xref: /freebsd-12.1/sys/netinet/raw_ip.c (revision 62985abe)
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 **, int, struct ip_fw_args *);
104 void	(*ip_divert_ptr)(struct mbuf *, int);
105 int	(*ng_ipfw_input_p)(struct mbuf **, int,
106 			struct ip_fw_args *, int);
107 
108 #ifdef INET
109 /*
110  * Hooks for multicast routing. They all default to NULL, so leave them not
111  * initialized and rely on BSS being set to 0.
112  */
113 
114 /*
115  * The socket used to communicate with the multicast routing daemon.
116  */
117 VNET_DEFINE(struct socket *, ip_mrouter);
118 
119 /*
120  * The various mrouter and rsvp functions.
121  */
122 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
123 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
124 int (*ip_mrouter_done)(void);
125 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
126 		   struct ip_moptions *);
127 int (*mrt_ioctl)(u_long, caddr_t, int);
128 int (*legal_vif_num)(int);
129 u_long (*ip_mcast_src)(int);
130 
131 int (*rsvp_input_p)(struct mbuf **, int *, int);
132 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
133 void (*ip_rsvp_force_done)(struct socket *);
134 #endif /* INET */
135 
136 extern	struct protosw inetsw[];
137 
138 u_long	rip_sendspace = 9216;
139 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
140     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
141 
142 u_long	rip_recvspace = 9216;
143 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
144     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
145 
146 /*
147  * Hash functions
148  */
149 
150 #define INP_PCBHASH_RAW_SIZE	256
151 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \
152         (((proto) + (laddr) + (faddr)) % (mask) + 1)
153 
154 #ifdef INET
155 static void
rip_inshash(struct inpcb * inp)156 rip_inshash(struct inpcb *inp)
157 {
158 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
159 	struct inpcbhead *pcbhash;
160 	int hash;
161 
162 	INP_INFO_WLOCK_ASSERT(pcbinfo);
163 	INP_WLOCK_ASSERT(inp);
164 
165 	if (inp->inp_ip_p != 0 &&
166 	    inp->inp_laddr.s_addr != INADDR_ANY &&
167 	    inp->inp_faddr.s_addr != INADDR_ANY) {
168 		hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr,
169 		    inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask);
170 	} else
171 		hash = 0;
172 	pcbhash = &pcbinfo->ipi_hashbase[hash];
173 	CK_LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
174 }
175 
176 static void
rip_delhash(struct inpcb * inp)177 rip_delhash(struct inpcb *inp)
178 {
179 
180 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
181 	INP_WLOCK_ASSERT(inp);
182 
183 	CK_LIST_REMOVE(inp, inp_hash);
184 }
185 #endif /* INET */
186 
187 /*
188  * Raw interface to IP protocol.
189  */
190 
191 /*
192  * Initialize raw connection block q.
193  */
194 static void
rip_zone_change(void * tag)195 rip_zone_change(void *tag)
196 {
197 
198 	uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets);
199 }
200 
201 static int
rip_inpcb_init(void * mem,int size,int flags)202 rip_inpcb_init(void *mem, int size, int flags)
203 {
204 	struct inpcb *inp = mem;
205 
206 	INP_LOCK_INIT(inp, "inp", "rawinp");
207 	return (0);
208 }
209 
210 void
rip_init(void)211 rip_init(void)
212 {
213 
214 	in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE,
215 	    1, "ripcb", rip_inpcb_init, IPI_HASHFIELDS_NONE);
216 	EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL,
217 	    EVENTHANDLER_PRI_ANY);
218 }
219 
220 #ifdef VIMAGE
221 static void
rip_destroy(void * unused __unused)222 rip_destroy(void *unused __unused)
223 {
224 
225 	in_pcbinfo_destroy(&V_ripcbinfo);
226 }
227 VNET_SYSUNINIT(raw_ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, rip_destroy, NULL);
228 #endif
229 
230 #ifdef INET
231 static int
rip_append(struct inpcb * last,struct ip * ip,struct mbuf * n,struct sockaddr_in * ripsrc)232 rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
233     struct sockaddr_in *ripsrc)
234 {
235 	int policyfail = 0;
236 
237 	INP_LOCK_ASSERT(last);
238 
239 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
240 	/* check AH/ESP integrity. */
241 	if (IPSEC_ENABLED(ipv4)) {
242 		if (IPSEC_CHECK_POLICY(ipv4, n, last) != 0)
243 			policyfail = 1;
244 	}
245 #endif /* IPSEC */
246 #ifdef MAC
247 	if (!policyfail && mac_inpcb_check_deliver(last, n) != 0)
248 		policyfail = 1;
249 #endif
250 	/* Check the minimum TTL for socket. */
251 	if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
252 		policyfail = 1;
253 	if (!policyfail) {
254 		struct mbuf *opts = NULL;
255 		struct socket *so;
256 
257 		so = last->inp_socket;
258 		if ((last->inp_flags & INP_CONTROLOPTS) ||
259 		    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
260 			ip_savecontrol(last, &opts, ip, n);
261 		SOCKBUF_LOCK(&so->so_rcv);
262 		if (sbappendaddr_locked(&so->so_rcv,
263 		    (struct sockaddr *)ripsrc, n, opts) == 0) {
264 			/* should notify about lost packet */
265 			m_freem(n);
266 			if (opts)
267 				m_freem(opts);
268 			SOCKBUF_UNLOCK(&so->so_rcv);
269 		} else
270 			sorwakeup_locked(so);
271 	} else
272 		m_freem(n);
273 	return (policyfail);
274 }
275 
276 /*
277  * Setup generic address and protocol structures for raw_input routine, then
278  * pass them along with mbuf chain.
