xref: /freebsd/sys/netinet/raw_ip.c (revision c66ec88fed842fbaad62c30d510644ceb7bd2d71)
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 #include "opt_route.h"
42 
43 #include <sys/param.h>
44 #include <sys/jail.h>
45 #include <sys/kernel.h>
46 #include <sys/eventhandler.h>
47 #include <sys/lock.h>
48 #include <sys/malloc.h>
49 #include <sys/mbuf.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/protosw.h>
53 #include <sys/rmlock.h>
54 #include <sys/rwlock.h>
55 #include <sys/signalvar.h>
56 #include <sys/socket.h>
57 #include <sys/socketvar.h>
58 #include <sys/sx.h>
59 #include <sys/sysctl.h>
60 #include <sys/systm.h>
61 
62 #include <vm/uma.h>
63 
64 #include <net/if.h>
65 #include <net/if_var.h>
66 #include <net/route.h>
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/in_fib.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/in_var.h>
74 #include <netinet/if_ether.h>
75 #include <netinet/ip.h>
76 #include <netinet/ip_var.h>
77 #include <netinet/ip_mroute.h>
78 #include <netinet/ip_icmp.h>
79 
80 #include <netipsec/ipsec_support.h>
81 
82 #include <machine/stdarg.h>
83 #include <security/mac/mac_framework.h>
84 
85 VNET_DEFINE(int, ip_defttl) = IPDEFTTL;
86 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_VNET | CTLFLAG_RW,
87     &VNET_NAME(ip_defttl), 0,
88     "Maximum TTL on IP packets");
89 
90 VNET_DEFINE(struct inpcbhead, ripcb);
91 VNET_DEFINE(struct inpcbinfo, ripcbinfo);
92 
93 #define	V_ripcb			VNET(ripcb)
94 #define	V_ripcbinfo		VNET(ripcbinfo)
95 
96 /*
97  * Control and data hooks for ipfw, dummynet, divert and so on.
98  * The data hooks are not used here but it is convenient
99  * to keep them all in one place.
100  */
101 VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL;
102 VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL;
103 
104 int	(*ip_dn_ctl_ptr)(struct sockopt *);
105 int	(*ip_dn_io_ptr)(struct mbuf **, struct ip_fw_args *);
106 void	(*ip_divert_ptr)(struct mbuf *, bool);
107 int	(*ng_ipfw_input_p)(struct mbuf **, struct ip_fw_args *, bool);
108 
109 #ifdef INET
110 /*
111  * Hooks for multicast routing. They all default to NULL, so leave them not
112  * initialized and rely on BSS being set to 0.
113  */
114 
115 /*
116  * The socket used to communicate with the multicast routing daemon.
117  */
118 VNET_DEFINE(struct socket *, ip_mrouter);
119 
120 /*
121  * The various mrouter and rsvp functions.
122  */
123 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
124 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
125 int (*ip_mrouter_done)(void);
126 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
127 		   struct ip_moptions *);
128 int (*mrt_ioctl)(u_long, caddr_t, int);
129 int (*legal_vif_num)(int);
130 u_long (*ip_mcast_src)(int);
131 
132 int (*rsvp_input_p)(struct mbuf **, int *, int);
133 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
134 void (*ip_rsvp_force_done)(struct socket *);
135 #endif /* INET */
136 
137 extern	struct protosw inetsw[];
138 
139 u_long	rip_sendspace = 9216;
140 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
141     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
142 
143 u_long	rip_recvspace = 9216;
144 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
145     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
146 
147 /*
148  * Hash functions
149  */
150 
151 #define INP_PCBHASH_RAW_SIZE	256
152 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \
153         (((proto) + (laddr) + (faddr)) % (mask) + 1)
154 
155 #ifdef INET
156 static void
157 rip_inshash(struct inpcb *inp)
158 {
159 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
160 	struct inpcbhead *pcbhash;
161 	int hash;
162 
163 	INP_INFO_WLOCK_ASSERT(pcbinfo);
164 	INP_WLOCK_ASSERT(inp);
165 
166 	if (inp->inp_ip_p != 0 &&
167 	    inp->inp_laddr.s_addr != INADDR_ANY &&
168 	    inp->inp_faddr.s_addr != INADDR_ANY) {
169 		hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr,
170 		    inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask);
171 	} else
172 		hash = 0;
173 	pcbhash = &pcbinfo->ipi_hashbase[hash];
174 	CK_LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
175 }
176 
177 static void
178 rip_delhash(struct inpcb *inp)
179 {
180 
181 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
182 	INP_WLOCK_ASSERT(inp);
183 
184 	CK_LIST_REMOVE(inp, inp_hash);
185 }
186 #endif /* INET */
187 
188 /*
189  * Raw interface to IP protocol.
190  */
191 
192 /*
193  * Initialize raw connection block q.
