP_HDRINCL:
627 			optval = inp->inp_flags & INP_HDRINCL;
628 			error = sooptcopyout(sopt, &optval, sizeof optval);
629 			break;
630 
631 		case IP_FW3:	/* generic ipfw v.3 functions */
632 		case IP_FW_ADD:	/* ADD actually returns the body... */
633 		case IP_FW_GET:
634 		case IP_FW_TABLE_GETSIZE:
635 		case IP_FW_TABLE_LIST:
636 		case IP_FW_NAT_GET_CONFIG:
637 		case IP_FW_NAT_GET_LOG:
638 			if (V_ip_fw_ctl_ptr != NULL)
639 				error = V_ip_fw_ctl_ptr(sopt);
640 			else
641 				error = ENOPROTOOPT;
642 			break;
643 
644 		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
645 		case IP_DUMMYNET_GET:
646 			if (ip_dn_ctl_ptr != NULL)
647 				error = ip_dn_ctl_ptr(sopt);
648 			else
649 				error = ENOPROTOOPT;
650 			break ;
651 
652 		case MRT_INIT:
653 		case MRT_DONE:
654 		case MRT_ADD_VIF:
655 		case MRT_DEL_VIF:
656 		case MRT_ADD_MFC:
657 		case MRT_DEL_MFC:
658 		case MRT_VERSION:
659 		case MRT_ASSERT:
660 		case MRT_API_SUPPORT:
661 		case MRT_API_CONFIG:
662 		case MRT_ADD_BW_UPCALL:
663 		case MRT_DEL_BW_UPCALL:
664 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
665 			if (error != 0)
666 				return (error);
667 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
668 				EOPNOTSUPP;
669 			break;
670 
671 		default:
672 			error = ip_ctloutput(so, sopt);
673 			break;
674 		}
675 		break;
676 
677 	case SOPT_SET:
678 		switch (sopt->sopt_name) {
679 		case IP_HDRINCL:
680 			error = sooptcopyin(sopt, &optval, sizeof optval,
681 					    sizeof optval);
682 			if (error)
683 				break;
684 			if (optval)
685 				inp->inp_flags |= INP_HDRINCL;
686 			else
687 				inp->inp_flags &= ~INP_HDRINCL;
688 			break;
689 
690 		case IP_FW3:	/* generic ipfw v.3 functions */
691 		case IP_FW_ADD:
692 		case IP_FW_DEL:
693 		case IP_FW_FLUSH:
694 		case IP_FW_ZERO:
695 		case IP_FW_RESETLOG:
696 		case IP_FW_TABLE_ADD:
697 		case IP_FW_TABLE_DEL:
698 		case IP_FW_TABLE_FLUSH:
699 		case IP_FW_NAT_CFG:
700 		case IP_FW_NAT_DEL:
701 			if (V_ip_fw_ctl_ptr != NULL)
702 				error = V_ip_fw_ctl_ptr(sopt);
703 			else
704 				error = ENOPROTOOPT;
705 			break;
706 
707 		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
708 		case IP_DUMMYNET_CONFIGURE:
709 		case IP_DUMMYNET_DEL:
710 		case IP_DUMMYNET_FLUSH:
711 			if (ip_dn_ctl_ptr != NULL)
712 				error = ip_dn_ctl_ptr(sopt);
713 			else
714 				error = ENOPROTOOPT ;
715 			break ;
716 
717 		case IP_RSVP_ON:
718 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
719 			if (error != 0)
720 				return (error);
721 			error = ip_rsvp_init(so);
722 			break;
723 
724 		case IP_RSVP_OFF:
725 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
726 			if (error != 0)
727 				return (error);
728 			error = ip_rsvp_done();
729 			break;
730 
731 		case IP_RSVP_VIF_ON:
732 		case IP_RSVP_VIF_OFF:
733 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
734 			if (error != 0)
735 				return (error);
736 			error = ip_rsvp_vif ?
737 				ip_rsvp_vif(so, sopt) : EINVAL;
738 			break;
739 
740 		case MRT_INIT:
741 		case MRT_DONE:
742 		case MRT_ADD_VIF:
743 		case MRT_DEL_VIF:
744 		case MRT_ADD_MFC:
745 		case MRT_DEL_MFC:
746 		case MRT_VERSION:
747 		case MRT_ASSERT:
748 		case MRT_API_SUPPORT:
749 		case MRT_API_CONFIG:
750 		case MRT_ADD_BW_UPCALL:
751 		case MRT_DEL_BW_UPCALL:
752 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
753 			if (error != 0)
754 				return (error);
755 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
756 					EOPNOTSUPP;
757 			break;
758 
759 		default:
760 			error = ip_ctloutput(so, sopt);
761 			break;
762 		}
763 		break;
764 	}
765 
766 	return (error);
767 }
768 
769 /*
770  * This function exists solely to receive the PRC_IFDOWN messages which are
771  * sent by if_down().  It looks for an ifaddr whose ifa_addr is sa, and calls
772  * in_ifadown() to remove all routes corresponding to that address.  It also
773  * receives the PRC_IFUP messages from if_up() and reinstalls the interface
774  * routes.