279  */
280 int
rip_input(struct mbuf ** mp,int * offp,int proto)281 rip_input(struct mbuf **mp, int *offp, int proto)
282 {
283 	struct ifnet *ifp;
284 	struct mbuf *m = *mp;
285 	struct ip *ip = mtod(m, struct ip *);
286 	struct inpcb *inp, *last;
287 	struct sockaddr_in ripsrc;
288 	struct epoch_tracker et;
289 	int hash;
290 
291 	*mp = NULL;
292 
293 	bzero(&ripsrc, sizeof(ripsrc));
294 	ripsrc.sin_len = sizeof(ripsrc);
295 	ripsrc.sin_family = AF_INET;
296 	ripsrc.sin_addr = ip->ip_src;
297 	last = NULL;
298 
299 	ifp = m->m_pkthdr.rcvif;
300 
301 	hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr,
302 	    ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
303 	INP_INFO_RLOCK_ET(&V_ripcbinfo, et);
304 	CK_LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) {
305 		if (inp->inp_ip_p != proto)
306 			continue;
307 #ifdef INET6
308 		/* XXX inp locking */
309 		if ((inp->inp_vflag & INP_IPV4) == 0)
310 			continue;
311 #endif
312 		if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
313 			continue;
314 		if (inp->inp_faddr.s_addr != ip->ip_src.s_addr)
315 			continue;
316 		if (last != NULL) {
317 			struct mbuf *n;
318 
319 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
320 			if (n != NULL)
321 			    (void) rip_append(last, ip, n, &ripsrc);
322 			/* XXX count dropped packet */
323 			INP_RUNLOCK(last);
324 			last = NULL;
325 		}
326 		INP_RLOCK(inp);
327 		if (__predict_false(inp->inp_flags2 & INP_FREED))
328 			goto skip_1;
329 		if (jailed_without_vnet(inp->inp_cred)) {
330 			/*
331 			 * XXX: If faddr was bound to multicast group,
332 			 * jailed raw socket will drop datagram.
333 			 */
334 			if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
335 				goto skip_1;
336 		}
337 		last = inp;
338 		continue;
339 	skip_1:
340 		INP_RUNLOCK(inp);
341 	}
342 	CK_LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) {
343 		if (inp->inp_ip_p && inp->inp_ip_p != proto)
344 			continue;
345 #ifdef INET6
346 		/* XXX inp locking */
347 		if ((inp->inp_vflag & INP_IPV4) == 0)
348 			continue;
349 #endif
350 		if (!in_nullhost(inp->inp_laddr) &&
351 		    !in_hosteq(inp->inp_laddr, ip->ip_dst))
352 			continue;
353 		if (!in_nullhost(inp->inp_faddr) &&
354 		    !in_hosteq(inp->inp_faddr, ip->ip_src))
355 			continue;
356 		if (last != NULL) {
357 			struct mbuf *n;
358 
359 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
360 			if (n != NULL)
361 				(void) rip_append(last, ip, n, &ripsrc);
362 			/* XXX count dropped packet */
363 			INP_RUNLOCK(last);
364 			last = NULL;
365 		}
366 		INP_RLOCK(inp);
367 		if (__predict_false(inp->inp_flags2 & INP_FREED))
368 			goto skip_2;
369 		if (jailed_without_vnet(inp->inp_cred)) {
370 			/*
371 			 * Allow raw socket in jail to receive multicast;
372 			 * assume process had PRIV_NETINET_RAW at attach,
373 			 * and fall through into normal filter path if so.
374 			 */
375 			if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
376 			    prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
377 				goto skip_2;
378 		}
379 		/*
380 		 * If this raw socket has multicast state, and we
381 		 * have received a multicast, check if this socket
382 		 * should receive it, as multicast filtering is now
383 		 * the responsibility of the transport layer.