194  */
195 static void
196 rip_zone_change(void *tag)
197 {
198 
199 	uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets);
200 }
201 
202 static int
203 rip_inpcb_init(void *mem, int size, int flags)
204 {
205 	struct inpcb *inp = mem;
206 
207 	INP_LOCK_INIT(inp, "inp", "rawinp");
208 	return (0);
209 }
210 
211 void
212 rip_init(void)
213 {
214 
215 	in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE,
216 	    1, "ripcb", rip_inpcb_init, IPI_HASHFIELDS_NONE);
217 	EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL,
218 	    EVENTHANDLER_PRI_ANY);
219 }
220 
221 #ifdef VIMAGE
222 static void
223 rip_destroy(void *unused __unused)
224 {
225 
226 	in_pcbinfo_destroy(&V_ripcbinfo);
227 }
228 VNET_SYSUNINIT(raw_ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, rip_destroy, NULL);
229 #endif
230 
231 #ifdef INET
232 static int
233 rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
234     struct sockaddr_in *ripsrc)
235 {
236 	int policyfail = 0;
237 
238 	INP_LOCK_ASSERT(last);
239 
240 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
241 	/* check AH/ESP integrity. */
242 	if (IPSEC_ENABLED(ipv4)) {
243 		if (IPSEC_CHECK_POLICY(ipv4, n, last) != 0)
244 			policyfail = 1;
245 	}
246 #endif /* IPSEC */
247 #ifdef MAC
248 	if (!policyfail && mac_inpcb_check_deliver(last, n) != 0)
249 		policyfail = 1;
250 #endif
251 	/* Check the minimum TTL for socket. */
252 	if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
253 		policyfail = 1;
254 	if (!policyfail) {
255 		struct mbuf *opts = NULL;
256 		struct socket *so;
257 
258 		so = last->inp_socket;
259 		if ((last->inp_flags & INP_CONTROLOPTS) ||
260 		    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
261 			ip_savecontrol(last, &opts, ip, n);
262 		SOCKBUF_LOCK(&so->so_rcv);
263 		if (sbappendaddr_locked(&so->so_rcv,
264 		    (struct sockaddr *)ripsrc, n, opts) == 0) {
265 			/* should notify about lost packet */
266 			m_freem(n);
267 			if (opts)
268 				m_freem(opts);
269 			SOCKBUF_UNLOCK(&so->so_rcv);
270 		} else
271 			sorwakeup_locked(so);
272 	} else
273 		m_freem(n);
274 	return (policyfail);
275 }
276 
277 /*
278  * Setup generic address and protocol structures for raw_input routine, then
279  * pass them along with mbuf chain.
280  */
281 int
282 rip_input(struct mbuf **mp, int *offp, int proto)
283 {
284 	struct ifnet *ifp;
285 	struct mbuf *m = *mp;
286 	struct ip *ip = mtod(m, struct ip *);
287 	struct inpcb *inp, *last;
288 	struct sockaddr_in ripsrc;
289 	int hash;
290 
291 	NET_EPOCH_ASSERT();
292 
293 	*mp = NULL;
294 
295 	bzero(&ripsrc, sizeof(ripsrc));
296 	ripsrc.sin_len = sizeof(ripsrc);
297 	ripsrc.sin_family = AF_INET;
298 	ripsrc.sin_addr = ip->ip_src;
299 	last = NULL;
300 
301 	ifp = m->m_pkthdr.rcvif;
302 
303 	hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr,
304 	    ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
305 	CK_LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) {
306 		if (inp->inp_ip_p != proto)
307 			continue;
308 #ifdef INET6
309 		/* XXX inp locking */
310 		if ((inp->inp_vflag & INP_IPV4) == 0)
311 			continue;
312 #endif
313 		if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
314 			continue;
315 		if (inp->inp_faddr.s_addr != ip->ip_src.s_addr)
316 			continue;
317 		if (last != NULL) {
318 			struct mbuf *n;
319 
320 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
321 			if (n != NULL)
322 			    (void) rip_append(last, ip, n, &ripsrc);
323 			/* XXX count dropped packet */
324 			INP_RUNLOCK(last);
325 			last = NULL;
326 		}
327 		INP_RLOCK(inp);
328 		if (__predict_false(inp->inp_flags2 & INP_FREED))
329 			goto skip_1;
330 		if (jailed_without_vnet(inp->inp_cred)) {
331 			/*
332 			 * XXX: If faddr was bound to multicast group,
333 			 * jailed raw socket will drop datagram.