775  */
776 void
777 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
778 {
779 	struct rm_priotracker in_ifa_tracker;
780 	struct in_ifaddr *ia;
781 	struct ifnet *ifp;
782 	int err;
783 	int flags;
784 
785 	switch (cmd) {
786 	case PRC_IFDOWN:
787 		IN_IFADDR_RLOCK(&in_ifa_tracker);
788 		CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
789 			if (ia->ia_ifa.ifa_addr == sa
790 			    && (ia->ia_flags & IFA_ROUTE)) {
791 				ifa_ref(&ia->ia_ifa);
792 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
793 				/*
794 				 * in_scrubprefix() kills the interface route.
795 				 */
796 				in_scrubprefix(ia, 0);
797 				/*
798 				 * in_ifadown gets rid of all the rest of the
799 				 * routes.  This is not quite the right thing
800 				 * to do, but at least if we are running a
801 				 * routing process they will come back.
802 				 */
803 				in_ifadown(&ia->ia_ifa, 0);
804 				ifa_free(&ia->ia_ifa);
805 				break;
806 			}
807 		}
808 		if (ia == NULL)		/* If ia matched, already unlocked. */
809 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
810 		break;
811 
812 	case PRC_IFUP:
813 		IN_IFADDR_RLOCK(&in_ifa_tracker);
814 		CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
815 			if (ia->ia_ifa.ifa_addr == sa)
816 				break;
817 		}
818 		if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) {
819 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
820 			return;
821 		}
822 		ifa_ref(&ia->ia_ifa);
823 		IN_IFADDR_RUNLOCK(&in_ifa_tracker);
824 		flags = RTF_UP;
825 		ifp = ia->ia_ifa.ifa_ifp;
826 
827 		if ((ifp->if_flags & IFF_LOOPBACK)
828 		    || (ifp->if_flags & IFF_POINTOPOINT))
829 			flags |= RTF_HOST;
830 
831 		err = ifa_del_loopback_route((struct ifaddr *)ia, sa);
832 
833 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
834 		if (err == 0)
835 			ia->ia_flags |= IFA_ROUTE;
836 
837 		err = ifa_add_loopback_route((struct ifaddr *)ia, sa);
838 
839 		ifa_free(&ia->ia_ifa);
840 		break;
841 	}
842 }
843 
844 static int
845 rip_attach(struct socket *so, int proto, struct thread *td)
846 {
847 	struct inpcb *inp;
848 	int error;
849 
850 	inp = sotoinpcb(so);
851 	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
852 
853 	error = priv_check(td, PRIV_NETINET_RAW);
854 	if (error)
855 		return (error);
856 	if (proto >= IPPROTO_MAX || proto < 0)
857 		return EPROTONOSUPPORT;
858 	error = soreserve(so, rip_sendspace, rip_recvspace);
859 	if (error)
860 		return (error);
861 	INP_INFO_WLOCK(&V_ripcbinfo);
862 	error = in_pcballoc(so, &V_ripcbinfo);
863 	if (error) {
864 		INP_INFO_WUNLOCK(&V_ripcbinfo);
865 		return (error);
866 	}
867 	inp = (struct inpcb *)so->so_pcb;
868 	inp->inp_vflag |= INP_IPV4;
869 	inp->inp_ip_p = proto;
870 	inp->inp_ip_ttl = V_ip_defttl;
871 	rip_inshash(inp);
872 	INP_INFO_WUNLOCK(&V_ripcbinfo);
873 	INP_WUNLOCK(inp);
874 	return (0);
875 }
876 
877 static void
878 rip_detach(struct socket *so)
879 {
880 	struct inpcb *inp;
881 
882 	inp = sotoinpcb(so);
883 	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
884 	KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
885 	    ("rip_detach: not closed"));
886 
887 	INP_INFO_WLOCK(&V_ripcbinfo);
888 	INP_WLOCK(inp);
889 	rip_delhash(inp);
890 	if (so == V_ip_mrouter && ip_mrouter_done)
891 		ip_mrouter_done();
892 	if (ip_rsvp_force_done)
893 		ip_rsvp_force_done(so);
894 	if (so == V_ip_rsvpd)
895 		ip_rsvp_done();
896 	in_pcbdetach(inp);
897 	in_pcbfree(inp);
898 	INP_INFO_WUNLOCK(&V_ripcbinfo);
899 }
900 