384 		 */
385 		if (inp->inp_moptions != NULL &&
386 		    IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
387 			/*
388 			 * If the incoming datagram is for IGMP, allow it
389 			 * through unconditionally to the raw socket.
390 			 *
391 			 * In the case of IGMPv2, we may not have explicitly
392 			 * joined the group, and may have set IFF_ALLMULTI
393 			 * on the interface. imo_multi_filter() may discard
394 			 * control traffic we actually need to see.
395 			 *
396 			 * Userland multicast routing daemons should continue
397 			 * filter the control traffic appropriately.
398 			 */
399 			int blocked;
400 
401 			blocked = MCAST_PASS;
402 			if (proto != IPPROTO_IGMP) {
403 				struct sockaddr_in group;
404 
405 				bzero(&group, sizeof(struct sockaddr_in));
406 				group.sin_len = sizeof(struct sockaddr_in);
407 				group.sin_family = AF_INET;
408 				group.sin_addr = ip->ip_dst;
409 
410 				blocked = imo_multi_filter(inp->inp_moptions,
411 				    ifp,
412 				    (struct sockaddr *)&group,
413 				    (struct sockaddr *)&ripsrc);
414 			}
415 
416 			if (blocked != MCAST_PASS) {
417 				IPSTAT_INC(ips_notmember);
418 				goto skip_2;
419 			}
420 		}
421 		last = inp;
422 		continue;
423 	skip_2:
424 		INP_RUNLOCK(inp);
425 	}
426 	INP_INFO_RUNLOCK_ET(&V_ripcbinfo, et);
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
rip_output(struct mbuf * m,struct socket * so,...)448 rip_output(struct mbuf *m, struct socket *so, ...)
449 {
450 	struct ip *ip;
451 	int error;
452 	struct inpcb *inp = sotoinpcb(so);
453 	va_list ap;
454 	u_long dst;
455 	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
456 	    IP_ALLOWBROADCAST;
457 	int cnt, hlen;
458 	u_char opttype, optlen, *cp;
459 
460 	va_start(ap, so);
461 	dst = va_arg(ap, u_long);
462 	va_end(ap);
463 
464 	/*
465 	 * If the user handed us a complete IP packet, use it.  Otherwise,
466 	 * allocate an mbuf for a header and fill it in.
467 	 */
468 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
469 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
470 			m_freem(m);
471 			return(EMSGSIZE);
472 		}
473 		M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
474 		if (m == NULL)
475 			return(ENOBUFS);
476 
477 		INP_RLOCK(inp);
478 		ip = mtod(m, struct ip *);
479 		ip->ip_tos = inp->inp_ip_tos;
480 		if (inp->inp_flags & INP_DONTFRAG)
481 			ip->ip_off = htons(IP_DF);
482 		else
483 			ip->ip_off = htons(0);
484 		ip->ip_p = inp->inp_ip_p;
485 		ip->ip_len = htons(m->m_pkthdr.len);
486 		ip->ip_src = inp->inp_laddr;
487 		ip->ip_dst.s_addr = dst;
488 		if (jailed(inp->inp_cred)) {
489 			/*
490 			 * prison_local_ip4() would be good enough but would
491 			 * let a source of INADDR_ANY pass, which we do not
492 			 * want to see from jails.
493 			 */
494 			if (ip->ip_src.s_addr == INADDR_ANY) {
495 				error = in_pcbladdr(inp, &ip->ip_dst, &ip->ip_src,
496 				    inp->inp_cred);
497 			} else {
498 				error = prison_local_ip4(inp->inp_cred,
499 				    &ip->ip_src);
500 			}
501 			if (error != 0) {
502 				INP_RUNLOCK(inp);
503 				m_freem(m);
504 				return (error);
505 			}
506 		}
507 		ip->ip_ttl = inp->inp_ip_ttl;
508 	} else {
509 		if (m->m_pkthdr.len > IP_MAXPACKET) {
510 			m_freem(m);
511 			return(EMSGSIZE);
512 		}
513 		ip = mtod(m, struct ip *);
514 		hlen = ip->ip_hl << 2;
515 		if (m->m_len < hlen) {
516 			m = m_pullup(m, hlen);
517 			if (m == NULL)
518 				return (EINVAL);
519 			ip = mtod(m, struct ip *);
520 		}
521 
522 		INP_RLOCK(inp);
523 		/*
524 		 * Don't allow both user specified and setsockopt options,
525 		 * and don't allow packet length sizes that will crash.