334 			 */
335 			if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
336 				goto skip_1;
337 		}
338 		last = inp;
339 		continue;
340 	skip_1:
341 		INP_RUNLOCK(inp);
342 	}
343 	CK_LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) {
344 		if (inp->inp_ip_p && inp->inp_ip_p != proto)
345 			continue;
346 #ifdef INET6
347 		/* XXX inp locking */
348 		if ((inp->inp_vflag & INP_IPV4) == 0)
349 			continue;
350 #endif
351 		if (!in_nullhost(inp->inp_laddr) &&
352 		    !in_hosteq(inp->inp_laddr, ip->ip_dst))
353 			continue;
354 		if (!in_nullhost(inp->inp_faddr) &&
355 		    !in_hosteq(inp->inp_faddr, ip->ip_src))
356 			continue;
357 		if (last != NULL) {
358 			struct mbuf *n;
359 
360 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
361 			if (n != NULL)
362 				(void) rip_append(last, ip, n, &ripsrc);
363 			/* XXX count dropped packet */
364 			INP_RUNLOCK(last);
365 			last = NULL;
366 		}
367 		INP_RLOCK(inp);
368 		if (__predict_false(inp->inp_flags2 & INP_FREED))
369 			goto skip_2;
370 		if (jailed_without_vnet(inp->inp_cred)) {
371 			/*
372 			 * Allow raw socket in jail to receive multicast;
373 			 * assume process had PRIV_NETINET_RAW at attach,
374 			 * and fall through into normal filter path if so.
375 			 */
376 			if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
377 			    prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
378 				goto skip_2;
379 		}
380 		/*
381 		 * If this raw socket has multicast state, and we
382 		 * have received a multicast, check if this socket
383 		 * should receive it, as multicast filtering is now
384 		 * the responsibility of the transport layer.
385 		 */
386 		if (inp->inp_moptions != NULL &&
387 		    IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
388 			/*
389 			 * If the incoming datagram is for IGMP, allow it
390 			 * through unconditionally to the raw socket.
391 			 *
392 			 * In the case of IGMPv2, we may not have explicitly
393 			 * joined the group, and may have set IFF_ALLMULTI
394 			 * on the interface. imo_multi_filter() may discard
395 			 * control traffic we actually need to see.
396 			 *
397 			 * Userland multicast routing daemons should continue
398 			 * filter the control traffic appropriately.
399 			 */
400 			int blocked;
401 
402 			blocked = MCAST_PASS;
403 			if (proto != IPPROTO_IGMP) {
404 				struct sockaddr_in group;
405 
406 				bzero(&group, sizeof(struct sockaddr_in));
407 				group.sin_len = sizeof(struct sockaddr_in);
408 				group.sin_family = AF_INET;
409 				group.sin_addr = ip->ip_dst;
410 
411 				blocked = imo_multi_filter(inp->inp_moptions,
412 				    ifp,
413 				    (struct sockaddr *)&group,
414 				    (struct sockaddr *)&ripsrc);
415 			}
416 
417 			if (blocked != MCAST_PASS) {
418 				IPSTAT_INC(ips_notmember);
419 				goto skip_2;
420 			}
421 		}
422 		last = inp;
423 		continue;
424 	skip_2:
425 		INP_RUNLOCK(inp);
426 	}
427 	if (last != NULL) {
428 		if (rip_append(last, ip, m, &ripsrc) != 0)
429 			IPSTAT_INC(ips_delivered);
430 		INP_RUNLOCK(last);
431 	} else {
432 		if (inetsw[ip_protox[ip->ip_p]].pr_input == rip_input) {
433 			IPSTAT_INC(ips_noproto);
434 			IPSTAT_DEC(ips_delivered);
435 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PROTOCOL, 0, 0);
436 		} else {
437 			m_freem(m);
438 		}
439 	}
440 	return (IPPROTO_DONE);
441 }
442 
443 /*
444  * Generate IP header and pass packet to ip_output.  Tack on options user may
445  * have setup with control call.
446  */
447 int
448 rip_output(struct mbuf *m, struct socket *so, ...)
449 {
450 	struct epoch_tracker et;
451 	struct ip *ip;
452 	int error;
453 	struct inpcb *inp = sotoinpcb(so);
454 	va_list ap;
455 	u_long dst;
456 	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
457 	    IP_ALLOWBROADCAST;
458 	int cnt, hlen;
459 	u_char opttype, optlen, *cp;
460 
461 	va_start(ap, so);
462 	dst = va_arg(ap, u_long);
463 	va_end(ap);
464 
465 	/*
466 	 * If the user handed us a complete IP packet, use it.  Otherwise,
467 	 * allocate an mbuf for a header and fill it in.