901 static void
902 rip_dodisconnect(struct socket *so, struct inpcb *inp)
903 {
904 	struct inpcbinfo *pcbinfo;
905 
906 	pcbinfo = inp->inp_pcbinfo;
907 	INP_INFO_WLOCK(pcbinfo);
908 	INP_WLOCK(inp);
909 	rip_delhash(inp);
910 	inp->inp_faddr.s_addr = INADDR_ANY;
911 	rip_inshash(inp);
912 	SOCK_LOCK(so);
913 	so->so_state &= ~SS_ISCONNECTED;
914 	SOCK_UNLOCK(so);
915 	INP_WUNLOCK(inp);
916 	INP_INFO_WUNLOCK(pcbinfo);
917 }
918 
919 static void
920 rip_abort(struct socket *so)
921 {
922 	struct inpcb *inp;
923 
924 	inp = sotoinpcb(so);
925 	KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
926 
927 	rip_dodisconnect(so, inp);
928 }
929 
930 static void
931 rip_close(struct socket *so)
932 {
933 	struct inpcb *inp;
934 
935 	inp = sotoinpcb(so);
936 	KASSERT(inp != NULL, ("rip_close: inp == NULL"));
937 
938 	rip_dodisconnect(so, inp);
939 }
940 
941 static int
942 rip_disconnect(struct socket *so)
943 {
944 	struct inpcb *inp;
945 
946 	if ((so->so_state & SS_ISCONNECTED) == 0)
947 		return (ENOTCONN);
948 
949 	inp = sotoinpcb(so);
950 	KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
951 
952 	rip_dodisconnect(so, inp);
953 	return (0);
954 }
955 
956 static int
957 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
958 {
959 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
960 	struct inpcb *inp;
961 	int error;
962 
963 	if (nam->sa_len != sizeof(*addr))
964 		return (EINVAL);
965 
966 	error = prison_check_ip4(td->td_ucred, &addr->sin_addr);
967 	if (error != 0)
968 		return (error);
969 
970 	inp = sotoinpcb(so);
971 	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
972 
973 	if (CK_STAILQ_EMPTY(&V_ifnet) ||
974 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
975 	    (addr->sin_addr.s_addr &&
976 	     (inp->inp_flags & INP_BINDANY) == 0 &&
977 	     ifa_ifwithaddr_check((struct sockaddr *)addr) == 0))
978 		return (EADDRNOTAVAIL);
979 
980 	INP_INFO_WLOCK(&V_ripcbinfo);
981 	INP_WLOCK(inp);
982 	rip_delhash(inp);
983 	inp->inp_laddr = addr->sin_addr;
984 	rip_inshash(inp);
985 	INP_WUNLOCK(inp);
986 	INP_INFO_WUNLOCK(&V_ripcbinfo);
987 	return (0);
988 }
989 
990 static int
991 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
992 {
993 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
994 	struct inpcb *inp;
995 
996 	if (nam->sa_len != sizeof(*addr))
997 		return (EINVAL);
998 	if (CK_STAILQ_EMPTY(&V_ifnet))
999 		return (EADDRNOTAVAIL);
1000 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
1001 		return (EAFNOSUPPORT);
1002 
1003 	inp = sotoinpcb(so);
1004 	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
1005 
1006 	INP_INFO_WLOCK(&V_ripcbinfo);
1007 	INP_WLOCK(inp);
1008 	rip_delhash(inp);
1009 	inp->inp_faddr = addr->sin_addr;
1010 	rip_inshash(inp);
1011 	soisconnected(so);
1012 	INP_WUNLOCK(inp);
1013 	INP_INFO_WUNLOCK(&V_ripcbinfo);
1014 	return (0);
1015 }
1016 
1017 static int
1018 rip_shutdown(struct socket *so)
1019 {
1020 	struct inpcb *inp;
1021 
1022 	inp = sotoinpcb(so);
1023 	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
1024 
1025 	INP_WLOCK(inp);
1026 	socantsendmore(so);
1027 	INP_WUNLOCK(inp);
1028 	return (0);
1029 }
1030 
1031 static int
1032 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
1033     struct mbuf *control, struct thread *td)
1034 {
1035 	struct inpcb *inp;
1036 	u_long dst;
1037 
1038 	inp = sotoinpcb(so);
1039 	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
1040 
1041 	/*
1042 	 * Note: 'dst' reads below are unlocked.