526 		 */
527 		if ((hlen < sizeof (*ip))
528 		    || ((hlen > sizeof (*ip)) && inp->inp_options)
529 		    || (ntohs(ip->ip_len) != m->m_pkthdr.len)) {
530 			INP_RUNLOCK(inp);
531 			m_freem(m);
532 			return (EINVAL);
533 		}
534 		error = prison_check_ip4(inp->inp_cred, &ip->ip_src);
535 		if (error != 0) {
536 			INP_RUNLOCK(inp);
537 			m_freem(m);
538 			return (error);
539 		}
540 		/*
541 		 * Don't allow IP options which do not have the required
542 		 * structure as specified in section 3.1 of RFC 791 on
543 		 * pages 15-23.
544 		 */
545 		cp = (u_char *)(ip + 1);
546 		cnt = hlen - sizeof (struct ip);
547 		for (; cnt > 0; cnt -= optlen, cp += optlen) {
548 			opttype = cp[IPOPT_OPTVAL];
549 			if (opttype == IPOPT_EOL)
550 				break;
551 			if (opttype == IPOPT_NOP) {
552 				optlen = 1;
553 				continue;
554 			}
555 			if (cnt < IPOPT_OLEN + sizeof(u_char)) {
556 				INP_RUNLOCK(inp);
557 				m_freem(m);
558 				return (EINVAL);
559 			}
560 			optlen = cp[IPOPT_OLEN];
561 			if (optlen < IPOPT_OLEN + sizeof(u_char) ||
562 			    optlen > cnt) {
563 				INP_RUNLOCK(inp);
564 				m_freem(m);
565 				return (EINVAL);
566 			}
567 		}
568 		/*
569 		 * This doesn't allow application to specify ID of zero,
570 		 * but we got this limitation from the beginning of history.
571 		 */
572 		if (ip->ip_id == 0)
573 			ip_fillid(ip);
574 
575 		/*
576 		 * XXX prevent ip_output from overwriting header fields.
577 		 */
578 		flags |= IP_RAWOUTPUT;
579 		IPSTAT_INC(ips_rawout);
580 	}
581 
582 	if (inp->inp_flags & INP_ONESBCAST)
583 		flags |= IP_SENDONES;
584 
585 #ifdef MAC
586 	mac_inpcb_create_mbuf(inp, m);
587 #endif
588 
589 	error = ip_output(m, inp->inp_options, NULL, flags,
590 	    inp->inp_moptions, inp);
591 	INP_RUNLOCK(inp);
592 	return (error);
593 }
594 
595 /*
596  * Raw IP socket option processing.
597  *
598  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
599  * only be created by a privileged process, and as such, socket option
600  * operations to manage system properties on any raw socket were allowed to
601  * take place without explicit additional access control checks.  However,
602  * raw sockets can now also be created in jail(), and therefore explicit
603  * checks are now required.  Likewise, raw sockets can be used by a process
604  * after it gives up privilege, so some caution is required.  For options
605  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
606  * performed in ip_ctloutput() and therefore no check occurs here.
607  * Unilaterally checking priv_check() here breaks normal IP socket option
608  * operations on raw sockets.
609  *
610  * When adding new socket options here, make sure to add access control
611  * checks here as necessary.
612  *
613  * XXX-BZ inp locking?
614  */
615 int
rip_ctloutput(struct socket * so,struct sockopt * sopt)616 rip_ctloutput(struct socket *so, struct sockopt *sopt)
617 {
618 	struct	inpcb *inp = sotoinpcb(so);
619 	int	error, optval;
620 
621 	if (sopt->sopt_level != IPPROTO_IP) {
622 		if ((sopt->sopt_level == SOL_SOCKET) &&
623 		    (sopt->sopt_name == SO_SETFIB)) {
624 			inp->inp_inc.inc_fibnum = so->so_fibnum;
625 			return (0);
626 		}
627 		return (EINVAL);
628 	}
629 
630 	error = 0;
631 	switch (sopt->sopt_dir) {
632 	case SOPT_GET:
633 		switch (sopt->sopt_name) {
634 		case IP_HDRINCL:
635 			optval = inp->inp_flags & INP_HDRINCL;
636 			error = sooptcopyout(sopt, &optval, sizeof optval);
637 			break;
638 
639 		case IP_FW3:	/* generic ipfw v.3 functions */
640 		case IP_FW_ADD:	/* ADD actually returns the body... */
641 		case IP_FW_GET:
642 		case IP_FW_TABLE_GETSIZE:
643 		case IP_FW_TABLE_LIST:
644 		case IP_FW_NAT_GET_CONFIG:
645 		case IP_FW_NAT_GET_LOG:
646 			if (V_ip_fw_ctl_ptr != NULL)
647 				error = V_ip_fw_ctl_ptr(sopt);
648 			else
649 				error = ENOPROTOOPT;
650 			break;
651 
652 		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
653 		case IP_DUMMYNET_GET:
654 			if (ip_dn_ctl_ptr != NULL)
655 				error = ip_dn_ctl_ptr(sopt);
656 			else
657 				error = ENOPROTOOPT;
658 			break ;
659 
660 		case MRT_INIT:
661 		case MRT_DONE:
662 		case MRT_ADD_VIF:
663 		case MRT_DEL_VIF:
664 		case MRT_ADD_MFC:
665 		case MRT_DEL_MFC:
666 		case MRT_VERSION:
667 		case MRT_ASSERT:
668 		case MRT_API_SUPPORT:
669 		case MRT_API_CONFIG:
670 		case MRT_ADD_BW_UPCALL:
671 		case MRT_DEL_BW_UPCALL:
672 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
673 			if (error != 0)
674 				return (error);
675 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
676 				EOPNOTSUPP;
677 			break;
678 
679 		default:
680 			error = ip_ctloutput(so, sopt);
681 			break;
682 		}
683 		break;
684 
685 	case SOPT_SET:
686 		switch (sopt->sopt_name) {
687 		case IP_HDRINCL:
688 			error = sooptcopyin(sopt, &optval, sizeof optval,
689 					    sizeof optval);
690 			if (error)
691 				break;
692 			if (optval)
693 				inp->inp_flags |= INP_HDRINCL;
694 			else
695 				inp->inp_flags &= ~INP_HDRINCL;
696 			break;
697 
698 		case IP_FW3:	/* generic ipfw v.3 functions */
699 		case IP_FW_ADD:
700 		case IP_FW_DEL:
701 		case IP_FW_FLUSH:
702 		case IP_FW_ZERO:
703 		case IP_FW_RESETLOG:
704 		case IP_FW_TABLE_ADD:
705 		case IP_FW_TABLE_DEL:
706 		case IP_FW_TABLE_FLUSH:
707 		case IP_FW_NAT_CFG:
708 		case IP_FW_NAT_DEL:
709 			if (V_ip_fw_ctl_ptr != NULL)
710 				error = V_ip_fw_ctl_ptr(sopt);
711 			else
712 				error = ENOPROTOOPT;
713 			break;
714 
715 		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
716 		case IP_DUMMYNET_CONFIGURE:
717 		case IP_DUMMYNET_DEL:
718 		case IP_DUMMYNET_FLUSH:
719 			if (ip_dn_ctl_ptr != NULL)
720 				error = ip_dn_ctl_ptr(sopt);
721 			else
722 				error = ENOPROTOOPT ;
723 			break ;
724 
725 		case IP_RSVP_ON:
726 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
727 			if (error != 0)
728 				return (error);
729 			error = ip_rsvp_init(so);
730 			break;
731 
732 		case IP_RSVP_OFF:
733 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
734 			if (error != 0)
735 				return (error);
736 			error = ip_rsvp_done();
737 			break;
738 
739 		case IP_RSVP_VIF_ON:
740 		case IP_RSVP_VIF_OFF:
741 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
742 			if (error != 0)
743 				return (error);
744 			error = ip_rsvp_vif ?
745 				ip_rsvp_vif(so, sopt) : EINVAL;
746 			break;
747 
748 		case MRT_INIT:
749 		case MRT_DONE:
750 		case MRT_ADD_VIF:
751 		case MRT_DEL_VIF:
752 		case MRT_ADD_MFC:
753 		case MRT_DEL_MFC:
754 		case MRT_VERSION:
755 		case MRT_ASSERT:
756 		case MRT_API_SUPPORT:
757 		case MRT_API_CONFIG:
758 		case MRT_ADD_BW_UPCALL:
759 		case MRT_DEL_BW_UPCALL:
760 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
761 			if (error != 0)
762 				return (error);
763 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
764 					EOPNOTSUPP;
765 			break;
766 
767 		default:
768 			error = ip_ctloutput(so, sopt);
769 			break;
770 		}
771 		break;
772 	}
773 
774 	return (error);
775 }
776 
777 /*
778  * This function exists solely to receive the PRC_IFDOWN messages which are
779  * sent by if_down().  It looks for an ifaddr whose ifa_addr is sa, and calls
780  * in_ifadown() to remove all routes corresponding to that address.  It also
781  * receives the PRC_IFUP messages from if_up() and reinstalls the interface
782  * routes.