468 	 */
469 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
470 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
471 			m_freem(m);
472 			return(EMSGSIZE);
473 		}
474 		M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
475 		if (m == NULL)
476 			return(ENOBUFS);
477 
478 		INP_RLOCK(inp);
479 		ip = mtod(m, struct ip *);
480 		ip->ip_tos = inp->inp_ip_tos;
481 		if (inp->inp_flags & INP_DONTFRAG)
482 			ip->ip_off = htons(IP_DF);
483 		else
484 			ip->ip_off = htons(0);
485 		ip->ip_p = inp->inp_ip_p;
486 		ip->ip_len = htons(m->m_pkthdr.len);
487 		ip->ip_src = inp->inp_laddr;
488 		ip->ip_dst.s_addr = dst;
489 #ifdef ROUTE_MPATH
490 		if (CALC_FLOWID_OUTBOUND) {
491 			uint32_t hash_type, hash_val;
492 
493 			hash_val = fib4_calc_software_hash(ip->ip_src,
494 			    ip->ip_dst, 0, 0, ip->ip_p, &hash_type);
495 			m->m_pkthdr.flowid = hash_val;
496 			M_HASHTYPE_SET(m, hash_type);
497 			flags |= IP_NODEFAULTFLOWID;
498 		}
499 #endif
500 		if (jailed(inp->inp_cred)) {
501 			/*
502 			 * prison_local_ip4() would be good enough but would
503 			 * let a source of INADDR_ANY pass, which we do not
504 			 * want to see from jails.
505 			 */
506 			if (ip->ip_src.s_addr == INADDR_ANY) {
507 				NET_EPOCH_ENTER(et);
508 				error = in_pcbladdr(inp, &ip->ip_dst,
509 				    &ip->ip_src, inp->inp_cred);
510 				NET_EPOCH_EXIT(et);
511 			} else {
512 				error = prison_local_ip4(inp->inp_cred,
513 				    &ip->ip_src);
514 			}
515 			if (error != 0) {
516 				INP_RUNLOCK(inp);
517 				m_freem(m);
518 				return (error);
519 			}
520 		}
521 		ip->ip_ttl = inp->inp_ip_ttl;
522 	} else {
523 		if (m->m_pkthdr.len > IP_MAXPACKET) {
524 			m_freem(m);
525 			return(EMSGSIZE);
526 		}
527 		ip = mtod(m, struct ip *);
528 		hlen = ip->ip_hl << 2;
529 		if (m->m_len < hlen) {
530 			m = m_pullup(m, hlen);
531 			if (m == NULL)
532 				return (EINVAL);
533 			ip = mtod(m, struct ip *);
534 		}
535 #ifdef ROUTE_MPATH
536 		if (CALC_FLOWID_OUTBOUND) {
537 			uint32_t hash_type, hash_val;
538 
539 			hash_val = fib4_calc_software_hash(ip->ip_dst,
540 			    ip->ip_src, 0, 0, ip->ip_p, &hash_type);
541 			m->m_pkthdr.flowid = hash_val;
542 			M_HASHTYPE_SET(m, hash_type);
543 			flags |= IP_NODEFAULTFLOWID;
544 		}
545 #endif
546 		INP_RLOCK(inp);
547 		/*
548 		 * Don't allow both user specified and setsockopt options,
549 		 * and don't allow packet length sizes that will crash.
550 		 */
551 		if ((hlen < sizeof (*ip))
552 		    || ((hlen > sizeof (*ip)) && inp->inp_options)
553 		    || (ntohs(ip->ip_len) != m->m_pkthdr.len)) {
554 			INP_RUNLOCK(inp);
555 			m_freem(m);
556 			return (EINVAL);
557 		}
558 		error = prison_check_ip4(inp->inp_cred, &ip->ip_src);
559 		if (error != 0) {
560 			INP_RUNLOCK(inp);
561 			m_freem(m);
562 			return (error);
563 		}
564 		/*
565 		 * Don't allow IP options which do not have the required
566 		 * structure as specified in section 3.1 of RFC 791 on
567 		 * pages 15-23.
568 		 */
569 		cp = (u_char *)(ip + 1);
570 		cnt = hlen - sizeof (struct ip);
571 		for (; cnt > 0; cnt -= optlen, cp += optlen) {
572 			opttype = cp[IPOPT_OPTVAL];
573 			if (opttype == IPOPT_EOL)
574 				break;
575 			if (opttype == IPOPT_NOP) {
576 				optlen = 1;
577 				continue;
578 			}
579 			if (cnt < IPOPT_OLEN + sizeof(u_char)) {
580 				INP_RUNLOCK(inp);
581 				m_freem(m);
582 				return (EINVAL);
583 			}
584 			optlen = cp[IPOPT_OLEN];
585 			if (optlen < IPOPT_OLEN + sizeof(u_char) ||
586 			    optlen > cnt) {
587 				INP_RUNLOCK(inp);
588 				m_freem(m);
589 				return (EINVAL);
590 			}
591 		}
592 		/*
593 		 * This doesn't allow application to specify ID of zero,
594 		 * but we got this limitation from the beginning of history.
595 		 */
596 		if (ip->ip_id == 0)
597 			ip_fillid(ip);
598 
599 		/*
600 		 * XXX prevent ip_output from overwriting header fields.