1043 	 */
1044 	if (so->so_state & SS_ISCONNECTED) {
1045 		if (nam) {
1046 			m_freem(m);
1047 			return (EISCONN);
1048 		}
1049 		dst = inp->inp_faddr.s_addr;	/* Unlocked read. */
1050 	} else {
1051 		if (nam == NULL) {
1052 			m_freem(m);
1053 			return (ENOTCONN);
1054 		}
1055 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
1056 	}
1057 	return (rip_output(m, so, dst));
1058 }
1059 #endif /* INET */
1060 
1061 static int
1062 rip_pcblist(SYSCTL_HANDLER_ARGS)
1063 {
1064 	int error, i, n;
1065 	struct inpcb *inp, **inp_list;
1066 	inp_gen_t gencnt;
1067 	struct xinpgen xig;
1068 	struct epoch_tracker et;
1069 
1070 	/*
1071 	 * The process of preparing the TCB list is too time-consuming and
1072 	 * resource-intensive to repeat twice on every request.
1073 	 */
1074 	if (req->oldptr == 0) {
1075 		n = V_ripcbinfo.ipi_count;
1076 		n += imax(n / 8, 10);
1077 		req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
1078 		return (0);
1079 	}
1080 
1081 	if (req->newptr != 0)
1082 		return (EPERM);
1083 
1084 	/*
1085 	 * OK, now we're committed to doing something.
1086 	 */
1087 	INP_INFO_WLOCK(&V_ripcbinfo);
1088 	gencnt = V_ripcbinfo.ipi_gencnt;
1089 	n = V_ripcbinfo.ipi_count;
1090 	INP_INFO_WUNLOCK(&V_ripcbinfo);
1091 
1092 	bzero(&xig, sizeof(xig));
1093 	xig.xig_len = sizeof xig;
1094 	xig.xig_count = n;
1095 	xig.xig_gen = gencnt;
1096 	xig.xig_sogen = so_gencnt;
1097 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1098 	if (error)
1099 		return (error);
1100 
1101 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
1102 
1103 	INP_INFO_RLOCK_ET(&V_ripcbinfo, et);
1104 	for (inp = CK_LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n;
1105 	     inp = CK_LIST_NEXT(inp, inp_list)) {
1106 		INP_WLOCK(inp);
1107 		if (inp->inp_gencnt <= gencnt &&
1108 		    cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
1109 			in_pcbref(inp);
1110 			inp_list[i++] = inp;
1111 		}
1112 		INP_WUNLOCK(inp);
1113 	}
1114 	INP_INFO_RUNLOCK_ET(&V_ripcbinfo, et);
1115 	n = i;
1116 
1117 	error = 0;
1118 	for (i = 0; i < n; i++) {
1119 		inp = inp_list[i];
1120 		INP_RLOCK(inp);
1121 		if (inp->inp_gencnt <= gencnt) {
1122 			struct xinpcb xi;
1123 
1124 			in_pcbtoxinpcb(inp, &xi);
1125 			INP_RUNLOCK(inp);
1126 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1127 		} else
1128 			INP_RUNLOCK(inp);
1129 	}
1130 	INP_INFO_WLOCK(&V_ripcbinfo);
1131 	for (i = 0; i < n; i++) {
1132 		inp = inp_list[i];
1133 		INP_RLOCK(inp);
1134 		if (!in_pcbrele_rlocked(inp))
1135 			INP_RUNLOCK(inp);
1136 	}
1137 	INP_INFO_WUNLOCK(&V_ripcbinfo);
1138 
1139 	if (!error) {
1140 		struct epoch_tracker et;
1141 		/*
1142 		 * Give the user an updated idea of our state.  If the
1143 		 * generation differs from what we told her before, she knows
1144 		 * that something happened while we were processing this
1145 		 * request, and it might be necessary to retry.
1146 		 */
1147 		INP_INFO_RLOCK_ET(&V_ripcbinfo, et);
1148 		xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1149 		xig.xig_sogen = so_gencnt;
1150 		xig.xig_count = V_ripcbinfo.ipi_count;
1151 		INP_INFO_RUNLOCK_ET(&V_ripcbinfo, et);
1152 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1153 	}
1154 	free(inp_list, M_TEMP);
1155 	return (error);
1156 }
1157 
1158 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist,
1159     CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0,
1160     rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
1161 
1162 #ifdef INET
1163 struct pr_usrreqs rip_usrreqs = {
1164 	.pru_abort =		rip_abort,
1165 	.pru_attach =		rip_attach,
1166 	.pru_bind =		rip_bind,
1167 	.pru_connect =		rip_connect,
1168 	.pru_control =		in_control,
1169 	.pru_detach =		rip_detach,
1170 	.pru_disconnect =	rip_disconnect,
1171 	.pru_peeraddr =		in_getpeeraddr,
1172 	.pru_send =		rip_send,
1173 	.pru_shutdown =		rip_shutdown,
1174 	.pru_sockaddr =		in_getsockaddr,
1175 	.pru_sosetlabel =	in_pcbsosetlabel,
1176 	.pru_close =		rip_close,
1177 };
1178 #endif /* INET */
1179