783  */
784 void
rip_ctlinput(int cmd,struct sockaddr * sa,void * vip)785 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
786 {
787 	struct rm_priotracker in_ifa_tracker;
788 	struct in_ifaddr *ia;
789 	struct ifnet *ifp;
790 	int err;
791 	int flags;
792 
793 	switch (cmd) {
794 	case PRC_IFDOWN:
795 		IN_IFADDR_RLOCK(&in_ifa_tracker);
796 		CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
797 			if (ia->ia_ifa.ifa_addr == sa
798 			    && (ia->ia_flags & IFA_ROUTE)) {
799 				ifa_ref(&ia->ia_ifa);
800 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
801 				/*
802 				 * in_scrubprefix() kills the interface route.
803 				 */
804 				in_scrubprefix(ia, 0);
805 				/*
806 				 * in_ifadown gets rid of all the rest of the
807 				 * routes.  This is not quite the right thing
808 				 * to do, but at least if we are running a
809 				 * routing process they will come back.
810 				 */
811 				in_ifadown(&ia->ia_ifa, 0);
812 				ifa_free(&ia->ia_ifa);
813 				break;
814 			}
815 		}
816 		if (ia == NULL)		/* If ia matched, already unlocked. */
817 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
818 		break;
819 
820 	case PRC_IFUP:
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 				break;
825 		}
826 		if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) {
827 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
828 			return;
829 		}
830 		ifa_ref(&ia->ia_ifa);
831 		IN_IFADDR_RUNLOCK(&in_ifa_tracker);
832 		flags = RTF_UP;
833 		ifp = ia->ia_ifa.ifa_ifp;
834 
835 		if ((ifp->if_flags & IFF_LOOPBACK)
836 		    || (ifp->if_flags & IFF_POINTOPOINT))
837 			flags |= RTF_HOST;
838 
839 		err = ifa_del_loopback_route((struct ifaddr *)ia, sa);
840 
841 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
842 		if (err == 0)
843 			ia->ia_flags |= IFA_ROUTE;
844 
845 		err = ifa_add_loopback_route((struct ifaddr *)ia, sa);
846 
847 		ifa_free(&ia->ia_ifa);
848 		break;
849 	}
850 }
851 
852 static int
rip_attach(struct socket * so,int proto,struct thread * td)853 rip_attach(struct socket *so, int proto, struct thread *td)
854 {
855 	struct inpcb *inp;
856 	int error;
857 
858 	inp = sotoinpcb(so);
859 	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
860 
861 	error = priv_check(td, PRIV_NETINET_RAW);
862 	if (error)
863 		return (error);
864 	if (proto >= IPPROTO_MAX || proto < 0)
865 		return EPROTONOSUPPORT;
866 	error = soreserve(so, rip_sendspace, rip_recvspace);
867 	if (error)
868 		return (error);
869 	INP_INFO_WLOCK(&V_ripcbinfo);
870 	error = in_pcballoc(so, &V_ripcbinfo);
871 	if (error) {
872 		INP_INFO_WUNLOCK(&V_ripcbinfo);
873 		return (error);
874 	}
875 	inp = (struct inpcb *)so->so_pcb;
876 	inp->inp_vflag |= INP_IPV4;
877 	inp->inp_ip_p = proto;
878 	inp->inp_ip_ttl = V_ip_defttl;
879 	rip_inshash(inp);
880 	INP_INFO_WUNLOCK(&V_ripcbinfo);
881 	INP_WUNLOCK(inp);
882 	return (0);
883 }
884 
885 static void
rip_detach(struct socket * so)886 rip_detach(struct socket *so)
887 {
888 	struct inpcb *inp;
889 
890 	inp = sotoinpcb(so);
891 	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
892 	KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
893 	    ("rip_detach: not closed"));
894 
895 	INP_INFO_WLOCK(&V_ripcbinfo);
896 	INP_WLOCK(inp);
897 	rip_delhash(inp);
898 	if (so == V_ip_mrouter && ip_mrouter_done)
899 		ip_mrouter_done();
900 	if (ip_rsvp_force_done)
901 		ip_rsvp_force_done(so);
902 	if (so == V_ip_rsvpd)
903 		ip_rsvp_done();
904 	in_pcbdetach(inp);
905 	in_pcbfree(inp);
906 	INP_INFO_WUNLOCK(&V_ripcbinfo);
907 }
908 
909 static void
rip_dodisconnect(struct socket * so,struct inpcb * inp)910 rip_dodisconnect(struct socket *so, struct inpcb *inp)
911 {
912 	struct inpcbinfo *pcbinfo;
913 