601 		 */
602 		flags |= IP_RAWOUTPUT;
603 		IPSTAT_INC(ips_rawout);
604 	}
605 
606 	if (inp->inp_flags & INP_ONESBCAST)
607 		flags |= IP_SENDONES;
608 
609 #ifdef MAC
610 	mac_inpcb_create_mbuf(inp, m);
611 #endif
612 
613 	NET_EPOCH_ENTER(et);
614 	error = ip_output(m, inp->inp_options, NULL, flags,
615 	    inp->inp_moptions, inp);
616 	NET_EPOCH_EXIT(et);
617 	INP_RUNLOCK(inp);
618 	return (error);
619 }
620 
621 /*
622  * Raw IP socket option processing.
623  *
624  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
625  * only be created by a privileged process, and as such, socket option
626  * operations to manage system properties on any raw socket were allowed to
627  * take place without explicit additional access control checks.  However,
628  * raw sockets can now also be created in jail(), and therefore explicit
629  * checks are now required.  Likewise, raw sockets can be used by a process
630  * after it gives up privilege, so some caution is required.  For options
631  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
632  * performed in ip_ctloutput() and therefore no check occurs here.
633  * Unilaterally checking priv_check() here breaks normal IP socket option
634  * operations on raw sockets.
635  *
636  * When adding new socket options here, make sure to add access control
637  * checks here as necessary.
638  *
639  * XXX-BZ inp locking?
640  */
641 int
642 rip_ctloutput(struct socket *so, struct sockopt *sopt)
643 {
644 	struct	inpcb *inp = sotoinpcb(so);
645 	int	error, optval;
646 
647 	if (sopt->sopt_level != IPPROTO_IP) {
648 		if ((sopt->sopt_level == SOL_SOCKET) &&
649 		    (sopt->sopt_name == SO_SETFIB)) {
650 			inp->inp_inc.inc_fibnum = so->so_fibnum;
651 			return (0);
652 		}
653 		return (EINVAL);
654 	}
655 
656 	error = 0;
657 	switch (sopt->sopt_dir) {
658 	case SOPT_GET:
659 		switch (sopt->sopt_name) {
660 		case IP_HDRINCL:
661 			optval = inp->inp_flags & INP_HDRINCL;
662 			error = sooptcopyout(sopt, &optval, sizeof optval);
663 			break;
664 
665 		case IP_FW3:	/* generic ipfw v.3 functions */
666 		case IP_FW_ADD:	/* ADD actually returns the body... */
667 		case IP_FW_GET:
668 		case IP_FW_TABLE_GETSIZE:
669 		case IP_FW_TABLE_LIST:
670 		case IP_FW_NAT_GET_CONFIG:
671 		case IP_FW_NAT_GET_LOG:
672 			if (V_ip_fw_ctl_ptr != NULL)
673 				error = V_ip_fw_ctl_ptr(sopt);
674 			else
675 				error = ENOPROTOOPT;
676 			break;
677 
678 		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
679 		case IP_DUMMYNET_GET:
680 			if (ip_dn_ctl_ptr != NULL)
681 				error = ip_dn_ctl_ptr(sopt);
682 			else
683 				error = ENOPROTOOPT;
684 			break ;
685 
686 		case MRT_INIT:
687 		case MRT_DONE:
688 		case MRT_ADD_VIF:
689 		case MRT_DEL_VIF:
690 		case MRT_ADD_MFC:
691 		case MRT_DEL_MFC:
692 		case MRT_VERSION:
693 		case MRT_ASSERT:
694 		case MRT_API_SUPPORT:
695 		case MRT_API_CONFIG:
696 		case MRT_ADD_BW_UPCALL:
697 		case MRT_DEL_BW_UPCALL:
698 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
699 			if (error != 0)
700 				return (error);
701 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
702 				EOPNOTSUPP;
703 			break;
704 
705 		default:
706 			error = ip_ctloutput(so, sopt);
707 			break;
708 		}
709 		break;
710 
711 	case SOPT_SET:
712 		switch (sopt->sopt_name) {
713 		case IP_HDRINCL:
714 			error = sooptcopyin(sopt, &optval, sizeof optval,
715 					    sizeof optval);
716 			if (error)
717 				break;
718 			if (optval)
719 				inp->inp_flags |= INP_HDRINCL;
720 			else
721 				inp->inp_flags &= ~INP_HDRINCL;
722 			break;
723 
724 		case IP_FW3:	/* generic ipfw v.3 functions */
725 		case IP_FW_ADD:
726 		case IP_FW_DEL:
727 		case IP_FW_FLUSH:
728 		case IP_FW_ZERO:
729 		case IP_FW_RESETLOG:
730 		case IP_FW_TABLE_ADD:
731 		case IP_FW_TABLE_DEL:
732 		case IP_FW_TABLE_FLUSH:
733 		case IP_FW_NAT_CFG:
734 		case IP_FW_NAT_DEL:
735 			if (V_ip_fw_ctl_ptr != NULL)
736 				error = V_ip_fw_ctl_ptr(sopt);
737 			else
738 				error = ENOPROTOOPT;
739 			break;
740 
741 		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
742 		case IP_DUMMYNET_CONFIGURE:
743 		case IP_DUMMYNET_DEL:
744 		case IP_DUMMYNET_FLUSH:
745 			if (ip_dn_ctl_ptr != NULL)
746 				error = ip_dn_ctl_ptr(sopt);
747 			else
748 				error = ENOPROTOOPT ;
749 			break ;
750 
751 		case IP_RSVP_ON:
752 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
753 			if (error != 0)
754 				return (error);
755 			error = ip_rsvp_init(so);
756 			break;
757 
758 		case IP_RSVP_OFF:
759 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
760 			if (error != 0)
761 				return (error);
762 			error = ip_rsvp_done();
763 			break;
764 
765 		case IP_RSVP_VIF_ON:
766 		case IP_RSVP_VIF_OFF:
767 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
768 			if (error != 0)
769 				return (error);
770 			error = ip_rsvp_vif ?
771 				ip_rsvp_vif(so, sopt) : EINVAL;
772 			break;
773 
774 		case MRT_INIT:
775 		case MRT_DONE:
776 		case MRT_ADD_VIF:
777 		case MRT_DEL_VIF:
778 		case MRT_ADD_MFC:
779 		case MRT_DEL_MFC:
780 		case MRT_VERSION:
781 		case MRT_ASSERT:
782 		case MRT_API_SUPPORT:
783 		case MRT_API_CONFIG:
784 		case MRT_ADD_BW_UPCALL:
785 		case MRT_DEL_BW_UPCALL:
786 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
787 			if (error != 0)
788 				return (error);
789 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
790 					EOPNOTSUPP;
791 			break;
792 
793 		default:
794 			error = ip_ctloutput(so, sopt);
795 			break;
796 		}
797 		break;
798 	}
799 
800 	return (error);
801 }
802 
803 /*
804  * This function exists solely to receive the PRC_IFDOWN messages which are
805  * sent by if_down().  It looks for an ifaddr whose ifa_addr is sa, and calls
806  * in_ifadown() to remove all routes corresponding to that address.  It also
807  * receives the PRC_IFUP messages from if_up() and reinstalls the interface
808  * routes.
809  */
810 void
811 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
812 {
813 	struct rm_priotracker in_ifa_tracker;
814 	struct in_ifaddr *ia;
815 	struct ifnet *ifp;
816 	int err;
817 	int flags;
818 
819 	switch (cmd) {
820 	case PRC_IFDOWN:
821 		IN_IFADDR_RLOCK(&in_ifa_tracker);
822 		CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
823 			if (ia->ia_ifa.ifa_addr == sa
824 			    && (ia->ia_flags & IFA_ROUTE)) {
825 				ifa_ref(&ia->ia_ifa);
826 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
827 				/*
828 				 * in_scrubprefix() kills the interface route.
829 				 */
830 				in_scrubprefix(ia, 0);
831 				/*
832 				 * in_ifadown gets rid of all the rest of the
833 				 * routes.  This is not quite the right thing
834 				 * to do, but at least if we are running a
835 				 * routing process they will come back.
836 				 */
837 				in_ifadown(&ia->ia_ifa, 0);
838 				ifa_free(&ia->ia_ifa);
839 				break;
840 			}
841 		}
842 		if (ia == NULL)		/* If ia matched, already unlocked. */
843 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
844 		break;
845 
846 	case PRC_IFUP:
847 		IN_IFADDR_RLOCK(&in_ifa_tracker);
848 		CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
849 			if (ia->ia_ifa.ifa_addr == sa)
850 				break;
851 		}
852 		if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) {
853 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
854 			return;
855 		}
856 		ifa_ref(&ia->ia_ifa);
857 		IN_IFADDR_RUNLOCK(&in_ifa_tracker);
858 		flags = RTF_UP;
859 		ifp = ia->ia_ifa.ifa_ifp;
860 
861 		if ((ifp->if_flags & IFF_LOOPBACK)
862 		    || (ifp->if_flags & IFF_POINTOPOINT))
863 			flags |= RTF_HOST;
864 
865 		err = ifa_del_loopback_route((struct ifaddr *)ia, sa);
866 
867 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
868 		if (err == 0)
869 			ia->ia_flags |= IFA_ROUTE;
870 
871 		err = ifa_add_loopback_route((struct ifaddr *)ia, sa);
872 
873 		ifa_free(&ia->ia_ifa);
874 		break;
875 	}
876 }
877 
878 static int
879 rip_attach(struct socket *so, int proto, struct thread *td)
880 {
881 	struct inpcb *inp;
882 	int error;
883 
884 	inp = sotoinpcb(so);
885 	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
886 
887 	error = priv_check(td, PRIV_NETINET_RAW);
888 	if (error)
889 		return (error);
890 	if (proto >= IPPROTO_MAX || proto < 0)
891 		return EPROTONOSUPPORT;
892 	error = soreserve(so, rip_sendspace, rip_recvspace);
893 	if (error)
894 		return (error);
895 	INP_INFO_WLOCK(&V_ripcbinfo);
896 	error = in_pcballoc(so, &V_ripcbinfo);
897 	if (error) {
898 		INP_INFO_WUNLOCK(&V_ripcbinfo);
899 		return (error);
900 	}
901 	inp = (struct inpcb *)so->so_pcb;
902 	inp->inp_vflag |= INP_IPV4;
903 	inp->inp_ip_p = proto;
904 	inp->inp_ip_ttl = V_ip_defttl;
905 	rip_inshash(inp);
906 	INP_INFO_WUNLOCK(&V_ripcbinfo);
907 	INP_WUNLOCK(inp);
908 	return (0);
909 }
910 
911 static void
912 rip_detach(struct socket *so)
913 {
914 	struct inpcb *inp;
915 
916 	inp = sotoinpcb(so);