914 	pcbinfo = inp->inp_pcbinfo;
915 	INP_INFO_WLOCK(pcbinfo);
916 	INP_WLOCK(inp);
917 	rip_delhash(inp);
918 	inp->inp_faddr.s_addr = INADDR_ANY;
919 	rip_inshash(inp);
920 	SOCK_LOCK(so);
921 	so->so_state &= ~SS_ISCONNECTED;
922 	SOCK_UNLOCK(so);
923 	INP_WUNLOCK(inp);
924 	INP_INFO_WUNLOCK(pcbinfo);
925 }
926 
927 static void
rip_abort(struct socket * so)928 rip_abort(struct socket *so)
929 {
930 	struct inpcb *inp;
931 
932 	inp = sotoinpcb(so);
933 	KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
934 
935 	rip_dodisconnect(so, inp);
936 }
937 
938 static void
rip_close(struct socket * so)939 rip_close(struct socket *so)
940 {
941 	struct inpcb *inp;
942 
943 	inp = sotoinpcb(so);
944 	KASSERT(inp != NULL, ("rip_close: inp == NULL"));
945 
946 	rip_dodisconnect(so, inp);
947 }
948 
949 static int
rip_disconnect(struct socket * so)950 rip_disconnect(struct socket *so)
951 {
952 	struct inpcb *inp;
953 
954 	if ((so->so_state & SS_ISCONNECTED) == 0)
955 		return (ENOTCONN);
956 
957 	inp = sotoinpcb(so);
958 	KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
959 
960 	rip_dodisconnect(so, inp);
961 	return (0);
962 }
963 
964 static int
rip_bind(struct socket * so,struct sockaddr * nam,struct thread * td)965 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
966 {
967 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
968 	struct inpcb *inp;
969 	int error;
970 
971 	if (nam->sa_len != sizeof(*addr))
972 		return (EINVAL);
973 
974 	error = prison_check_ip4(td->td_ucred, &addr->sin_addr);
975 	if (error != 0)
976 		return (error);
977 
978 	inp = sotoinpcb(so);
979 	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
980 
981 	if (CK_STAILQ_EMPTY(&V_ifnet) ||
982 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
983 	    (addr->sin_addr.s_addr &&
984 	     (inp->inp_flags & INP_BINDANY) == 0 &&
985 	     ifa_ifwithaddr_check((struct sockaddr *)addr) == 0))
986 		return (EADDRNOTAVAIL);
987 
988 	INP_INFO_WLOCK(&V_ripcbinfo);
989 	INP_WLOCK(inp);
990 	rip_delhash(inp);
991 	inp->inp_laddr = addr->sin_addr;
992 	rip_inshash(inp);
993 	INP_WUNLOCK(inp);
994 	INP_INFO_WUNLOCK(&V_ripcbinfo);
995 	return (0);
996 }
997 
998 static int
rip_connect(struct socket * so,struct sockaddr * nam,struct thread * td)999 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1000 {
1001 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
1002 	struct inpcb *inp;
1003 
1004 	if (nam->sa_len != sizeof(*addr))
1005 		return (EINVAL);
1006 	if (CK_STAILQ_EMPTY(&V_ifnet))
1007 		return (EADDRNOTAVAIL);
1008 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
1009 		return (EAFNOSUPPORT);
1010 
1011 	inp = sotoinpcb(so);
1012 	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
1013 
1014 	INP_INFO_WLOCK(&V_ripcbinfo);
1015 	INP_WLOCK(inp);
1016 	rip_delhash(inp);
1017 	inp->inp_faddr = addr->sin_addr;
1018 	rip_inshash(inp);
1019 	soisconnected(so);
1020 	INP_WUNLOCK(inp);
1021 	INP_INFO_WUNLOCK(&V_ripcbinfo);
1022 	return (0);
1023 }
1024 
1025 static int
rip_shutdown(struct socket * so)1026 rip_shutdown(struct socket *so)
1027 {
1028 	struct inpcb *inp;
1029 
1030 	inp = sotoinpcb(so);
1031 	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
1032 
1033 	INP_WLOCK(inp);
1034 	socantsendmore(so);
1035 	INP_WUNLOCK(inp);
1036 	return (0);
1037 }
1038 
1039 static int
rip_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct thread * td)1040 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
1041     struct mbuf *control, struct thread *td)
1042 {
1043 	struct inpcb *inp;
1044 	u_long dst;
1045 
1046 	inp = sotoinpcb(so);
1047 	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
1048 
1049 	/*
1050 	 * Note: 'dst' reads below are unlocked.