917 	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
918 	KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
919 	    ("rip_detach: not closed"));
920 
921 	INP_INFO_WLOCK(&V_ripcbinfo);
922 	INP_WLOCK(inp);
923 	rip_delhash(inp);
924 	if (so == V_ip_mrouter && ip_mrouter_done)
925 		ip_mrouter_done();
926 	if (ip_rsvp_force_done)
927 		ip_rsvp_force_done(so);
928 	if (so == V_ip_rsvpd)
929 		ip_rsvp_done();
930 	in_pcbdetach(inp);
931 	in_pcbfree(inp);
932 	INP_INFO_WUNLOCK(&V_ripcbinfo);
933 }
934 
935 static void
936 rip_dodisconnect(struct socket *so, struct inpcb *inp)
937 {
938 	struct inpcbinfo *pcbinfo;
939 
940 	pcbinfo = inp->inp_pcbinfo;
941 	INP_INFO_WLOCK(pcbinfo);
942 	INP_WLOCK(inp);
943 	rip_delhash(inp);
944 	inp->inp_faddr.s_addr = INADDR_ANY;
945 	rip_inshash(inp);
946 	SOCK_LOCK(so);
947 	so->so_state &= ~SS_ISCONNECTED;
948 	SOCK_UNLOCK(so);
949 	INP_WUNLOCK(inp);
950 	INP_INFO_WUNLOCK(pcbinfo);
951 }
952 
953 static void
954 rip_abort(struct socket *so)
955 {
956 	struct inpcb *inp;
957 
958 	inp = sotoinpcb(so);
959 	KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
960 
961 	rip_dodisconnect(so, inp);
962 }
963 
964 static void
965 rip_close(struct socket *so)
966 {
967 	struct inpcb *inp;
968 
969 	inp = sotoinpcb(so);
970 	KASSERT(inp != NULL, ("rip_close: inp == NULL"));
971 
972 	rip_dodisconnect(so, inp);
973 }
974 
975 static int
976 rip_disconnect(struct socket *so)
977 {
978 	struct inpcb *inp;
979 
980 	if ((so->so_state & SS_ISCONNECTED) == 0)
981 		return (ENOTCONN);
982 
983 	inp = sotoinpcb(so);
984 	KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
985 
986 	rip_dodisconnect(so, inp);
987 	return (0);
988 }
989 
990 static int
991 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
992 {
993 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
994 	struct inpcb *inp;
995 	int error;
996 
997 	if (nam->sa_len != sizeof(*addr))
998 		return (EINVAL);
999 
1000 	error = prison_check_ip4(td->td_ucred, &addr->sin_addr);
1001 	if (error != 0)
1002 		return (error);
1003 
1004 	inp = sotoinpcb(so);
1005 	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
1006 
1007 	if (CK_STAILQ_EMPTY(&V_ifnet) ||
1008 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
1009 	    (addr->sin_addr.s_addr &&
1010 	     (inp->inp_flags & INP_BINDANY) == 0 &&
1011 	     ifa_ifwithaddr_check((struct sockaddr *)addr) == 0))
1012 		return (EADDRNOTAVAIL);
1013 
1014 	INP_INFO_WLOCK(&V_ripcbinfo);
1015 	INP_WLOCK(inp);
1016 	rip_delhash(inp);
1017 	inp->inp_laddr = addr->sin_addr;
1018 	rip_inshash(inp);
1019 	INP_WUNLOCK(inp);
1020 	INP_INFO_WUNLOCK(&V_ripcbinfo);
1021 	return (0);
1022 }
1023 
1024 static int
1025 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1026 {
1027 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
1028 	struct inpcb *inp;
1029 
1030 	if (nam->sa_len != sizeof(*addr))
1031 		return (EINVAL);
1032 	if (CK_STAILQ_EMPTY(&V_ifnet))
1033 		return (EADDRNOTAVAIL);
1034 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
1035 		return (EAFNOSUPPORT);
1036 
1037 	inp = sotoinpcb(so);
1038 	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
1039 
1040 	INP_INFO_WLOCK(&V_ripcbinfo);
1041 	INP_WLOCK(inp);
1042 	rip_delhash(inp);
1043 	inp->inp_faddr = addr->sin_addr;
1044 	rip_inshash(inp);
1045 	soisconnected(so);
1046 	INP_WUNLOCK(inp);
1047 	INP_INFO_WUNLOCK(&V_ripcbinfo);
1048 	return (0);
1049 }
1050 
1051 static int
1052 rip_shutdown(struct socket *so)
1053 {
1054 	struct inpcb *inp;
1055 
1056 	inp = sotoinpcb(so);
1057 	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
1058 
1059 	INP_WLOCK(inp);
1060 	socantsendmore(so);
1061 	INP_WUNLOCK(inp);
1062 	return (0);
1063 }
1064 
1065 static int
1066 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
1067     struct mbuf *control, struct thread *td)
1068 {
1069 	struct inpcb *inp;
1070 	u_long dst;
1071 
1072 	inp = sotoinpcb(so);
1073 	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
1074 
1075 	/*
1076 	 * Note: 'dst' reads below are unlocked.