1051 	 */
1052 	if (so->so_state & SS_ISCONNECTED) {
1053 		if (nam) {
1054 			m_freem(m);
1055 			return (EISCONN);
1056 		}
1057 		dst = inp->inp_faddr.s_addr;	/* Unlocked read. */
1058 	} else {
1059 		if (nam == NULL) {
1060 			m_freem(m);
1061 			return (ENOTCONN);
1062 		}
1063 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
1064 	}
1065 	return (rip_output(m, so, dst));
1066 }
1067 #endif /* INET */
1068 
1069 static int
rip_pcblist(SYSCTL_HANDLER_ARGS)1070 rip_pcblist(SYSCTL_HANDLER_ARGS)
1071 {
1072 	int error, i, n;
1073 	struct inpcb *inp, **inp_list;
1074 	inp_gen_t gencnt;
1075 	struct xinpgen xig;
1076 	struct epoch_tracker et;
1077 
1078 	/*
1079 	 * The process of preparing the TCB list is too time-consuming and
1080 	 * resource-intensive to repeat twice on every request.
1081 	 */
1082 	if (req->oldptr == 0) {
1083 		n = V_ripcbinfo.ipi_count;
1084 		n += imax(n / 8, 10);
1085 		req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
1086 		return (0);
1087 	}
1088 
1089 	if (req->newptr != 0)
1090 		return (EPERM);
1091 
1092 	/*
1093 	 * OK, now we're committed to doing something.
1094 	 */
1095 	INP_INFO_WLOCK(&V_ripcbinfo);
1096 	gencnt = V_ripcbinfo.ipi_gencnt;
1097 	n = V_ripcbinfo.ipi_count;
1098 	INP_INFO_WUNLOCK(&V_ripcbinfo);
1099 
1100 	bzero(&xig, sizeof(xig));
1101 	xig.xig_len = sizeof xig;
1102 	xig.xig_count = n;
1103 	xig.xig_gen = gencnt;
1104 	xig.xig_sogen = so_gencnt;
1105 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1106 	if (error)
1107 		return (error);
1108 
1109 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
1110 
1111 	INP_INFO_RLOCK_ET(&V_ripcbinfo, et);
1112 	for (inp = CK_LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n;
1113 	     inp = CK_LIST_NEXT(inp, inp_list)) {
1114 		INP_WLOCK(inp);
1115 		if (inp->inp_gencnt <= gencnt &&
1116 		    cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
1117 			in_pcbref(inp);
1118 			inp_list[i++] = inp;
1119 		}
1120 		INP_WUNLOCK(inp);
1121 	}
1122 	INP_INFO_RUNLOCK_ET(&V_ripcbinfo, et);
1123 	n = i;
1124 
1125 	error = 0;
1126 	for (i = 0; i < n; i++) {
1127 		inp = inp_list[i];
1128 		INP_RLOCK(inp);
1129 		if (inp->inp_gencnt <= gencnt) {
1130 			struct xinpcb xi;
1131 
1132 			in_pcbtoxinpcb(inp, &xi);
1133 			INP_RUNLOCK(inp);
1134 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1135 		} else
1136 			INP_RUNLOCK(inp);
1137 	}
1138 	INP_INFO_WLOCK(&V_ripcbinfo);
1139 	for (i = 0; i < n; i++) {
1140 		inp = inp_list[i];
1141 		INP_RLOCK(inp);
1142 		if (!in_pcbrele_rlocked(inp))
1143 			INP_RUNLOCK(inp);
1144 	}
1145 	INP_INFO_WUNLOCK(&V_ripcbinfo);
1146 
1147 	if (!error) {
1148 		struct epoch_tracker et;
1149 		/*
1150 		 * Give the user an updated idea of our state.  If the
1151 		 * generation differs from what we told her before, she knows
1152 		 * that something happened while we were processing this
1153 		 * request, and it might be necessary to retry.
1154 		 */
1155 		INP_INFO_RLOCK_ET(&V_ripcbinfo, et);
1156 		xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1157 		xig.xig_sogen = so_gencnt;
1158 		xig.xig_count = V_ripcbinfo.ipi_count;
1159 		INP_INFO_RUNLOCK_ET(&V_ripcbinfo, et);
1160 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1161 	}
1162 	free(inp_list, M_TEMP);
1163 	return (error);
1164 }
1165 
1166 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist,
1167     CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0,
1168     rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
1169 
1170 #ifdef INET
1171 struct pr_usrreqs rip_usrreqs = {
1172 	.pru_abort =		rip_abort,
1173 	.pru_attach =		rip_attach,
1174 	.pru_bind =		rip_bind,
1175 	.pru_connect =		rip_connect,
1176 	.pru_control =		in_control,
1177 	.pru_detach =		rip_detach,
1178 	.pru_disconnect =	rip_disconnect,
1179 	.pru_peeraddr =		in_getpeeraddr,
1180 	.pru_send =		rip_send,
1181 	.pru_shutdown =		rip_shutdown,
1182 	.pru_sockaddr =		in_getsockaddr,
1183 	.pru_sosetlabel =	in_pcbsosetlabel,
1184 	.pru_close =		rip_close,
1185 };
1186 #endif /* INET */
1187