1077 	 */
1078 	if (so->so_state & SS_ISCONNECTED) {
1079 		if (nam) {
1080 			m_freem(m);
1081 			return (EISCONN);
1082 		}
1083 		dst = inp->inp_faddr.s_addr;	/* Unlocked read. */
1084 	} else {
1085 		if (nam == NULL) {
1086 			m_freem(m);
1087 			return (ENOTCONN);
1088 		}
1089 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
1090 	}
1091 	return (rip_output(m, so, dst));
1092 }
1093 #endif /* INET */
1094 
1095 static int
1096 rip_pcblist(SYSCTL_HANDLER_ARGS)
1097 {
1098 	struct xinpgen xig;
1099 	struct epoch_tracker et;
1100 	struct inpcb *inp;
1101 	int error;
1102 
1103 	if (req->newptr != 0)
1104 		return (EPERM);
1105 
1106 	if (req->oldptr == 0) {
1107 		int n;
1108 
1109 		n = V_ripcbinfo.ipi_count;
1110 		n += imax(n / 8, 10);
1111 		req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
1112 		return (0);
1113 	}
1114 
1115 	if ((error = sysctl_wire_old_buffer(req, 0)) != 0)
1116 		return (error);
1117 
1118 	bzero(&xig, sizeof(xig));
1119 	xig.xig_len = sizeof xig;
1120 	xig.xig_count = V_ripcbinfo.ipi_count;
1121 	xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1122 	xig.xig_sogen = so_gencnt;
1123 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1124 	if (error)
1125 		return (error);
1126 
1127 	NET_EPOCH_ENTER(et);
1128 	for (inp = CK_LIST_FIRST(V_ripcbinfo.ipi_listhead);
1129 	    inp != NULL;
1130 	    inp = CK_LIST_NEXT(inp, inp_list)) {
1131 		INP_RLOCK(inp);
1132 		if (inp->inp_gencnt <= xig.xig_gen &&
1133 		    cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
1134 			struct xinpcb xi;
1135 
1136 			in_pcbtoxinpcb(inp, &xi);
1137 			INP_RUNLOCK(inp);
1138 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1139 			if (error)
1140 				break;
1141 		} else
1142 			INP_RUNLOCK(inp);
1143 	}
1144 	NET_EPOCH_EXIT(et);
1145 
1146 	if (!error) {
1147 		/*
1148 		 * Give the user an updated idea of our state.  If the
1149 		 * generation differs from what we told her before, she knows
1150 		 * that something happened while we were processing this
1151 		 * request, and it might be necessary to retry.
1152 		 */
1153 		xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1154 		xig.xig_sogen = so_gencnt;
1155 		xig.xig_count = V_ripcbinfo.ipi_count;
1156 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1157 	}
1158 
1159 	return (error);
1160 }
1161 
1162 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist,
1163     CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
1164     rip_pcblist, "S,xinpcb",
1165     "List of active raw IP sockets");
1166 
1167 #ifdef INET
1168 struct pr_usrreqs rip_usrreqs = {
1169 	.pru_abort =		rip_abort,
1170 	.pru_attach =		rip_attach,
1171 	.pru_bind =		rip_bind,
1172 	.pru_connect =		rip_connect,
1173 	.pru_control =		in_control,
1174 	.pru_detach =		rip_detach,
1175 	.pru_disconnect =	rip_disconnect,
1176 	.pru_peeraddr =		in_getpeeraddr,
1177 	.pru_send =		rip_send,
1178 	.pru_shutdown =		rip_shutdown,
1179 	.pru_sockaddr =		in_getsockaddr,
1180 	.pru_sosetlabel =	in_pcbsosetlabel,
1181 	.pru_close =		rip_close,
1182 };
1183 #endif /* INET */
1184